Taking the Pulse:Polling Citizen Opinionon Tax and Bond Questions Presented to the North Carolina Association of County Commissioners 24 January 2008 By Mark Wm. Hertzog, Ph.D.
Contents • Why Poll? • How Polling is Done: • Questionnaire Development • Sampling and Sampling Error • Fielding: Interviewing and Quality Control • Post-Fielding: Topline Results, Reporting, and Public Release • Two Examples: Durham County 2007 • Last Thoughts
Why Poll? • Show your citizens that you care what they think by asking them. • Nothing provokes more citizen objections in a referendum campaign than the feeling they are not being consulted. • Test-drive potential tax and bond issues. • Certain measures may pass easily, while others will encounter strong opposition. • Anticipate Citizen Concerns • Find out underlying voter concerns and objections, so you can address them, either by adjusting the proposal to meet them, or answering them in your campaign. • Track Campaign Progress • Periodic tracking studies can show your citizen advocacy group how the campaign is doing, and what effect (if any) news reports, advertising and endorsements for both sides, and other issues are affecting voter attitudes and likelihood to vote.
Questionnaire Development • Basic Construction • Done in direct consultation with sponsors. • Consists of both “closed-ended” responses (yes/no, ratings on a scale, etc.) and “open-ended” verbatim responses. • Includes essential demographic questions, as well as attitudinal and behavioral ones. • Final draft is programmed into a computer-assisted telephone interviewing (CATI) system. • Research house will manually test the survey program multiple times, to find bugs and get them corrected. • Automated testing then is done with computer-generated mock data prior to final approval. • IMPORTANT: Assume 5 business days from final approval of the questionnaire to the start of interviewing.
Questionnaire Development • Sponsor Identification • Questionnaire either identifies the sponsor (e.g., “the Durham County government”), or is “blinded”—that is, the polling company is identified, but not the sponsor. • If the county is sponsoring the poll, identifying the sponsor is more likely to encourage cooperation. • If a citizen group is sponsoring the poll, it may be better to “blind” the survey, so as not to prejudice the respondents’ answers. • Rules of thumb • Need to be brief and to the point to maximize responses. • Ideally 10 minutes or less. • Response categories should be appropriate to the question asked. • Questions must not lead respondents to a “right” answer.
Sampling and Sampling Error • Sampling • Best source of sample: voter registration lists. • A randomized subset of these are selected to obtain telephone number matches from a commercial vendor. • Calls are made from the subset with phone numbers. • If voter registration lists aren’t available, use random-digit dialing (RDD) to residential phone numbers, plus screening questions. • How Many Interviews? • Minimum is 384 to obtain a sampling error of no more than +/- 5% overall. • Comparisons between two or more subsets of voters will have higher sampling errors. • To reduce sampling error to +/- 4%, need 625 interviews.
Sampling and Sampling Error • Assuring the Sample Looks Like the Electorate • Essential criterion: Talk to registered voters who are at least somewhat likely to vote. • Use stratified random sampling: that is, set minimum or maximum limits (quotas) of interviews from specific demographic categories, based on previous turnout and current voter registration. • State Board of Elections breaks out voters in recent elections by gender, race, age cluster, and precinct. • Voter registration lists include additional demographics: age in years, zip code, municipality (if any)
Fielding • Briefing/Training • Once the CATI-programmed questionnaire is finalized and the sample is loaded into the system, interviewer training and a client briefing are held. • You and the project manager brief the interviewers on the locality, the issues they’ll be asking about, important things to know about the community that respondents are likely to mention during an interview, and correct pronunciation of proper names. • The interviewers then read through the questionnaire aloud, and anything that needs correction can be pointed out. • Interviewing and QA • After the client briefing, interviewing commences. • During the interviewing period: • Quality assurance monitors from the fielding house listen in regularly on the interviewers and provide coaching. • The account manager and research associate also perform periodic monitoring to assure our quality standards are being met.
Post-Fielding • Preliminary Topline Report • Once the last interview is complete, you can receive a preliminary report of the “topline” results. • Data Management • Verbatim responses are read and assigned numerical codes for quantitative analysis. • Corrections are made to occasional miskeyed responses. • Data are formatted for analysis and tabulated. • Data Analysis and Reporting • The project manager analyzes the results, breaking out statistically significant differences between demographic groups. • A full report/presentation is prepared.
Release (or Not) of Results • How results are released • Public meetings, such as County Commission sessions, or open meetings of your citizen advocacy group. • Press releases (accompanied if desired by a news conference). • Internet publication. • When You Must Release Results • If the county is the identified sponsor of the survey. • Advantage of having a citizen group sponsor your poll: You don’t have to release results. • Disadvantage of having a citizen group sponsor your poll: You don’t have to release results. • If a survey is being taken, the local press, bloggers, and opposition campaigners will get wind of it. • You can hang on to results that would affect how you run your campaign. • However, releasing data selectively undermines the credibility of the results. • Releases must never be done out of context.
