Your SlideShare is downloading. ×
HOW ONLINE ADS MOVE VOTES IN POLITICAL CAMPAIGNS1
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

HOW ONLINE ADS MOVE VOTES IN POLITICAL CAMPAIGNS1

729
views

Published on

Published in: News & Politics, Business

0 Comments
2 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
729
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
28
Comments
0
Likes
2
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. HOW ONLINE ADS MOVE VOTES IN POLITICAL CAMPAIGNS A STUDY OF THFLORIDA’S 11 CONGRESSIONAL DISTRICT IN THE 2010 ELECTION CYCLE BY THE CALIFORNIA GROUP
  • 2. TABLE OF CONTENTSExecutive Summary Page 3Background Page 5Experimental Method and Survey Page 7MethodologyKey Findings Page 12Conclusions Page 20Appendices Authors’ Bios Page 21 Methodological Issues Page 23 Example of Digital Ad Used Page 25 Campaign Grid Targeting Techniques Page 26 2
  • 3. EXECUTIVE SUMMARYStudies show that political campaigns spend considerably less on onlineadvertising as a percentage of their total ad buy than do commercial advertisers.However, projections for 2010 showed that online ads would make up less than1% of all campaign advertising dollars spent1, but the same digital media waspredicted to outpace in 2010 the 12.9% market share it comprised in thecommercial world in 2009.2Campaigns, with their smaller budgets and restricted time frames, can ill-afford tospend money on a medium that does not give high impact. On the other hand,when engaged in a zero-sum competition, can campaigns afford to not use amedium that gives an advantage, even a small one?Our study of the impact of 14 million impressions of a negative online messageby Republican challenger Mike Prendergast against Rep. Kathy Castor (D, FL-11) in the last eight days of a Florida Congressional campaign shows thefollowing:Online Advertising Changes Votes ! Four percent of the electorate who could recall the message delivered via online ads – i.e. those for whom the message was “burned in” – changed their vote over the course of the campaign in favor of Prendergast. NOTE: in the 2010 cycle, 31 races in the House of Representatives were decided by 4% of the vote or less. ! Measured another way, in a post-election survey, those voters who could recall the message – those for whom the message was burned in – voted for Kathy Castor at a rate 8 percentage points less than those who could not recall the message.Online Advertising is Highly Targetable ! Within the targeted audience, Republican men, survey respondents being able to recall the message went from 4% to 22% over the course of the eight-day online ad campaign.1 Borrell Associates, 2010 Political Advertising Outlook: The Endless Campaign (Feb ’10)2 USA Today Social Media Lounge, May 14, 2010 “Advertising Spending Looks Up in 2010” 3
  • 4. Online Advertising is Highly Targetable (cont’d) ! The percent of Republican men who were able to identify Kathy Castor with the message went from 53% in the initial poll to 68% in the final poll. ! In the post-election poll, Republican men were the demographic sub- group that most associated Kathy Castor with the message. Sixty-eight percent of Republican men associated the negative men with Castor.Online Advertising Has a Significant Impact at the Last Minute ! In the above example of Republican men recalling the message, it took 9.3 million targeted impressions and six days of advertising to get a seven percentage point increase from 4% to 11%; it took an additional 5.3 million impressions over two days to get an additional 11 percentage points recalling the message, to 22%. ! Among all voters, the rate of increase was shallower but still substantial. The increase of all voters being able to recall the message phrase rose 3 percentage points with the first six days and 9.3 million targeted impressions and another 3 percentage points with the next 5.3 million impressions and two days of advertising. 4
  • 5. BACKGROUNDPrior to the advent of the Internet, the newest technology in voter contact camewith the use of television. Perhaps one of the first and still one of the mostfamous television commercials was the Lyndon B. Johnson “Daisy” commercialin 1964, which, ironically, was aired only once by the campaign, but airedhundreds of times on television newscasts.3 The Internet has yet to have such amoment.Despite the early impact of TV commercials such as “Daisy,” it wasn’t until nearlytwenty years later that political consultants were able to fully quantify the impactof television commercials on public opinion. This study was modeled after effortsin the early eighties in California and elsewhere that refined the use of televisioncommercials.California, with its large campaign budgets and twelve distinct media markets,lent itself to such experimentation. The 1982 Pete Wilson for U.S. Senatecampaign against then ex-governor and now current governor Jerry Brown wasat the forefront of these efforts. The Wilson campaign tested a unique messagein a distinct market. A unique message is defined as a message not being usedin that media market in any other medium. A distinct media market might be onesuch as Fresno or