Ref. Ares(2018)2554324 - 16/05/2018
Ref. Ares(2018)2751838 - 28/05/2018
Review of the likely effects of Google’s proposed Commitments dated October 21, 2013
(“Second Commitments”)
December 9, 2013
Professors David J. Franklyn ("Franklyn") and David A. Hyman ("Hyman")1
1. Introduction
In May, 2012, the European Commission published its preliminary conclusion that Google limits
choice in Internet search by (among other things) giving Google’s vertical search services
(including Google Shopping and Google Hotels) more prominent page placement and links than
it gives to its rivals in these areas. On April 25, 2013, in its first set of proposed voluntary
commitments (hereinafter Google’s “First Commitments”), Google proposed to resolve these
concerns by providing more prominent placement for selected competitors’ websites (“three rival
links,”) and by adding a label and disclosure that would indicate which services are and are not
provided by Google.2 The stated purpose of these proposals is to give consumers an effective
choice among vertical search services offered by Google and its rivals.
We submitted an expert report in this matter on July 1, 2013, in which we analyzed the impact of
the First Commitments. Based on market studies that we conducted in the U.K., we found that
the First Commitments, if enacted, would not address the Commission’s stated concerns and
materially restore competition in the vertical search markets at issue. The European Commission
did not accept the First Commitments.
On October 21, 2013, Google submitted a second set of revised commitments (hereinafter
Google’s “Second Commitments”) to the European Commission. On October 28, 2013,
Google’s Second Commitments were released to various parties, including FairSearch Europe,
for comment. The modifications include, but are not limited to, certain enlargements of the three
rival links box, changes in the display of the three rival links for mobile devices, and changes to
the language of the proposed disclosure pop-up box that is supposed to enable consumers to
clearly distinguish between paid and unpaid links, and between Google’s specialized (vertical)
search results (including Google Shopping and Google Hotels) and Google’s generic (unpaid)
search results.
We were retained by Clifford Chance, advisors to FairSearch Europe (hereinafter “Fairsearch”),
to conduct market tests on the proposed changes, and to render opinions as to the likely impact of
Google's Second Commitments proposals on Internet users/shoppers. Specifically, we were
1
Experts’ background and qualifications are detailed in an Appendix. A copy of Franklyn's Curriculum
Vitae is attached as Exhibit A. A copy of Hyman’s Curriculum Vitae is attached as Exhibit B. Franklyn and
Hyman have written two law review articles about consumer perceptions and the Internet, including an article
published in Spring, 2013 in the Harvard Journal of Law and Technology, and a forthcoming article on search bias
and search neutrality. Copies of these articles are attached as Exhibits C & D.
2
See April 25, 2013 Google Commitment Proposal.
asked to test the visibility and consumer understanding of Google's Second Commitments
proposals.
We designed and conducted three online surveys in the United Kingdom that measured the
likelihood that consumers would click on any of Google’s Second Commitments three rival
links. We tested the likelihood of consumer click through on images similar to those that would
appear on non-mobile and on mobile devices. If consumers do not click on the three rival links,
then Google’s proposed remedy would not be meaningfully visible to consumers or likely to
reflect the natural choices of consumers--and therefore would not be likely to restore competition
in these areas, and maximize consumer welfare.
The survey also measured whether respondents recognized and understood Google’s modified
label and disclosure. If consumers do not recognize and understand Google’s label and
disclosure, then they would be even less likely to exercise a meaningful choice between
Google’s vertical search services and the same services offered by Google’s competitors.
As explained more fully below, we found Google’s Second Commitments are not likely to
materially increase or restore consumer choice or competition in the vertical search markets at
issue. Google’s Second Commitments do not achieve the Commission’s stated objective of
materially increasing the visibility of rivals and the awareness of those rivals by consumers.
Specifically, we found that:
A. Google is the first stop for the overwhelming majority of users that employ a search
engine to locate products and hotels.
B. The “three rival links” gather only a modest number of consumer clicks, both in absolute
terms and relative to clicks on Google Shopping and Google Hotels.
C. Incremental increases in consumer attention that may accrue under the Second
Commitments are unstable and easily destroyed by changes to Google’s search results
page.