Durham County, May 2007 • Situation • Durham County was considering two amounts for a school bond: the preferred amount, $195 million, and a lesser amount, $150 million. • It also wanted to float three smaller bond issues. • Before the General Assembly session, it also wanted to evaluate support for alternative tax measures. • Method • Hertzog Research adapted a study design another research house had used for the county four years before. • Essential to that was stating, in the question, the estimated cost of each bond issue in property taxes on a $200,000 house.
Durham County, May 2007 • Result • Surprisingly, support was equally strong for a school bond at $195 million, and at $150 million. • Those who supported it at $150 million, also did so at $195 million. • Likewise, those who opposed it at $195 million also opposed it at $150 million. • Two of the three smaller bond issues had wide support; one did not. • Voters were amenable to a prepared-food tax, but not to a 1% land-transfer tax. • County Action • The $195 million school bond was placed on the ballot. • The two smaller bonds that had strong majority support were also placed on the ballot; the one with less support was not.
Question Wording Example $150 Million School BondQ4. Another possibility is a Durham Public Schools bond issue with fewer projects, but for a smaller amount, at the $150 million level. This COULD increase the property tax on a $200,000 home by $100 per year. Considering this possibility, how would you vote on a Durham Public Schools bond issue for $150 million? Q4A. Would you say you lean more toward voting “yes,” or toward voting “no,” on the school bond issue, or are you really not sure?Q5. (Asked only of respondents who answered “no” or “lean no”) What would your reasons be for voting “no” on the school bond referendum?
Durham Public Schools: $150 Million Bond Q4: School bond at $150 million • Whites were somewhat more likely than blacks to give “don’t know” responses. • As age increases, likelihood to vote for the bond decreases: the “yes” vote diminishes from 77% among those 25 to 34, down to 38% among those 65 or older. • Those with children under 18 at home are significantly more likely to vote “yes.” • Those with children enrolled in the DPS system are significantly more likely to vote “yes.” • Those who said they definitely will vote were less likely to be undecided than were those who said they just probably will vote. • A plurality of those who said they are well informed about DPS construction projects were no voters, while those who said they were somewhat informed supported the bonds nearly two-to-one.
Q5. Reasons for responses of “no” or “lean no” on school bond White women were more likely than other groups to cite the fact that they did not have children in the Durham Public Schools as a reason for voting no. Those who said they were somewhat informed about DPS construction projects were more likely to cite the cost to taxpayers than were either those who said they were well informed, or those who said they were not well informed. Durham Public Schools: “No” Vote Reasons
Alternative Tax Proposals Q9: Prepared food tax • Majorities of black and white voters support the proposed tax, while a majority of the other-races respondents oppose it. • Support is strongest in zip codes 27703 and 27705, and weakest in zips 27712 and 27713; while the “don’t know” share is significant in zips 27701 and 27707. Q10: Land transfer tax • Homeowners and renters differ drastically. Homeowners, the huge majority of respondents, oppose a land transfer tax by 25%-66%, with 9% undecided. Most renters favor the land transfer tax, 56%-31%, with 13% undecided. (In contrast, views on the prepared food tax were almost identical among both homeowners and renters.) • African-Americans are significantly more likely to support the tax than are whites and other-race respondents. This appears to be directly related to the higher share of home renters among black respondents compared with white and other-race respondents. • Support is weakest in zip code 27713, where 17% are in favor, 65% opposed, and 18% not sure. The smallest share of “no” votes is in zip 27701, where 32% are in favor, 48% opposed, and 19% not sure.
Alternative Tax Proposals Q11. Now let’s suppose the bond issues we’ve discussed were financed largely by a prepared food tax, or a land transfer tax, or both, rather than higher property taxes. Would this make you more likely, or less likely, to vote for the bond issues, or would it make no difference in your vote?
After being asked their opinions on each of the two tax proposals, we asked the respondents this question: “Under North Carolina law, even if the voters pass both of these tax proposals, the County could levy only one: EITHER the sales tax increase, OR the land transfer tax, BUT NOT BOTH.Regardless of how you would vote on these proposals, if you HAD to choose between them, which one would you rather see enacted—the one-quarter of 1% sales tax increase on non-food items, OR the four-tenths of 1% land transfer tax paid by sellers of real estate?” The answers appear at right. Preferences between the Proposals
Demographic Check: Sample v. Self-Report Just as occurred in the May survey, about 2-3% of all voters, who are classified in voter registration records as black, categorize themselves otherwise in the survey, most often “multiracial” or “some other race.” If voter registration records are used, rather than self-categorization, then 30.4% of respondents are black, and 2.7% belong to “all other racial categories.” This table compares the survey respondents to the turnout by race in Durham County in the November 2006 general election, and to the Durham County voter registration rolls as of 18 April 2007.