Monterrey that, at the time, did not get broadcast stations fromany other market.The Wilson campaign conducted a baseline poll of the awareness of themessage in that market. An example might be “Do you agree or disagree withthe statement ‘Jerry Brown is soft on crime,’” which, for the sake of argument,might get a 25% agreement. Then the campaign would run television adsillustrating that Jerry Brown was soft on crime in the media market and usetracking polls to assess the results.The Wilson campaign found that once you put 600 gross rating points of mediabehind an ad – that is to say, once you aired an ad a sufficient number of timesthat everyone in the market had theoretically seen the ad six times – it “movednumbers.” To move numbers might mean that a later poll would show a 35%agreement with the agree/disagree question. In political parlance, the change invoters’ opinions about Brown’s stance on crime would be called “burning in the3 The Daisy commercial featured a four year old girl picking petals off of a flower and countingthem as she does so. When she reaches nine, her voice changes to a male voice counting downa missile launch. The camera freezes on her face, and zooms in to her pupil, blacking out thescreen. A mushroom cloud consumes the screen as the male voice reaches zero. Johnson’svoice enters in a voice over saying: "These are the stakes! To make a world in which all of Godschildren can live, or to go into the dark. We must either love each other, or we must die." Anothermale voice follows with "Vote for President Johnson on November 3. The stakes are too high foryou to stay home." 5
  • 6. message.” The standard is now 1000 gross rating points, owing to the public’sincreased resistance over the last thirty years to political messaging, fracturing ofthe national consciousness with cable TV and the internet, and in general thestronger competing demands for Americans’ attention in 2010 vs. 1982. 6
  • 7. EXPERIMENTAL METHOD & SURVEY METHODOLOGYGeorge Gorton was the lead consultant on the Pete Wilson for U.S. Senatecampaign in 1982. In 2010, Google contracted with Gorton’s firm, The CaliforniaGroup, to adapt the methodology used in the television era to the Internet.Gorton hired Steven Moore, a pollster with experience in large-scale, non-traditional polling and a pioneer in online advertising for political purposes, tohelp with the study.Florida’s 11th Congressional District was home to a second term Democraticincumbent defending against a long shot GOP challenger in a fairly strongDemocratic district. Mike Prendergast, a retired Army colonel, ran against Rep.Kathy Castor.District ProfileDistrict: FL-11Democratic Incumbent: Kathy CastorRepublican Challenger: Mike PrendergastCook PVI Rating: D+11Barack Obama 2008 Vote: 66%2008 Election Result Castor (D) 71% Adams (R) 29%2010 Election Result Castor (D) 60% Prendergast (R) 40%Campaign Committee Spending Castor (D) $357,492 Prendergast (R) $496,080Independent Expenditure Spending NoneThe Prendergast campaign agreed to use a unique, negative message.A “unique, negative message” is a message that a) was being used only online;the message was not being used in any other medium (TV, radio, mail, etc.)while the test was being conducted and b) was being targeted at the opponentand designed to decrease the opponents’ favorable rating and increase theopponent’s unfavorable rating. We chose negative messaging because, in otherforms of voter contact, negative messaging about the opponent penetrates morequickly and effectively than positive messaging about the candidate. Weassumed the same would hold for digital ads. 7
  • 8. The message the campaign chose was one of wasteful spending. Specifically,Castor voted for a program that would spend $71,000 to study the effects ofcocaine on monkeys. In a baseline poll of CD 11 voters, this message made49% of all voters less likely to vote for Kathy Castor, and 80% of Republican menless likely to vote for Kathy Castor. The creative for the ads featured the phrase“Does your monkey need rehab?” (see Appendices for examples of the bannerads).The message that Castor voted for a program that would spend $71,000 to studythe effects of cocaine on monkeys had not been used in paid media by thePrendergast campaign prior to the test. Neither had it been used effectively inearned media. A search of the Tampa Tribune website yielded one reference tothe program, on March 16 of 2010.Regardless of the awareness of the program prior to the test, the specificmessage phrase “Does your monkey need rehab?” had not been used in anyform of media – paid or earned – by the Prendergast campaign.We conducted a benchmark survey (n=400) on October 24th, 25th and 26th. Thepurpose of this survey was to establish a baseline for voter opinions in the areaswe would be testing.We conducted tracking polling (n=200 nightly) on October 29th, 30st and 31st inadvance of the November 2 election.Finally, we conducted a post-election survey (n=400) on November 3rd and 4th.In compliance with FEC regulations, the results of the poll were not shared withthe campaign prior to Election Day.We wanted to test changes in at least four areas: 1) change in awareness ofInternet advertising 2) change in favorability ratings among those who had seenthe ads 3) change in voting behavior among those who had seen the ads and 4)effectiveness of “burning in” a message.Although we tested awareness of Internet advertising by asking respondentsdirectly, this is among the least important of the data we are measuring.The typical cycle of negative political advertising (independent of media) is asfollows:1) voters receive negative information about Candidate A via a politicaladvertisement by Candidate B 8