D. Presenting rival links in a competitively neutral fashion is a relatively straightforward
matter. Parity of rival link presentation improves consumer welfare.
E. Consumer confusion persists as to the difference between Google’s vertical search results
and other search results. Instead of reducing consumer confusion about search results,
Google’s proposed disclosure/explanation materially increases confusion.
II. Survey Methodology
Between November 12 and November 20, 2013, we completed three on-line surveys of a total of
approximately 3,500 respondents residing in the United Kingdom.3 More specifically, using a
series of simulated searches, the surveys examined:
3
The survey was administered by Survey Sampling International (SSI) to a demographically representative
sample of Google users in the United Kingdom. We excluded from the analysis surveys that were not completed
from an IP address in the UK.
2
• Whether the Second Commitments resulted in a material number of clicks on any
of the proposed rival links;
• The effect of other modification to search page layout and labeling on click
patterns;
• The effectiveness of the disclosure accompanying the rival links in
communicating basic information.
To examine these issues, we designed a series of multi-part interactive Internet surveys. The
surveys are attached as Addendum A. The results presented in this report are based on
respondents who completed the entire survey (with no missing results), and took more than 5
minutes to do so. We were left with a total sample of approximately 2,500 completed, useable
respondents, fairly evenly divided across all three surveys.4 For the combined results, the
margin of error on the figures we report is +2% for a 95% confidence interval. For any given
survey, the margin of error is roughly +3.5% for a 95% confidence interval.
Each survey began with two questions designed to measure participants’ general online search
behavior. In Survey 1, we then asked respondents to run two sets of searches – one for an Apple
iPod and the other for a Nikon camera. In Survey 2, we asked respondents to run two sets of
searches – one for an Apple iPod and the other for a Razer headset. In Survey 3, we asked
respondents to search for a hotel in Madrid.
We chose these three searches in order to test three hypothetical sample product searches (here
iPod, Nikon camera, and Razer), and one hypothetical locational search (here a search for a hotel
in Madrid). Respondents were asked to click on the link they would have selected if they were
trying to buy an iPod, Nikon camera, or Razer headset, or trying to find a hotel in Madrid. For
each survey, respondents first viewed a “native” page of search results (
i.e., the search results
that would have appeared if an actual Google search had been run without the proposed
Commitments being implemented) followed by a series of modified search results, reflecting
Google’s proposed Commitments, and several variations we developed.
In order to advance through the survey, respondents had to repeatedly run the search query, and
then click through on whatever link they selected. Respondents were not allowed to go back to
an earlier question after progressing past it. Because respondents had to click on something to
advance, our results represent an upper-bound on clicks that would occur in a real-world search
environment, where users can simply close the page without clicking on anything.
After viewing the Google native page for each of these searches, respondents saw search results
pages incorporating the three rival links specified in the Second Commitments. For the three
rival links for the products search, we chose well-known companies that would qualify for
inclusion under the terms of Google’s Second Commitments; namely, Kelkoo, Pricerunner and
Bizrate. These are the same three rival links we used in our July, 2013 market test of Google’s
First Commitments.
4
See Addendum B, Table 1. We separately analyzed the results for the excluded surveys, and found
virtually identical results. Thus, our exclusion criteria do not affect our findings.
3
To validate our findings, we also tested the mobile version of rival links specified in the Second
Commitments, along with multiple variations in which we modified the page layout and labeling
of the search results page.5 Respondents were presented with these variations after re-running
various Google searches. Respondents were asked to click on the site where they were most
likely to find what they were looking for,
i.e., an iPod, Nikon camera, Razer headset or a hotel in
Madrid.
After completing these tasks, we showed respondents a series of screen shots of actual search
results and asked them a variety of questions to determine whether they could tell the difference
between the results in various regions of a typical search results page – that is, if they could
correctly differentiate between paid and unpaid regions of the page and also identify how certain
type of results were generated. We also asked respondents to view Google’s proposed
disclosure/explanation regarding its specialized vertical search results and asked respondents a
series of questions to test their understanding of the proposed disclosure/explanation. We
concluded with a series of questions about the demographics of those completing the survey.
