In our earlier look at a potential Rick Perry-Greg Abbott gubernatorial primary contest, we assessed their relative strengths and weaknesses in a hypothetical 2014 Republican gubernatorial primary. A subtext of our argument was that the trial ballot — a speculative choice between candidates — has inherent flaws this far before an election. It’s a topic that comes up frequently, so we want to lay out some of the limitations of trial ballots and the hazards of over-reliance on the Perry-Abbott item in making summary judgments about their standing with GOP primary electorate.
Trial ballots, when asked close to an election with clearly established candidates after a long and full election season, with a reasonably defined set of likely voters, can very accurately estimate the actual outcomes. As we move farther and farther away from that ideal, the trial ballot becomes increasingly imprecise. Take for example our February 2011 survey in which we tried to assess the upcoming Republican primary election for U.S. senator. We included six named candidates, the option of “another Republican candidate,” as well as a “don’t know” response. Ted Cruz received the outright support of only 3 percent of the potential Republican primary electorate, 52 percent selected “don’t know,” and in the end, only two of the candidates whom we tested (Cruz and David Dewhurst) ended up on the actual ballot — even though there were nine candidates in the primary. At the time, the people we chose to test were those being discussed as potential GOP candidates for the Senate seat, and the reality is that this item was designed to gauge name recognition at that point in time — not to predict the eventual election outcome.
This points to a key distinction with trial ballots in particular, and survey questions more generally. The sine qua non of a good survey question is to measure a single attitude. Things begin to get less precise when we — usually unintentionally — measure multiple attitudes. In the example above, it was clear that most Texas Republicans hadn’t yet thought much about the Senate primary, but if they were forced into a voting booth at that moment, a plurality would have favored Dewhurst, mostly because they recognized his name. Tiny minorities would have supported other candidates, again, because of some level of recognition and positive association, while the majority was still persuadable and waiting for the campaign to educate them on their choices. The campaign, in fact, changed things over the course of a year, and Cruz is now a senator.
The Perry-Abbott trial ballot is more complicated than that 2011 question: As a matter of question design, it measures at least four different attitudes, and potentially some combination of them. Reactions that respondents could have might include:
- Perry Support: “I know who Perry and Abbott are, but I like Perry and would support him again.”
- Abbott Support: “I know who Abbott and Perry are, and I would support Abbott given that choice.”
- Anti-Perry Vote: “I don’t like Perry, I may or may not know who Greg Abbott is, but I will register my displeasure with Perry by supporting this other candidate.”
- Perry Name Recognition: “I don’t really have an opinion on this as the primary is over a year away, but I know who Perry is, so I will choose him.”
In other words, while it may look as though respondents are registering a comparative preference for Abbott or Perry, it could be that they’re obliging us pollsters with a response — a response given with extremely limited information that in no way reflects the eventual choice that they would make once the candidates are finalized and have campaigned for their vote. Additionally, we can’t define what the Republican primary electorate will look like over a year from now. Together, these problems saddle the estimate with biases and error. Add this: Among people who chose Abbott in the head-to-head item, 31 percent were unable to express an opinion about him — positive or negative — when asked.
What to make of the results? Not much, unless you put them in the context of the results of other findings that more sensitively assess attitudes at this relatively early stage of the election cycle — especially in a race that has no meaning in most Texans’ mind, and that might never actually happen.