  • 9. 2) the voters’ opinions of both candidates decrease; Candidate A drops becauseof the negative information and Candidate B drops because most people don’tlike people who say negative things about other people3) voters forget the vehicle for the negative information about Candidate A andjust remember the negative information4) the voters’ opinion of Candidate B returns to around what it was prior to thedelivery of the negative ad5) the voters forget the negative information, and are just left with a negativeimpression of Candidate A.In addition to the California Group’s thirty-year experience, this cycle for negativepolitical television ads is confirmed by other studies. According to a 1998 studyby the University of Missouri School of Journalism: Janet Mullins, the 1988 Bush campaigns media director, claimed that "everybody hates negative ads; then they rate them most effective in terms of decision making. There isnt any long term effect . . . It is kind of like birth pains. Two days later, you forget how much it hurt. The same is true for negative advertising . . ."4People rarely admit to being influenced by political advertising, yet campaigncycle after campaign cycle shows that political advertising moves numbers. Assuch, the most important data is the change in favorability, change in votingbehavior, and amount to which the message was burned in.If poll respondents deny seeing an ad, but their opinions about a candidate havechanged in a direction that corresponds with the message of the ad, we have toassume that the ad is effective and respondents are either deceiving theinterviewer about the source of their opinions to make themselves look good, orthe respondents simply have forgotten the source of their opinion.Similarly, in our study, the final poll showed that the message was burned inamong 22% of the campaign’s target audience, Republican men. That is to say,in the final poll, 22% of Republican men could recall the message, while only17% of Republican men in the final poll reported having seen an Internet ad byeither candidate. As stated previously, the online ad was the only media bywhich the very specific message “Does your monkey need rehab?” be received,yet thirty percent more Republican men reported receiving the message thanreported seeing the medium through which the message was transmitted.One of the most common criticisms offered by detractors of the effectiveness ofonline advertising for political campaigns is based on a standard survey question4 http://www.scripps.ohiou.edu/wjmcr/vol02/2-1a-B.htm 9
  • 10. on campaign surveys asking respondents if they have seen, read, or heardsomething about Candidate A’s campaign, and then ask where the respondentgot that information. Respondents to such surveys tend to name online ads lessfrequently as a source of information about the campaign, regardless of howmuch is spent on online advertising.However, our findings seem to indicate that people lose recall of the source ofthe information they receive from online ads rather quickly and retain just theinformation, which is in line with what practitioners know and academics writeabout the cycle negative advertising. As such, data indicates that the standardseen/read/heard question used frequently by campaign pollsters is not a goodindicator of the effectiveness of digital ads.The Online BuyThe vendor who created and placed the online advertising was CampaignGrid,the largest online advertising platform dedicated exclusively to politicalcampaigns and causes. The firm placed several different types of onlineadvertising including banner ads placed on general interest, social media, news,entertainment, email portals and political sites as well as Google search text ads.Grid targeted registered voters using its proprietary voter data driven ad platformto burn in the message to first time viewers of the ad, and re-targeted people whoclicked on the ads and visited the campaign’s website (see Appendix III).CampaignGrid managed more than 140 online campaigns nationwide during the2010 cycle, including 76 general election races, among them Sharron Angle forU.S. Senate in Nevada, Christine O’Donnell for U.S. Senate in Delaware, andRichard Burr for U.S. Senate in North Carolina, as well as 55 House races and 9local races. 10