III. Findings
A. Google is the first stop for the overwhelming majority of users that employ a search
engine to locate products and hotels
We asked respondents how they would go about locating a specific product (iPod) or a hotel in a
city they didn’t know. Of those who responded that they used a search engine to perform these
tasks, Google accounted for 95% of respondents searching for an iPod, and 94% of those
searching for a hotel in a city they didn’t know. We obtained comparable results when we asked
a similar question previously (compare Tables 2A and 2B in our initial report).
B. The three rival links in the Second Commitments gather only a modest number of
consumer clicks, both in absolute terms, and relative to clicks on Google Shopping
and Google Hotels.
We tested the three rival links proposed in the Second Commitments.6 Table 1A presents the
results for Products, and Table 1B presents the results for Hotels.
As Table 1A shows, in a product search, the three rival links garnered 0.9%-2.9% of total clicks,
compared to 36.7% of total clicks on Google Shopping.7
5
Our goal in testing these variations was to provide a baseline for assessing the impact of layout and
labeling on consumer search behavior. Because our primary task in this engagement was to test the proposals in the
Second Commitments, we focus on them in this report. Copies of the images that are referenced in this report are
attached as Addendum C.
6
These variations are based on the images contained in Google’s Second Commitments.
7
By “Google Shopping” we are referring to all clicks in the specified region, and not simply the blue link at
the top of the region, “Shop for ipod on Google.”
The Second Commitments also authorize the shading of the region containing the rival links. We
4
Table 1A: Click Rates on Second Commitments Proposal – Non-mobile Product Searches
Type of Link
Pricerunner
2.9%
Rival Links
Kelkoo
1.2%
Bizrate
0.9%
Google Shopping
36.7%
All Vertical Search Options
41.8%
All Other Link Types
58.2%
Total
100%
Pricerunner
7.0%
Share of All Vertical
Kelkoo
3.0%
Search Options
Bizrate
2.3%
Google Shopping
87.7%
For ease of comparison, the bottom four rows in Table 1A (“Share of All Vertical Search
Options”) compute the percentage of clicks on each of the three rival links and Google Shopping
compared to the total clicks on all vertical search options on the search results page.8 This figure
is effectively the “market share” for each of the specified rival links and for Google Shopping if
we assume the Second Commitments are implemented. As these rows indicate, in a product
search, any given rival link secured 2.3%-7% of vertical search clicks, compared to 87.7% of
vertical search clicks for Google Shopping.
separately tested a shaded region, and found a modest increase in the absolute number of clicks on the three rival
links (from 5% to 7.7%), and a reduction in the number of clicks on Google Shopping (from 36.7% to 28.4%). In
combination, this meant that Google Shopping’s share of all vertical search options declined slightly, from 87.7% to
78.7%. Further testing will be necessary to determine whether this finding is robust.
8
We define clicks on all vertical search options as the sum of clicks on Google Shopping, Pricerunner,
Kelkoo and Bizrate.
5
Table 1B presents similar information for our hotel search.
Table 1B: Click Rates on Second Commitments Proposal – Non-mobile Hotel Searches
Type of Link
Expedia
3.8%
Rival Links
Priceline
0.9%
Trivago
2.3%
Google Hotels + Places
28.1%
All Vertical Search Options
35.1%
All Other Link Types
64.9%
Total
100%
Expedia
10.9%
Share of All Vertical Search
Priceline
2.5%
Options
Trivago
6.7%
Google Hotels + Places
79.9%
Table 1B shows that in a hotel search, the three rival links garnered 0.9%-3.8% of total clicks,
compared to 28.1% of total clicks on Google’s vertical search options (Google Hotels + Google
Places.9 As in Table 1A, the bottom four rows in Table 1B compute the percentage of clicks on
each of the three rival links compared to the total clicks on all vertical search options on the
search results page.10 As these rows indicate, in a hotel search, any given rival link secured
2.5%-10.9% of vertical search clicks, compared to 79.9% of vertical search clicks on Google’s
vertical search options.11
Table 2A presents comparable results for mobile search images for Products, and Table 2B
presents the same information for Hotels. We tested two versions: the “Plain” version (which
has the text “Other Sites” in the lower right hand corner of the Shopping and Hotels boxes), and
a “Pop-up” version (which adds to the Plain version a large notification balloon that contains the
text “Tap here for additional product (or hotel) search services,” along with a blue button that has
the word “OK” in it).12
9
By “Google Hotels,” we are referring to all clicks in the specified region, and not simply the blue link at
the top of the region, “Hotels in Madrid, Spain on Google.”
