“WOULD YOU STILL HANG UP ON THIS POLL IF YOU KNEW . . . :” EXPERIMENTS TO IMPROVE THE DESIGN OF POLITICAL MESSAGE TESTING POLLS Thomas M. Guterbock Director, Center for Survey Research University of Virginia TomG@virginia.edu Deborah L. Rexrode Research Analyst, Center for Survey Research University of Virginia dlr3r@virginia.edu Samantha Luks Vice-President for Special Projects YouGovPolimetrix sam.luks@yougov.com Presented at the Annual Meeting of the American Association for Public Opinion Research Hollywood, Florida May 15, 2009 UPDATED ABSTRACT May 2009 Political message testing polls occasionally generate trouble: complaints to AAPOR, negative press attention, denunciations from opposing campaigns, and criticism from others in the survey profession. These polls also present ethical issues, since they may not always adequately secure informed consent. Some generate large numbers of break-offs from annoyed respondents, suggesting that significant numbers are unsatisfied with their experience in interviews of this type. The exact form of message-testing surveys varies substantially across firms and campaigns. However, there has been no published research about which features of these surveys might affect the experience of those who respond. We report on a multi-factor experiments testing whether specific changes in the design of a message testing poll can produce improvement in the perception of the interview among respondents. The vehicle for these experiments is a message testing poll about the generic 2010 Congressional election. Three features of the questionnaire are manipulated in a trial fielded by Polimetrix/YouGov in March 2009. (1) An abrupt transition from ordinary preference questions to persuasive or “push” questions is tested against more informative transitions that forewarn the respondent that politically charged messages will ensue. (2) The sequencing and degree of balance (between positive messages about one party’s candidate for Congress and negative messages about the other) is varied. (3) The questions that are asked after each persuasive item are varied, ranging from questions about changed voting intention to less direct questions that ask about the convincingness, believability and importance of each item. The dependent variables are responses to a closing interrogatory in which respondents assess the fairness and availability of the questions, whether they felt fully informed about the interview, their degree of concern with partisan use of the survey results, whether the interview was comfortable or stressful, and their willingness to be interviewed in the future. We present and discuss results showing that these judgments are strongly affected by the features of the questionnaire. The results demonstrate that respondent survey experience in message testing polls can be improved by betterment of poll design. 2 SUMMARY OF EXPERIMENTAL DESIGN FACTOR 1 TRANSITIONAL INTRODUCTION Control or Generic Group: No Introduction Intro 1: Here are some statements you might hear from a political candidate running for office. Your thoughts and opinions in response to these statements are an important part of our research. Intro 2: The following statements are the sort that you might hear in a political campaign, or a campaign commercial These statements are different from what you might expect from a usual poll or survey. You might not agree with these statements, and some are negative. We are testing people’s reactions to these statements and we would like to warn you that these statements could cause some people to react strongly. This is ok. Your thoughts and opinions in response to these questions are an important part of our research. FACTOR 2 DESIGN BALANCE Each design will present three blocks of message items. The first block will contain messages about the favored party, and the second and third blocks will contain messages about the opposing party. The strongest or least “fair” negative messages will be included in the third block. Unbalanced design: 6 positive messages about favored party. 7 negative items about opposing party 7 more negative items about opposing party Partially balanced design: 4 positive and 2 negative messages about favored party 2 positive and 5 negative messages about opposing party 1 positive and 6 negative messages about opposing party 3 FACTOR 3 TEST QUESTIONS Test 1: When you hear the following statements about the Democrats in Congress, does knowing about this make you more likely or less likely to vote for a Democratic party candidate, and how strongly do you feel about that? 1 Much more likely 2 Somewhat more likely 3 Somewhat less likely 4 Much less likely 5 No opinion Test 2: When you read the following statements about the Democrats in Congress, how convincing are they? (Positive statements) 1 Very convincing 2 Somewhat convincing 3 Somewhat unconvincing 4 Not convincing 5 No opinion When you read the following statements about the Democrats in Congress, how much doubt do they create? (Negative statements) 1 Serious doubt 2 Some doubt 3 Very little doubt 4 No doubts 5 No opinion Test 3: When you read the following statements about the Democrats in Congress, how believable do you think this statement is? 1 Very believable 2 Somewhat believable 3 Somewhat unbelievable 4 Not believable 5 No opinion Considering these same statements, for you as a voter, how important is it for you to know this information? 