  • 11. The buy for the Florida campaign yielded more than 14 million impressions over8 days, or an average of 1.8 million impressions/day. The advertising buy wasback-loaded, so voters in the last days of the campaign would see moreimpressions than they saw in the first days of the buy. Following is a chart ofimpressions per day:For perspective on the size of the buy and the pacing of impressions, thePrendergast campaign was above the mean among CampaignGrid’s total clientbase when measured in terms of impressions per day, but was below the meanin terms of total impressions. This latter measurement is largely because of thelength of the campaign. The final three days, on an impression-per-day basis,was considerably above the mean. 11
  • 12. KEY FINDINGSInternet advertising changes the opinions of voters.We compared the poll results overtime (for instance, between theinitial and post-election polls), andinternally within the final poll togauge the impact of the onlineadvertising. Both methodsconfirmed the effectiveness ofonline ads.All Voters, Measured Over TimeAmong all voters, the rate at whichrespondents reported having heardthe phrase “Does your monkeyneed rehab?” went from 8% in theinitial poll to 12% in the final poll(n=400 in both cases), a 4percentage point increase. Perhaps more profound is that those who were able to label Kathy Castor as the politician most likely to be associated with a program to spend $71,000 on the effects of cocaine on monkeys increased 30% percent over time. In an open ended question, respondents increased by six percentage points, from 20% to 26%. 12
  • 13. The biggest change, though, occurred in the questions when respondents weregiven a choice between the two named Congressional candidates. Those whoidentified Castor with cocaine monkeys went from 28% in the first poll to 36% inthe final poll, an 8 percentage point increase. Under the same parameters,Prendergast remained statistically unchanged, going from 12% to 13%.While 4%, 6%, and 8% are not as large of movements as those currentlyassociated with television, the movements are consistent in direction and thedata proves the effectiveness of digital ads to change voters’ opinions. Again,the Internet is not yet the campaign howitzer that television advertising is, but theoutcomes of 31 House races decided in the 2010 by 4% or less could have beenchanged by the skillful application of digital ads.All Voters, Measured Within the Final PollLooking at the polls internally yielded similar results, albeit with smaller samplesizes and a much higher margin of error.Among all voters who had seen the Internet ads (n=56), the vote on the final pollwas 58% for Castor to 42% for Prendergast – a four-point swing when comparedto actual election results. 13
  • 14. 2010 Elec tio n Results 4 point swing in Prendergast’s FavorLooking at the results from another angle (below), those voters who had seen,read, or heard the phrase “Does your monkey need rehab?” – those with whomthe message was burned in (n=55) – voted for Castor at a rate of 53%. Thosesurvey respondents with whom the message was not burned in voted for Castorat a rate of 61%. 14
  • 15. Those voters who had seen, read or heard the phrase “Does your monkey needrehab?” had a 51-38 fav/unfav for Kathy Castor (n=55), while those who had notseen/read/heard the phrase had a 52-27 fav/unfav for Kathy Castor (n=346).That is to say, among those voters who had not seen seen/read/heard themessage phrase, Castor was +25 favorable to unfavorable rating. Among thosevoters who had seen/read/heard the message phrase, Castor was limited to a+13 favorable to unfavorable rating.This change in Castor’s favorable to unfavorable ratio is largely a factor ofCastor’s unfavorable rating increasing 11 points among those who had seen,read or heard the phrase “Does your monkey need rehab?” As one mightexpect, when the negative message against Castor is burned in, Castor’snegative rating goes up. Castor Fav Castor Unfav Fav – Unfav had not s/r/h message phrase 52 27 +25 s/r/h message phrase 51 38 +13 Change -1 +11 -12 15
  • 16. Digital advertising is dexterous and precise in terms of targeting.Not surprisingly, thosedemographics to which theadvertising was targeted werethe demographics in whichopinions were changed mostsignificantly.Our initial poll showed thatPrendergast had the vote ofonly 77% of the Republicans,and 74% of the Republican +1 8men. The campaign must Increasehave been seeing the samething, and felt that in the finaldays of the campaign thecandidate both had room togrow with Republican men,and that this would be a goodgroup to motivate for voterturnout. As such,CampaignGrid designed its campaign to target Republican men. (for a detaileddiscussion of CampaignGrid’s targeting techniques, see Appendix IV)The percentage of Republican men who reported having seen, read or heard thephrase “Does your monkey need rehab?” increased from 4% in the first poll to22% in the second poll, for a net change of 18 percentage points and an increaseof 550% between the first and final polls. 16
  • 17. When given a choice of the two Congressional candidates with whom toassociate spending $71,000 to study the effects of cocaine on monkeys,Republican men associated Castor at a rate of 53% initially, and 68% in the finalpoll. The percentage associating Prendergast dropped 4 points from 9% to 5%.Of the two candidates running for Congress in your Congressional district, Kathy Castor and MikePrendergast, which of the candidates would you associate with voting for a program to spend$71,000 of taxpayer money to fund a study of the effects of cocaine on monkeys? +1 5 IncreaseThe percentage of Republican men voting for Prendergast increased from 74%to 82% between the initial poll and the post-election poll. 17