10
We define clicks on vertical search options as the sum of clicks on Google Hotels, Google Places, Expedia,
Priceline, and Trivago.
11
The Second Commitments also authorize the shading of the region containing the rival links. We
separately tested a shaded region, and found a modest increase in the absolute number of clicks on the three rival
links (from 7% to 9.3%), and a slight reduction in the number of clicks on Google Hotels & Places (from 28.1% to
27.6%). In combination, this meant that Google Hotels & Places share of all vertical search options declined
slightly, from 79.9% to 74.8%. Further testing will be necessary to determine whether this finding is robust.
12
These variations are based on the images contained in Google’s Second Commitments.
6
Table 2A: Click Rates on Second Commitments Proposal – Mobile Product Searches
Commitments Variation
Plain
Pop-up
Type of Link
“Other Sites” Link
0.3%
0.2%
Google Shopping
43.3%
33.8%
All Vertical Search Options
43.6%
34%
All Other Link Types
56.4%
66%
Total
100%
100%
Share of All Vertical “Other Sites” Links
0.7%
0.6%
Search Options
Google Shopping
99.3%
99.4%
Table 2B: Click Rates on Second Commitments Proposal – Mobile Hotel Searches
Commitments Variation
Plain
Pop-up
Type of Link
“Other Sites” Links
0.0%
0.1%
Google Hotels + Places
29.4%
29.5%
All Vertical Search Options (Including Places)
29.4%
29.6%
All Other Link Types (Excluding Places)
70.6%
70.4%
Total
100%
100%
Share of All Vertical
“Other Sites” Links
0.0%
0.4%
Search Options
Google Hotels + Places 100.0%
99.6%
As Tables 2A and 2B reflect, when we test the Mobile version of the Second Commitments,
there are effectively no clicks on the “Other Sites” links in any of the variations. For ease of
comparison, the bottom two rows in Tables 2A & 2B (“Share of All Vertical Search Options”)
compute the percentage of clicks on the “Other Sites” links and Google’s vertical search options
compared to the total clicks on all vertical search options on the search results page.13 As this
row indicates, in the Mobile setting, “Other Sites” links account for 0.6%-0.7% of clicks for
products (Table 2A), and 0.0%-0.3% of clicks for hotels (Table 2B). (We treat the actual pop-up
notification balloon as an “Other Link Type) in both Tables 2A and 2B, since it only appears
once, and is not, in fact, a vertical search link/option.
One further complication: in mobile applications, a user must first click on the “Other sites” link,
and then select a particular rival link on which to click. We test only respondents’ initial click
through, so our results reflect the maximum percentage of clicks on all rival links combined –
and that only if we (implausibly) assume that no users click back or simply close the search
results page.
Tables 1A & 1B demonstrate that the percentage of non-mobile clicks on rival links is only
13
We define clicks on all vertical search options on product search results pages as the sum of clicks on
Google Shopping, Pricerunner, Kelkoo and Bizrate. We define clicks on vertical search options on hotel search
results pages as the sum of clicks on Google Hotels, Google Places, Expedia, Priceline, and Trivago.
7
modestly higher than in our market test of Google’s First Commitments (there, it was never more
than 2%). However, click through rates for rival links in the Second Commitments continue to
lag far behind click through rates for Google’s own vertical search results. And, in the mobile
context (Tables 2A & 2B), none of the versions we tested attracted more than a trivial number of
clicks, even when the pop-up notification balloon was present, thereby calling attention to the
“Other sites” link.14
C. Incremental increases in consumer attention that may accrue from the proposed
rival links are unstable and easily destroyed by changes to Google’s search results
page.
We found that any incremental increases in consumer attention that might arise as a result of
enhanced presentation of rival links in the Second Commitments is subject to dissipation if
Google were to introduce new search features, specifically a search feature that it has already
introduced elsewhere (e.g., in the United States). This search feature is allowable under the
Second Commitments.
