Study Summary

To observe the use of paid search advertising in the final hours of the 2006 US midterm elections, we queried Google for the names of candidates running in swing Senate races and analyzed the resulting 569 ads. We observed:

• Political pay-per-click advertisers use Google. Few political advertisers use Yahoo Paid Search.
• There are few political advertisers: the average search results page for queries in this study returned only 3.7 ads.
• The most prevalent advertisers within this query set were Accoona (search engine), Gather.com (social networking), CafePress (retailer), and GOPSenators.com (National Republican Senatorial Committee).
• “Red” ads (pro-Republican or anti-Democrat) outnumbered “blue ads” (pro-Democrat or anti-Republican) two-to-one.
• Blue ads were three times more likely to be negative than red ads.
• No campaign ads referenced President Bush.
• Iraq and Al-Qaeda were only mentioned only twice (stopidioticforeignpolicy.com, against Jim Webb)
• Only two ads linked directly to videos (against George Allen: “macaca” and prescription drugs).
• Blue ads were longer than red ads. There was no difference in the length of words in blue and red ads.
• Blue ads were more likely to include an exclamation point. Red ads were more likely to contain a question mark.

We conclude that paid political advertising on the search engines is still in its infancy. Note the cost of search advertising is a fraction of the cost of television advertising, yet often has comparable reach.

We anticipate search engine marketing will become increasingly prevalent to the American political process in coming elections. In the 2008 elections, we predict political advertisers will completely fill the paid search spots on the first page of Google search results pages; political advertisers will advertise heavily on Yahoo; political advertisers will use more video; and Democrats will increase spending to match Republicans.

Method

We queried Google and Yahoo for terms related to the 2006 US midterm elections from November 3, 2006 to November 6, 2006. We found virtually no political ads on Yahoo. Yahoo searches on candidate names returned poorly-targeted e-commerce ads. For example, a Yahoo search george allen returned paid results from AOL Movies, Gifts.com, Dealtime, PeopleFinders, FindPeopleToday, etc. This observation suggests political organizations have largely ignored Yahoo as an advertising channel.

Accordingly, we focus our analysis on paid search on Google AdWords.

Using FireFox with cookies disabled, we searched Google 11 times for 14 candidate names in swing US Senate elections, yielding 154 searches over the last few days of the campaign.

Candidate names used as search queries were bob casey, bob corker, conrad burns, george allen, harold ford, jim pederson, jim talent, jim webb, jon kyl, jon tester, lincoln chafee, mike dewine, rick santorum, sherrod brown (Table 1).

We also conducted general political searches (ie war on terror, war in iraq, iraq policy, national security, control of congress, house of representatives, close elections, economic policy, bush popularity, republican base, democratic base, election turnout, congressional elections). We found little campaign advertising on these general terms. Accordingly, we focus our analysis on candidate names.

Table 2 presents the 569 paid search ads displayed on these 154 Google queries.

Note that Google may serve different paid search ads for the same query based on the user’s geography, prior searches, and other factors. Different users conducting identical searches during this time frame may have received different Google ads than we did.

Advertisers

The most prevalent advertiser was Accoona.com, a search engine associated with the Chinese government, appearing on 56% of the 154 searches. In close second was Gather.com, a social networking site aimed at the 35-and-over demographic, appearing on 47% of searches. CafePress.com, an e-commerce site offering personalized products, came in third, appearing on 29% of the searches. The fourth most prevalent advertiser on this query set was the National Republican Senatorial Committee (gopsenators.com), appearing on 28% of the searches.

Table 3 presents counts of ads by advertiser.

Competition

Google search pages can contain 8 to 12 ads. Google sells these spots to advertisers using a dynamic real-time auction. On typical e-commerce searches on competitive terms (dvd, business computer, flowers), every slot on the first page of search results is filled.This was not the case for these searches. Even on the day prior to the election, the ad slots on Google search results pages were largely empty.

The average number of ads returned by a query in this search set was 3.7 (Table 4).

Red v. Blue

We characterized each ad as “red” (pro-Republican or anti-Democrat), “blue” (pro-Democrat or anti-Republican), “other” (political, but neither pro-Democrat nor pro-Republican), or “neither” (not supporting any political party). 58% of these ads did not support a political party. These were ads from CafePress (campaign products) and ads from news sites (election information).

Red advertising dominated blue: 26% of these ads in the sample were “red” (pro-Republican or anti-Democrat), versus only 16% “blue” (pro-Democrat or anti-Republican).

We observed three instances of ads (0.5%) run by the Libertarian party. These were trigged by the search phrase rick santorum and criticized both major parties (”Vote Libertarian in 2006 Don’t waste your vote on the two old parties. Vote Libertarian”)

Negative Ads

We characterized an ad as “negative” ad if it used negative language to characterize a politician or party. Table 5 presents the ads characterized as negative.

63% of blue ads were negative, versus 23% of red. This difference was statistically significant (chi-sq(1)=37.6, p=0.000).

The most negative ad in the sample was by blogpac.org on the phrase george allen, accusing the campaign of assaulting a constituent (”George Allen Beat Me Up Watch video of Allen staffers attacking a constituent - BlogPAC www.blogpac.org”).

Ad Length

Google AdWords consist of a 25 character title followed by 70 characters of text; thus, an Adwords ad has a maximum length of 95 characters, including spaces. Blue ads were typically longer than red ads by 9 characters, and this difference was statistically significant (83.4 vs. 74.4, t=5.8, p=0.000). There was no significant difference between the average length of words in red versus blue ads.

Ad Copy

No blue or red campaign ads mentioned President Bush. The word “Bush” only occurred in ads for littledemocrats.com (”Why Mommy is a Democrat The book George Bush doesn’t want your kids to read!”), running on the phrase sherrod brown.

23 three ads mentioned “Iraq”. These were anti-Jim Webb ads run by www.stopidioticforeignpolicy.com (”Webb and Bin Laden Both Agree We Should Leave Iraq Now - Be Smart - Vote Allen” and “Webb Wants the U.S. to leave Iraq now–so does Al-Qaeda–So who gets your vote?”) running on the phrases jim webb and george allen.

Blue ads were more likely than red to include an exclamation point (18% vs. 9%, chi-sq(1)=4.4, p=0.03).

Red ads were more likely than blue to contain a question mark (33% vs. 7%, chi-sq(1)=20.8, p=0.000).

The words “help” and “now” occurred frequently in these ads. Both red and blue ads used the word “help” frequently (27% of ads). 27% of red ads contained “now”, versus 0% for blue ads (chi-sq(1)=28.9, p=0.000).

Video And YouTube

Only two ads linked to video clips, both on YouTube. These were attack ads against Senator George Allen discussing his prescription drug policy and his “Macaca” comment.

Tracking

While it is difficult to assess the sophistication of an advertiser simply by inspecting their paid search ads, we suspect most political pay-pay-click advertisers are less advanced than their e-commerce counterparts. Only 30% of destination URLs contained tracking codes to enable the advertiser to monitor the traffic from the ad. Most of these embedded tracking codes tracked only the referring engine (”sid=google”) rather than a specific ad.

Conclusions

We believe political pay-per-click advertising is in its infancy. We base this opinion on the low utilization, limited tracking, Google-only perspective, and name-only search emphasis. Because of the greater cost efficiency of pay-per-click advertising, we predict campaigns and political action committees will move aggressively into paid search advertising as a marketing channel in future elections.

Additional Information

See all of the 2006 Election Study Tables

Download the PDF of the Study:
Google AdWords Paid Search Advertising In Close 2006 US Senate Races