  • 18. Internet advertising is effective at the last minute.One of the most surprising findings of the study was that Internet advertising canmake a significant difference in a short amount of time. And, unlike methods ofvoter contact such as mail and television, Internet advertising needs very littlelead-time to be introduced into a campaign, and as such can be created, targetedand served as late as Election Day itself.As shown in the chart below, GOP men who had received the messageincreased by seven percentage points over five days of advertising that yieldedsome 9.3 million impressions. Over the last two days, the percentage of GOPmen with whom the message was burned in jumped eleven percentage points asthe campaign showed more than 5.3 million impressions.The cause of the dramatic increase over the last two days is difficult to ascertainwith certainty. In the last two days, 2.5 million impressions per day were servedvs. 1.6 million impressions per day in the first six days, which could account forthe increase. This is particularly true if the last three days are counted as agroup, which is a reasonable thing to do, as the 11% polling number on 10/31was generated by three days of n=200 polling – 10/29, 10/30, 10/31. Such agrouping would create a 3 to 1 difference in the impressions per day for 3.0million impressions on the last three days and 1.1 million impressions for the firstfive days. This would imply that burning in an ad with a target audience is simplya function of packing impressions into days and increasing the frequency withwhich a target audience sees an ad. 18
  • 19. Another theory as to the reason for the dramatic increase over the last two daysof the campaign parallels political television advertising. TV ads are well knownby practitioners to have a “hockey stick” type result on polls. Some media buyerswould say that if you air an ad with only 600 gross rating points behind it, youmight as well not do it, because all the benefits on a poll accrue from 800-1000gross rating points. This would indicate that the necessary pre-requisite for thedramatic results of the last two days was the number of impressions that wereserved in the first six days.Unlike television advertising, digital advertising can be adjusted in real time. Ifone version of an ad is working better than another, then resources can instantlybe put behind the effective ad, increasing the efficiency of the buy. The“learning” nature of the digital ad realm could be another reason for the hockeystick shaped result.In either case, achieving a 550% increase in message recall by a target audiencein the last 8 days of a campaign is a huge benefit to a campaign.Comparing Digital Advertising to More Conventional MediaA value of online advertising not necessarily measured in polls that none the lessbecame apparent in this study is the ability to deploy messaging online long afterthe window closes on sending mail or putting commercials on TV.Typically, a TV commercial would need to be on air eight to ten days to getenough GRPs behind it to burn in a message. Producing a TV commercial cantake as little as 24 hours or several days, depending on the production values ofthe commercial. Getting the commercial in rotation at the television stationstakes time. As such, the last day to begin airing a TV commercial is probably tendays to two weeks prior to Election Day. Some media consultants might arguethat you could compress the timeline to air a commercial and have it still beeffective, or that a more powerful commercial needs fewer GRPs, but onethousand GRPs is the gold standard for airing a TV commercial.Similarly, mail takes time to produce and deliver. Most mail vendors will requirea finalized mail design be in their office ten days prior to Election Day.Digital advertising can be designed and online in hours, and the results of thissurvey prove that a large media buy in the last days of a campaign can movenumbers. In this study, 5.3 million impressions in the last two days of thecampaign moved 3% of voters, and 11% of targeted voters. As noted earlier, 31races in 2010 were decided by fewer than four percentage points. 19
  • 20. CONCLUSIONS! This study proves that digital advertising changes the opinions of voters.! Online advertising in its current form is much more precisely targetable than broadcast media such as radio and television. The next generation of digital advertising will be even more targetable, and the strategists and vendors who capitalize on this most capably will be the next cycle’s leaders in the world of political online advertising. Campaign Grid, for instance, is already working on a platform to incorporate consumer data and voting behavior into targeting for online advertising.! No other media can be deployed as quickly or effectively in the final ten days of a campaign as digital media, with the possible exception of automated phone calls. Longer digital media campaigns during earlier phases of the election cycle are also effective, but those were not measured here. As targeting becomes more sophisticated, the value of digital advertising in persuading voters and ginning up turnout in the final days of a campaign will increase greatly. 20