Specifically, we presented respondents with a page of search results that contained the following
banner-strip at the top of the page of search results that included the three rival links as proposed
in the Commitments:
14
In our test of Google’s First Commitments, we tested the proposed text for mobile uses, but we used a full-
size screen image. In our test of Google’s Second Commitments for mobile uses, we used an image that looked like
a mobile screen. We found comparable click through rates in the earlier report (0.1%-0.8%) versus our findings in
this report (0.0%-0.3% for the plain images for Product and Hotel searches). The first report did not contain a pop-
up image, so direct comparison of the percentage of clicks on the “Other sites” link for that image is not possible.
8
Once we added this banner-strip (which we treated as part of the Google Shopping/Hotels
region), click-through rates on shaded rival links on Products searches dropped from 5.1% to
3.4% for products, and from 7% to 4.9% for hotels. Click through rates on the Google Shopping
and Google Hotels + Places regions also increased dramatically. In combination, this meant the
share of clicks on rival links dropped from 12.3% to 6.5% for products, and from 20.1% to
10.4% for hotels. Thus, in both relative and absolute terms, the introduction of the banner-strip
largely negated the additional click-through effect of the Second Commitments (as compared to
the First Commitments). In unreported results, we were able to reduce the absolute and relative
share of clicks on the rival links further still, by adding a floating pop-up that said “Search on
Google Shopping” (or “Search on Google Travel”) to the search results page with the banner-
strip. The floating pop-up is also allowable under the Second Commitments.
D. Presenting rival links in a competitively neutral fashion is a relatively
straightforward matter. Parity of Rival Link Presentation Improves Consumer
Welfare
We also tested two variations that we created, to determine how difficult it would be to present
search results in a way that would substantially increase consumer attention and click-throughs
on rival sites.
In one alternative, we presented respondents with a page of search results that eliminated the
Google Shopping or Google Hotels boxes (along with the proposed Commitments three rival
9
links box) – and replaced those boxes with a single shaded box. The box was headed “Compare
prices on,” and had icons for four vertical search offerings. The new box contained only icons
and names of the vertical search offerings, and some abbreviated text describing the products
that were available. Thus, the four vertical search offerings were placed on an even playing field
(including equal visual footing) with one another. The image below shows an example of the
box from the Razer product search, but we tested this approach in all three surveys.
We presented respondents with two versions of this box: one in which Google appeared in the
box as the left-most icon/vertical search offering, and one in which Google appeared in the box
as the right-most icon/vertical search offering. This allows us to more effectively isolate the
extent to which observed click through patterns are attributable to various factors (including
search results page architecture, visually rich appearance of Google Shopping and Google
Hotels, and the strength of the individual brands).
Using this approach with a product search, we found that respondents clicked on the
individual rival links 3.1% to 9.2% of the time, and clicked on Google Shopping 10.6% of the
time when it was the left-most icon/vertical search offering, and 4.3% of the time when it was
the right-most icon/vertical search offering.15 With a hotel search, we found that respondents
clicked on the individual rival links 1.0% to 8.4% of the time, and clicked on Google Icon 6.4%
of the time when it was the left-most icon/vertical search offering, and 3.2% of the time when it
was the right-most icon/vertical search offering.16 Thus, when we create parity of appearance
and placement, we obtain substantially higher click through rates in both relative and absolute
terms for the rival links, reflecting the greater visibility of the rival links compared to the status
quo.17 And, click through rates on Google Shopping and Google Hotels drop dramatically,
indicating that position on the page and visually rich appearance is an important component in
15 If we look at click-through rates for each of the listed icons, we find Pricerunner received 9.2% of clicks when it
was the left-most image, and 6.1% when it was the 2nd image from the left. Kelkoo received 3.1% of clicks when it
was 2nd image from the left, and 3.7% when it was the 2nd image from the right. Bizrate received 4.7% of clicks
when it was the 3rd image from the left, and 3.5% when it was the right-most image.
16 If we look at click-through rates for each of the listed icons, we find Expedia received 8.4% of clicks when it was
the left-most image, and 5.7% when it was the 2nd image from the left. Priceline received 2.7% of clicks when it
was 2nd image from the left, and 1.2% when it was the 2nd image from the right. Trivago received 2.0% of clicks
when it was the 3rd image from the left, and 1.0% when it was the right-most image.