1 Very important 2 Somewhat important 3 Not important 4 No opinion 4 EXAMPLE OF PARTISAN STATEMENTS Republican Republicans have supported the war in Iraq which has taken a devastating toll on our military, our economy, and our reputation around the world. (Neg – War) The Republican Party has always been firm and strong regarding the War in Iraq. A withdrawal without completely securing our gains would be irresponsible. (Pos – War) Republicans are working hard to strengthen our economy by advancing free trade with growing economies around the world. (Pos – Foreign Trade) Because of the strong Republican support for the unlawful invasion of Iraq several years ago, America is now at odds with the rest of the world. (Neg – Diplomacy) Republicans act as if peace in the Middle East is not an achievable goal. Therefore, they do not try to facilitate any understanding and compromise. (Neg – Diplomacy) Democratic Democratic leaders’ emphasis on troop withdrawal in Iraq is undermining all that we have accomplished there. If we pull out now, we may lose everything we have worked for. (Neg – War) Democrats in Congress say we must strengthen America’s security by starting to reduce our troops in Iraq in a responsible way and work with other nations to bring stability. (Pos – War) Democrats work hard to keep American jobs on American soil. (Pos – Foreign Trade) While diplomacy is appropriate in some circumstances, the democrats are too soft when it comes to defense and need to recognize that military action is sometimes the best option. (Neg – Diplomacy) There will never be peace in the Middle East as long as terrorists are not fully and universally condemned, yet Democrats refuse to stand up to the Palestinians. (Neg – Diplomacy) 5 ATTRIBUTES OF QUESTIONS Thinking about the questions you were asked related to Congress… {FUTURE} If we contacted you in the future to do another survey like this, do you think you would participate? 1 2 3 4 Would definitely participate Would probably participate Would probably not participate Would definitely not participate {FAIR} How fair were the questions about Congress? 1 2 3 4 5 Very fair Somewhat fair Somewhat unfair Very unfair Not sure/Don’t know {BELIEVE} How believable was the information presented about Congress? 1 2 3 4 5 Very believable Somewhat believable Somewhat unbelievable Not believable at all Not sure/Don’t know {STRESS} Was answering these questions comfortable or stressful? Did answering feel… 1 2 3 4 5 Very comfortable Somewhat comfortable Somewhat stressful Very stressful Neither comfortable nor stressful {HOWUSED} Do you think these survey results will be used to promote or oppose a particular party’s congressional campaign? a. Yes [GO TO SUPPORT] b. No [GO TO MISLED] 6 {SUPPORT} How concerned are you that the survey results might be used to promote a particular party’s congressional campaign? 1 2 3 4 5 Very concerned Somewhat concerned Not very concerned Not concerned Not sure/Don’t know {EXPECTED} Thinking back to the introduction to this survey, were the questions what you expected? 1 2 3 4 Exactly what I expected Close to what I expected Somewhat different than I expected Not at all what I expected [GO TO MISLED] SCRIPT FOR PANEL PARTICIPANT DEBRIEFING AT THE COMPLETION OF THE SURVEY: A final note about this survey: This survey was part of an academic experiment conducted by Polimetrix on behalf of the Center for Survey Research at the University of Virginia to better understand how voters respond to surveys that test political campaign messages. Such surveys are frequently conducted by campaign consultants of both major political parties. The University does not conduct partisan polling or research for parties or candidates. This research does not reflect the political views of Polimetrix, the Center for Survey Research, or the University of Virginia. This study has been approved by the Institutional Review Board of the University. The design of the study involves an equal number of participants from each political party. The purpose of this survey is to test methods that would improve the practices of political polls and improve the experience for respondents like you. The study utilizes different sequencing of questions with some versions of the survey having a stronger negative or positive emphasis. Some versions favor Republicans; some favor Democrats. Your reactions to these various methods will help us develop more effective and responsible research methods for conducting these types of polls. Thank you very much for your assistance with our research 7 ATTRIBUTES BY TREATMENTS* Crosstab fairr Fairness of questions about Congress--recoded 