  • 21. APPENDIX I: Authors’ BiosGeorge GortonGeorge Gorton’s political career spans four decades and three continents.Most recently, Gorton was Arnold Schwarzenegger’s first political consultant,masterminding Proposition 49, an $8 million statewide initiative campaign thattransformed Schwarzenegger, in the eyes of the electorate, from an action star tothe leading candidate for governor going in to the 2003 special election.Gorton is also well known for leading a small group of American politicalconsultants who secretly advised Boris Yeltsin’s re-election campaign in 1996.Gorton and his team brought Yeltsin from fifth place in the polls at six percent toa thirteen point victory over his communist rival five months later. The campaignwas chronicled in a TIME Magazine cover story and the subject of “SpinningBoris,” a Showtime movie in which Jeff Goldblum played Gorton.Gorton has also played a strategic role in campaigns in Romania, Panama, theCzech Republic and Canada.During the eighties and nineties, Gorton was the innovator who ran all of PeteWilson’s statewide campaigns in California. Gorton helped quantify the impact oftelevision on public opinion through extensive testing of the medium duringWilson’s campaigns in the eighties. Gorton created methodology whichdetermined how many gross rating points a campaign needed to put behind acommercial in order to “burn in” a message.Gorton’s political career started out as youth director for the William Buckley forU.S. Senate campaign in New York. He subsequently was named national youthdirector for President Nixon’s 1972 re-election campaign.Steven E. MooreSteven Moore is a Republican political strategist with extensive global pollingexperience, having conducted surveys in Russia, Romania, Indonesia, Iraq, EastTimor and the United States.While in Iraq, Moore designed and conducted a monthly 12,000 sample surveyduring some of the most tumultuous times in that country. The complexitiesinvolved in conducting a poll with a sample size ten to twelve times larger than atypical national U.S. poll on a monthly basis using face-to-face interviews in acountry with no history of public opinion research during a shooting war are self-evident. The General staff used the information gathered in the survey to assess 21
  • 22. reconstruction efforts, gain insight into what Iraqis wanted in their governmentand evaluate the effectiveness of US military "soft power" efforts.Since returning from Iraq in 2006, Moore has been active in Congressionalcampaigns, and has been recognized as an innovator in using onlinecommunications to communicate with Congressional voters about issuesimportant to them.Moore brings a unique approach and skill set to his endeavors, by utilizingquantitative data to solve complex problems. His combination of quantitativeskills and online advertising innovator status made him uniquely qualified toauthor this groundbreaking study. He was not involved in the Prendergastcampaign.Formerly a partner in Gorton Moore International, Moore has been involved inpolitics for more than twenty years. 22
  • 23. APPENDIX II: Methodological IssuesTesting Awareness of Negative AdsTesting negative messaging posed a methodological challenge, but not oneuncommon to political surveys. When testing a positive message about, forexample, Mike Prendergast, one can simply ask, “Have you seen an Internet adabout Mike Prendergast?” Such a positive ad is about Mike Prendergast and hasdisclaimer language that says it is paid for by Mike Prendergast for Congress.However, when testing a negative ad, the negative ad is about Kathy Castor butis aired by Mike Prendergast for Congress. The subject matter of a negative adagainst Kathy Castor is, by definition, about Kathy Castor, but it was created andpaid for by Mike Prendergast. It even says so on the ad. So in analyzing theresults of the polls, we combined the respondents who said they saw the Castorad and the respondents who said they saw the Prendergast ad, then adjusted forthose who saw both ads. (see Venn diagram below) This was a particularlyeffective solution in the Florida race given that Castor’s ad buy was negligiblenext to Prendergast’s and no third parties were advertising on the Internet in thisrace.The blue circle represents those respondents who reported seeing the Castor ad.The off white circle represents those respondents who report seeing thePrendergast ad. Where the circle intersects is the group that has seen both ads.Seventy-three respondents reported seeing one ad or the other, and seventeenrespondents reported seeing both ads. To accurately count penetration of thead, we need a count of those respondents who have seen either ad. These 17that report seeing both ads are counted twice, in both the blue and the off-whitecircles. The 37 that reported seeing the Castor ad is comprised of 20 who sawthe Castor ad only and 17 that saw both. The 36 that reported seeing thePrendergast ad is comprised of 19 that saw the Prendergast ad only, and 17 thatsaw both. Thus, the 17 is counted twice and the real number for those who haveseen either ad is 56. 23