17 Viewed collectively, the three rival links/icons secured roughly 32% of clicks on vertical search options when
Google was the left-most icon, and 53% of clicks on vertical search options when was Google was the right-most
icon. The results were virtually identical for product searches and hotel searches. These “market share” figures are
substantially greater than those we observe when we test Google’s Second Commitments.
Compare Tables 1A &
1B
.
10
the dominance of Google Shopping and Google Hotels. Finally, the fact that click through rates
on Google Shopping and Google Hotels drop dramatically when it is in the right-most position
(compared to the left-most position) further confirm that position on the search results page, (i.e.,
the architecture of the search results page, which gives Google’s vertical search offerings greater
visibility and presence than competitors) rather than simply the strength of the brands of the
competing vertical search providers, is responsible for an appreciable proportion of observed
behavior.
We found a similar effect when we tested respondents’ responses to an alternative variation,
where we split the Google Shopping box into two equal-sized and equal-looking boxes. One box
was labeled Google Shopping Results and the other box was labeled “Compare results on
Kelkoo.” Both boxes contained the word “Sponsored,” and included two images of the products
in question, along with some text. This alternative variation is illustrated below:
We tested this variation with Google Shopping on the right and on the left. We found that parity
of presentation resulted in an increase in clicks on Kelkoo when it occupied the left-side box
(increasing to 12.2%), compared to the right-side box (6.9%). Clicks on Google Shopping were
23.2% when it occupied the left-side box, and 16.3% when it occupied the right-side box. Thus,
both Kelkoo and Google experienced significant increases in clicks when they were on the left
side than when they were on the right side -- even though everything else remained the same.
These findings show that click through rates on Google’s specialized search results are driven in
part by the prominent position on the search results page and the more visually attractive (larger
and photo-inclusive) presentation of Google’s own specialized search results.18
We find the highest degree of rival link visibility and resulting consumer attention--and rough
parity in click through rates --when rival links are displayed in a manner that is comparable to
the manner in which Google’s own specialized search results are displayed. The First and
Second Commitments do not achieve this objective. Our findings provide some insight into the
factors that affect click through rates on vertical search offerings.
18 Unsurprisingly, we also observe a brand effect. Google has higher absolute clicks than Kelkoo, irrespective of
position. But, the more significant issue is that visibility makes a difference, even after we control for brand effects.
Google is as strong a brand when it is on the left side as when it is on the right side. But, Google’s share of
combined clicks on vertical search options is 77% when it is on the left side, and only 57% when it is on the right
side.
11
E. Consumer confusion persists as to the difference between Google’s specialized
search results and other search results – and the disclosure actually increases
confusion
The Second Commitments provide that searchers who roll-over an icon that currently appears
next to the word “Sponsored” at the top right of Google’s vertical will see a pop-up disclosure
statement. The statement is intended to clarify the origin and type of search results they are
seeing.19 In unreported results, we asked respondents whether they had ever clicked on the icon
that appeared next to the word “Sponsored” to see what would pop-up. Only 12%-17% of
respondents responded that they had ever done so.
We also presented all respondents with the pop-up disclosure statement, and asked whether the
statement improved their understanding of why content appeared on the search results page,
made them more confused, or had no effect. Table 3 presents the results.
Table 3: Impact of Google's Proposed Pop-up Disclosure on Understanding of Why
Content Appears
Type of Search
Product
Hotel
Improved understanding
27%
28%
More Confused
43%
44%
No Effect on understanding
30%
27%
Table 3 shows that of those who actually read the disclosure statement, only 27%-28% thought it
improved their understanding of why content appeared on the search results page.
We also showed respondents images that reflected the impact of the Second
Commitments on the search results page (similar to those we used to generate the results in Table
1). But, these search results pages had each section of the webpage labeled, as “Region A,”
“Region B,” and so on. We then tested whether respondents were able to correctly explain the
reason why links appeared in particular regions, such as Google Shopping, the Rival Links Box,
and Algorithmic content. We provided seven possible explanations: unpaid algorithmic content;
links for which the listed companies pay to appear; price comparison service offered by Google;
price comparison service offered by independent third parties; a fanciful control (unpaid links
selected by Google’s Special Marketing Team); Other; and Don’t Know/Not Sure.20
19
The specific disclosure text reads as follows: “This link is inserted to show more results from Google’s
Shopping results. For shopping results from other relevant providers, click on the relevant links to other search sites
below or see Google’s other search results. Ads are based on your current search term. Visit Google’s Ads
Preferences Manager to learn more, block specific advertisers, or opt out of personalized ads.”
