Election polling media coverage typically focuses on the candidate ballot test because it’s what’s considered “news” and it’s an easy concept for journalists and voters to use to evaluate a candidate’s prospects in an upcoming election. Most people read the headline, see a candidate with a lead and go about their daily business. Yet, there are those who want to go beyond the headline that “Sarah Smith leads Jon Jones 45% to 39% with 16% undecided.” This guide is for this latter category of people.
Here is what I study when assessing the ballot test and what it may foretell for election day:
Margin of error: Let’s take the hypothetical Smith-Jones race above and the 45% to 39% ballot test results. Smith leads Jones by six percentage points, which is larger than the survey’s margin of error of ± 4.0%. However, the margin of error applies to both candidates, so Smith’s actual support could range from 41-49% and Jones’ could be 35%-43%, overlapping in the 41-43% range.
What you’re looking for: Is the lead more than double the margin of error? If it is, then it is likely statistically significant and therefore a real lead.Complementary questions: In a previous post, I wrote about four other polling questions that are worth reviewing in addition to the ballot test. Two of those questions can help you assess the ballot test results and put them into context: the generic ballot and incumbent job approval (if applicable).
What you’re looking for: Use these two questions to assess ballot support.The generic ballot measures the level of voter support for either a Democrat or Republican without named candidates. In our example, if Smith is the Democrat and sits at 45% and the generic Democrat receives 51%, then one might conclude the likelihood of Smith’s vote support increasing to a level that would win the contest is more likely given this additional information.
The incumbent job approval question measure just that – how many people approve or disapprove of the job the person is doing. If Jones is the incumbent and has a 43% job approval then their 41% current ballot standing suggests a much tougher time in getting to 50% + 1 vote to win; however, if Jones’ job approval is 49%, the task is still steep but perhaps more doable.
Party support (via crosstabs): Sometimes media stories will show you candidate support by party. If the pollster is fully transparent they will provide a link to the crosstabs that analyzes results by gender, age, political party and other demographic information. You can learn a great deal if you know where to look.
What you’re looking for: How well is each candidate doing with their own base? A general rule of thumb is that a candidate should garner support from at least 85% of their own partisans. So, in our example, Jones should be supported by at least 85% of the Republicans surveyed. A figure that falls well below that portends difficulty in winning unless Jones is overperforming with Democrats and independents. While you’re looking at the party support consider how independents are weighing in on the race. Does support for one candidate far outpace the other? What is the level of undecided among independent voters? Oftentimes a good indicator of the race lies with the independents.How many and who are the undecided voters (via crosstabs): The level of voters saying they are undecided may reveal where the race is headed. Knowing who these voters are provides insight into the race’s potential trajectory.
What you’re looking for: There are three important metrics.First, how many voters are undecided: with about a month to election day, most polls will have around 10% of the electorate saying they are undecided; a level significantly higher than that suggests the candidates haven’t broken through, haven’t locked down their own party’s support, or turned off independents.
Second, who are the undecided voters? Are they high propensity voters who truly haven’t made up their mind? Are they younger voters who tend to not vote as frequently in mid-term elections? Are they partisans who haven’t fully embraced their candidate? Understanding the makeup of these voters provides guidance on what to anticipate as election day nears.
Finally, if one of the candidates is an incumbent, their ballot support is below 50% and the undecideds are higher than 10%, that incumbent is in trouble. These voters aren’t undecided; they just aren’t ready to say they’re not voting for the incumbent or not voting at all. Generally speaking, at least two-thirds of undecided voters who actually vote will support the challenger, not the incumbent.
What is the trendline: If you have done all of the above, you’re deeply invested in this race. Take the next step and review the trendline – how the ballot has evolved over the course of the campaign. Many congressional and local races won’t have a lot of polling done, but statewide and national races almost always have at three or more measures along the way. If your news outlet doesn’t provide a trendline or hasn’t done extensive polling, check aggregators like RealClearPolitics or FiftyPlusOne.
What you’re looking for: How have the results changed over time? Are both candidates gaining support or just one of them? Is the percentage of undecided voters decreasing or staying the same? Is the gap between the two candidates closing? Examining the results across multiple surveys can tell you how the campaign is shaping up and where it might be headed.
Diving a bit deeper beyond the headlines and incorporating additional data points into your inquiry will add context (and perhaps clarity) to the ballot test results. Reading polling results with a new perspective should help you wake up the day after an election with less shock about what actually happened.