Durham County, August 2007 • Situation • After the General Assembly session, Durham County wanted to evaluate support for the two alternative tax measures: the 0.25% sales tax on non-food items, and the 0.4% land transfer tax. • It also wanted to see what effect placing an alternative tax measure on the ballot would have on support for the three bond issues. • Method • Hertzog Research adapted the study design from the May survey. • Essential to that was stating, in the question, the estimated cost of each bond issue in property taxes on a $200,000 house.
Durham County, August 2007 The “Before” Sequence • After establishing that they qualified to be interviewed, respondents were first asked how they would vote on each of the three bond issues on the November ballot. • Question wording was similar to that in the May bond survey, and specified the estimated cost per year in property taxes on a $200,000 home. • The order in which each bond was asked about was randomized. The alternative tax questions • Respondents were then asked about the two alternative tax proposals: the 0.25% sales tax increase on non-food items, and the 0.4% land transfer tax. • Each question specified the estimated first-year revenue, and the estimated effect on the property taxes on a $200,000 home. • Following this, they were asked, regardless of how they stood on either tax proposal, to name which one they would prefer if they had to choose one. The “After” Sequence • Finally, respondents were asked again how they would vote on each of the three bond issues, if either the sales tax increase or the land transfer tax (or both) were also on the November ballot. • The order in which each bond was asked about was again randomized
Durham County, August 2007 • Result • There was majority support for a sales tax increase. • The land transfer tax was deeply unpopular. • Surprisingly, support was just as strong for the $195 million school bond even if the sales tax increase were put on the ballot at the same time. • However, there was a decrease in support for the smaller bond issues if the sales tax were also to be placed on the ballot. • Both smaller bond issues still had majority support. • County Action • The sales tax measure was not placed on the ballot. • No serious public opposition arose to the bond issues. • The bond issues, as a result, passed with more than 80% of the vote in November. • Note that the dollar amounts of property tax effects on the “average” $200,000 home were not included on the ballot—as they were in our survey, in anticipation of opposition.
Alternative Tax Proposals Q4: 0.25% Sales tax increase on non-food items • Men are significantly more likely to support the sales tax increase, while women are more likely than men to be undecided on this question. • Voters aged 35-44 are somewhat more likely to oppose the tax than other age groups—but the tax enjoys majority support among this age group as well. • Remarkably, there are no other significant differences on this tax by respondent segment. Q5: 0.4% Land transfer tax • This measure does not receive more than 36% support from any demographic segment. • Women are more likely than men to say they are not sure about how they would vote on the land transfer tax, while men are more likely to be firmly opposed. • Opposition is greatest in zip 27707 and 27712, and least intense in zips 27704 and 27705.
Durham Public Schools: $195 Million Bond • NOTE: On this chart and those that follow, the term “Before” means the initial stand on the bond when the question is first asked, before any mention of the alternative tax proposals. “After” means the voters’ stands after the tax measures were asked about. We specifically ask how they would vote on each bond measure if either the sales tax increase or the land transfer tax, or both, should also appear on the November ballot.
Last Thoughts: Polling Challenges • Message overload • Many voters screen all calls, or refuse to answer calls from numbers they do not recognize, in order to avoid solicitation calls. • The cell phone challenge • An increasing number of people, estimated at 7% today, use (unlisted) cell phones as their primary phones, rather than a land line or internet-based phone. • However, these are predominantly more transient populations: students, immigrants, and newcomers to an area. • Dislike or distrust of polls • The proliferation of bogus polls, the misreporting or misuse of legitimate polls and politicians and media, and the belief that public officials use polls to determine their stands on issues, lead many in the public to have a bad attitude toward all opinion research.
Last Thoughts: When You Poll • Use the phone. Internet polls aren’t representative. • Internet usage is tilted heavily toward younger, higher-income people. • You lose those who aren’t online regularly—who often are among the most likely to vote. • Sample is self-selected. In a telephone poll using probability sampling, any registered voter is far more likely to be included. • Choose a research partner that cares as much about your project as you do. • Large research houses are geared toward ongoing relationships with very large commercial clients. • Your county may be large, but your polling needs are not. • Therefore, you will be treated as a small, one-off client, and your needs will fall through the cracks. • Choose a smaller house that specializes in custom research, and that understands the needs of your community.
Conclusion Questions and comments regarding this presentation may be directed to: Mark Wm. Hertzog, Ph.D., Principal Hertzog Research, LLC 601 Jones Ferry Road, Suite N7 Carrboro, NC 27510 (919) 960-3904 firstname.lastname@example.org