20
Because of the nature of the search results page, more than one explanation might be correct for some of
the regions/links. For example, a respondent could correctly answer that links in the Google Shopping region were
there because Google had been paid for the link to appear, or because it was a price comparison service offered by
Google. Similarly, for the rival links box, a respondent could correctly answer that links in that region were there
because Google had been paid for the link to appear, or because it was a price comparison service offered by
independent 3rd parties. For these regions, we treated both of the listed answers as correct.
12
We found little evidence that respondents effectively differentiated paid from unpaid
content, or understood that Google Shopping was offering vertical search results. Only 51% of
respondents correctly responded that the content in Google Shopping represented either paid
content or a price comparison service offered by Google, with responses split evenly among
these two choices. This means that 48% of respondents did not understand why links appeared
in the Google Shopping box – with responses split among those who thought it was unpaid
algorithmic content (14%); those who picked our fanciful Google Special Marketing Team
control (18%); those who thought it was a price comparison service offered by independent 3rd
parties (6%); and those who didn't know, weren’t sure, or picked other (11%). We obtained
similarly mixed results when we asked respondents why algorithmic links appeared (only 41.5%
correctly responded that these links were selected by Google’s computer formula and were
unpaid), and why the rival links appeared (only 51% correctly responded that these links were
either paid content or a price comparison service offered by independent third parties).
Finally, we evaluated the impact of the disclosure statement on respondents’ understanding of
why links appeared in the Google Shopping region. We did this by asking them the same
question twice – once before and then again after they had reviewed the disclosure statement.
We find evidence that the disclosure statement actually reduced respondents’ understanding of
the issue. In the first set of surveys, the percentage of respondents answering that the links that
appeared in Google Shopping were the result of a price comparison service offered by Google
declined from 26.8% to 14.4%. The percentage of respondents who didn’t know or were unsure
increased from 10.1% to 17.2%, and the percentage that selected our fanciful control (i.e., links
selected by the Special Marketing Team) increased from 17.8% to 22.1%.
Stated differently, after being exposed to the disclosure statement, the percentage of respondents
selecting a correct answer declined from 51.5% to 40.5%. We obtain comparable results for the
Hotels survey; after being exposed to the disclosure statement, the percentage of respondents
selecting a correct answer declined from 50.4% to 41.8%.
Appendix B includes more detailed information on all of these results. Addendum D contains
information on respondent demographics, including self-reported knowledge of search page
layout and labeling.
Summary/Conclusions:
• The proposed three rival links do not materially increase consumer awareness or the
visibility of the proposed links to large majorities of consumers.
• Large majorities of consumers do not understand the origin of the rival links or their
relationship to Google's other links.
• We found only slightly higher click-through rates for the Second Commitment proposals
than we found for the First Commitment proposals. We do not believe that the Second
Commitments are likely to command materially increased consumer attention or restore
competition for rivals that seek to compete with either Google Shopping or Google
13
Hotels and Places.
• We found that the proposal rival link remedy is unstable. Modest changes to the
appearance of the search results page, such as the addition of a banner-strip similar to the
one currently being used in the U.S., eliminate the observed slight increases in visibility
and prominence (as measured by the click-through rates on rival links).
• Google Shopping commands a high degree of consumer clicks and attention in part
because of its location on the search results page, and the visually-rich display (e.g.,
pictures and mini-boxes within a box). Google Hotels and Places benefit from similar
treatment. Confirming our earlier work, we find that page layout and design is more
important than search result labeling in determining visibility and click-through rates.
• We find evidence that consumer preferences reflect both the strength of the underlying
brands, as well as the way in which those brands are presented on the search results page.
Absent parity of presentation, an appreciable number of consumers will be diverted from
their preferred destination, which will reduce consumer welfare. Conversely, consumer
welfare is enhanced when artificial impediments to consumer choice are removed and
rival sites are positioned in a manner that is visibly equivalent to the positioning and
layout of Google’s own specialized search results.