1.00 very unfair 2.00 3.00 4.00 5.00 Total Count % within s elec ttreat compare generic with the better t reat ments s omewhat unfair Count % within s elec ttreat compare generic with the better t reat ments not s ure/don't know Count % within s elec ttreat compare generic with the better t reat ments s omewhat fair Count % within s elec ttreat compare generic with the better t reat ments very fair Count % within s elec ttreat compare generic with the better t reat ments Count % within s elec ttreat compare generic with the better t reat ments selectt reat compare generic with the better treatments 2.00 balanc ed with 1.00 any transit ion control/gener and alt ernate ic treatment tes t questions 73 10 Total 83 39.0% 4.2% 19.5% 30 20 50 16.0% 8.4% 11.7% 13 16 29 7.0% 6.7% 6.8% 28 75 103 15.0% 31.4% 24.2% 43 118 161 23.0% 49.4% 37.8% 187 239 426 100.0% 100.0% 100.0% selectt reat compare generic wit h the better treat ments 2.00 balanc ed with 1.00 any transit ion control/gener and alt ernate ic treat ment tes t questions 59 7 Total 66 Crosstab believer Believabilit y of info presented about Congress-recoded Total 1.00 not believable at all Count % within select treat compare generic with the bet ter t reatments 2.00 s omewhat Count believable % within select treat compare generic with the bet ter t reatments 3.00 not s ure/ don't know Count % within select treat compare generic with the bet ter t reatments 4.00 s omewhat Count believable % within select treat compare generic with the bet ter t reatments 5.00 very believable Count % within select treat compare generic with the bet ter t reatments Count % within select treat compare generic with the bet ter t reatments 31.6% 2.9% 15.5% 31 31 62 16.6% 13.0% 14.6% 11 23 34 5.9% 9.7% 8.0% 57 106 163 30.5% 44.5% 38.4% 29 71 100 15.5% 29.8% 23.5% 187 238 425 100.0% 100.0% 100.0% *Differences between treatments on all six of the attributes are significant. 8 Crosstab expectedr Questions matched expectations 1.00 not at all what I expected 2.00 s omewhat different than I expected 3.00 close to what I expected 4.00 exactly what I expected Total Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments selecttreat compare generic with the better treatments 2.00 balanced with 1.00 any transition control/gener and alternate ic treatment tes t questions 60 31 Total 91 32.8% 13.0% 21.6% 53 58 111 29.0% 24.3% 26.3% 53 115 168 29.0% 48.1% 39.8% 17 35 52 9.3% 14.6% 12.3% 183 239 422 100.0% 100.0% 100.0% Crosstab misuse concern that results will aid a campaign Total .00 won"t be used Count % within select treat compare generic with the bet ter t reat ments 1.00 not c oncerned, not Count sure % within select treat compare generic with the bet ter t reat ments 2.00 not very c onc erned Count % within select treat compare generic with the bet ter t reat ments 3.00 somewhat Count concerned % within select treat compare generic with the bet ter t reat ments 4.00 very c onc erned Count % within select treat compare generic with the bet ter t reat ments Count % within select treat compare generic with the bet ter t reat ments selectt reat compare generic wit h the better treat ments 2.00 balanc ed with 1.00 any transit ion control/gener and alt ernate ic treat ment tes t questions 58 102 Total 160 31.2% 43.2% 37.9% 25 42 67 13.4% 17.8% 15.9% 21 36 57 11.3% 15.3% 13.5% 36 34 70 19.4% 14.4% 16.6% 46 22 68 24.7% 9.3% 16.1% 186 236 422 100.0% 100.0% 100.0% 9 Crosstab stress r int erview comfortable or stress ful--recoded in ordinal sequence 1.00 very comfortable 2.00 s omewhat comfortable 3.00 neither c omfortable nor stressful 4.00 s omewhat st ress ful 5.00 very stres sful Total Count % within s elec ttreat compare generic with the better t reat ments Count % within s elec ttreat compare generic with the better t reat ments Count % within s elec ttreat compare generic with the better t reat ments Count % within s elec ttreat compare generic with the better t reat ments Count % within s elec ttreat compare generic with the better t reat ments Count % within s elec ttreat compare generic with the better t reat ments selectt reat compare generic with the better treatments 2.00 balanc ed with 1.00 any transit ion control/gener and alt ernate ic treatment tes t questions 40 97 Total 137 21.4% 40.4% 32.1% 32 60 92 17.1% 25.0% 21.5% 49 47 96 26.2% 19.6% 22.5% 51 32 83 27.3% 13.3% 19.4% 15 4 19 8.0% 1.7% 4.4% 187 240 427 100.0% 100.0% 100.0% Crosstab futurer Participate in future surveys -reversed 1.00 would definitely not participate 2.00 would probably not participate 3.00 would probably participate 4.00 would definitely participate Total Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments Count % within selecttreat compare generic with the better treatments selecttreat compare generic with the better treatments 2.00 balanced with 1.00 any transition control/gener and alternate ic treatment tes t questions 6 0 Total 6 3.2% .0% 1.4% 12 2 14 6.5% .8% 3.3% 38 70 108 20.4% 29.2% 25.4% 130 168 298 69.9% 70.0% 70.0% 186 240 426 100.0% 100.0% 100.0% 10