  • 24. Small Sample SizesOne of the issues associated with examining a form of voter contact that reachesbetween 15% and 30% of the electorate is that the sample size for subgroupsgets small rather quickly. However, we found that even using sample sizes thatmight yield, say, a margin of error of seven or eight percent, all the numbers weregoing in the same direction, yielding consistent trends, and increasing thebelievability of our conclusions. Smaller sample sizes are identified in thefindings. We also tried to mitigate the impact of small sample sizes by mergingthe tracking polls (n=600) with the data from the final poll (n=400). Obviously,the larger overall sample size yields larger samples for sub-groups.Accounting for “Phantom Reporting”In the initial survey, we ask whether voters have seen Internet ads from eithercandidate, and we get a positive response despite the fact that no Internet adshad been run by either campaign at that point in the race. This can beinterpreted a couple of different ways: 1) voters are reporting having seen ads byother candidates whose districts overlap with FL-11, such as ads for governor orUS Senate or 2) some small percentage of survey respondents will say anything.For example, experience has shown that a candidate who has only recentlymoved in to a district and done nothing to make voters aware of their candidacycan still get a name ID of 10% or 15%. Just as voters will report knowing a“phantom candidate,” it is possible they will also report a “phantom ad.”Regardless of the reason for the phantom reporting in the initial poll, the questionbecomes one of how to interpret subsequent polling – i.e. if 7% report seeing anInternet ad when none was present, and a subsequent poll shows that 20%report seeing the ad, do we assume a 7% “phantom number” in the subsequentpoll and report a “real” awareness of the ad of 13%? We chose not to. In theexample of the unknown candidate generating name ID in an initial poll, nocampaign we are aware of subtracts the phantom name ID generated in initialpolling when a subsequent poll is conducted for name ID in the presence of votercontact efforts. Meaning, if an initial poll shows 15% name ID for candidate BobAnderson who moved in to Florida’s 22nd Congressional District three weeks priorto the poll and has not conducted a campaign, and a subsequent poll conductedafter candidate Anderson has sent out five mail pieces shows him with 40%name ID, no campaign would knock his name ID down to 25%.In both the case of the candidate and the Internet ad, a subsequent poll does notoffer a way to distinguish between “phantom responses” and actual responses.As such, a campaign has to assume that all responses are valid responses. 24
  • 25. APPENDIX III: Example of Digital Ad Used 25
  • 26. APPENDIX IV: Campaign Grid Targeting TechniquesCampaignGrid used several forms of targeting to reach Republican men.1) Geographic targeting of Registered VotesCampaignGrid used its state of the art technology and precision online targetingto target registered voters in specific “voter zones”. (Note: voter zones are 2.5times smaller than traditional zip codes and allow for more precise targeting ---(see the “heat map” for Republicans in FL-11) Note that FL-11 is fairly mixed inits partisan affiliation. Very little is strong Republican, and very little is strongDemocrat. Given the limited Republican strongholds in FL-11, CampaignGridalso targeted using other methods. 26
  • 27. RetargetingDuring the 2010 election cycle, CampaignGrid also created a massive targetingand retargeting program, deploying cookies and then serving ads to thosecookied individuals, in conjunction with numerous conservative websites, such asFox News, allowing advertisers to reach directly self-identified conservativevoters. In Fl-11, 384,000 ads were served to people with conservative cookieson their browser.Site TargetingRepublican men visit certain sites. Among those sites targeted were AOL,nbcsports.com, realclearpolitics.com, cnn.com, and townhall.com.Keyword Targeting on FacebookFacebook provides easy keyword targeting through users’ use of affinity groups –Sarah Palin, Marco Rubio, America Speaking Out, Florida Republican Party, etc.– and this approach was used by CampaignGrid to target ads.About Campaign GridCampaignGrids patent pending, online voter targeting platform, combines 187million voter records with associated consumer data for each voter in theelectorate.Real time data mining allows advertisers to microtarget an audience on the weband cell phones using 172 different consumer and voting behaviors. Grid haspioneered hyper-local data driven online targeting by mapping the electorate tothe web.Interactive ads that feature static banners and videos are designed to list build,fund raise and persuade. Real time feedback on which ads work are used tooptimize the ad campaign. 27

×