• Modifications that do not materially affect page layout are unlikely to change preexisting
competitive dynamics in search. Modifications in page layout may or may not have an
impact, depending on their design attributes. Stated differently, we find the highest
degree of consumer attention and competition for clicks when rival links are displayed in
a manner that is comparable to the manner in which Google Shopping/Hotels results are
displayed.
• We find high levels of consumer confusion as to the source and type of search results and
a low level of ability to correctly differentiate between paid and unpaid results, or
between Google’s vertical search and generic search results.
• We find that the proposed Second Commitment disclosure statement does not effectively
communicate the necessary information (i.e., whether the region in question is paid v.
unpaid, and the location and significance of the “rival links.”) Indeed, the proposed
language confuses many respondents.
In summary, we find that the proposals offered by Google in the Second Commitments are
unlikely to materially increase consumer attention, offer consumers meaningful choices or
restore competition (as measured by click rates) in this space. By testing various alternatives, we
identified several factors that are more likely to draw consumer attention to rival links, offer
consumers meaningful choices and thereby improve competition (again, as measured by click
rates) in this space. We also found that Google’s proposed label and disclosure – which is aimed
at improving consumers’ ability to differentiate Google search results from rivals’ search results
– are actually confusing or misleading to a substantial number of consumers.
Submitted this 9th day of December, 2013
14
Professors David J. Franklyn & David A. Hyman
15
Appendix
Franklyn is a tenured, full professor of law, specializing in trademark law, at the University of
San Francisco School of Law (“USF”). Franklyn is the Executive Director of the McCarthy
Institute for Intellectual Property and Technology Law at USF and Director of the Center for the
Empirical Study of Trademark Law (CEST). The McCarthy Institute, of which Franklyn is
Executive Director, conducts wide-ranging empirical research on consumer perception issues in
both the United States and Europe. Franklyn is also the Director of the Masters of Law Program
for U.S. and foreign lawyers in Intellectual Property Law at USF. Franklyn teaches and writes
primarily about trademark law and consumer perception issues. Franklyn is editor-in-chief and
co-author of McCarthy’s Desk Encyclopedia of Intellectual Property Law.
Franklyn works in collaboration with Professor McCarthy on a number of projects, including his
treatise on trademark law. Professor McCarthy is the author of a seven-volume treatise on
trademark law. It is the most-widely cited treatise on trademark law in the United States, having
been cited by courts in over 3,500 judicial opinions, including in a recent decision by the United
States Supreme Court in a case in which Franklyn was retained as an expert.
Franklyn has consulted and/or served as an expert witness on behalf of numerous clients in
numerous cases involving consumer perception issues, including in matters in the United States,
Asia, the European Union, the Middle East and South America. Franklyn has been retained as
an expert in several matters involving the issue of consumer perceptions. Franklyn was recently
was retained by the International Olympic Committee to provide trademark advice and advice
regarding the consumer impact of various logos regarding the 2016 Olympic Games. Franklyn is
a licensed attorney in Illinois and California. Franklyn has written several law review articles
about consumer perceptions on the Internet. Franklyn has also acted as a consultant to several
clients in matters related to trademark law, consumer perceptions and the Internet.
Hyman is the H. Ross and Helen Workman Chair in Law and Professor of Medicine at the
University of Illinois. He is tenured in both the College of Law and the College of Medicine.
He heads the Epstein Program in Health Law and Policy at the University of Illinois College of
Law. He teaches and writes about empirical law and economics, principally involving the
regulation of health care, and competition law and policy. He is the author of more than 100
articles in student edited law reviews, and peer-reviewed medical, health policy, and law
journals.
In competition law and policy, Professor Hyman served as Special Counsel to the Federal Trade
Commission from 2001-2004. In that capacity, Professor Hyman was principal author and
project leader for the first joint report ever issued by the Federal Trade Commission and
Department of Justice, “Improving Health Care: A Dose of Competition” (2004). He has also
published numerous other articles involving competition law and policy, including a series of
articles with Professor William Kovacic, the former Chairman of the United States Federal Trade
Commission.
Hyman has been deposed in various cases involving health care fraud and pharmaceutical
pricing, and has testified in one case involving pharmaceutical pricing.
16