Election Polls: Why so different?

Donald Trump and Hillary Clinton

Once again the time has come for the general public to pay attention to arguably the greatest Right in the modern world. Voting to change the ruling Government, without violence, riots or coup attempts, is relatively uncommon for the vast majority – in part due to the fact that the process is supposed to be transparent and fully vetted. Putting aside those two requirements, especially as they apply to the 2016 presidential election, one of the major factors originally meant as a means to motivate voters into action is polling. A means for campaigns to gauge how well their message has been received.
As the past 3 or 4 decades have technologically grown, the purpose and reaction to polls has changed. In that timeframe political messages have devolved to 30 second soundbites in commercials, and 2 minute summations on cable news programs with many seeing polls as just another means to promote specific political agendas. While cable television reached more Americans, and provided 24 hour access, the content of that wider greater reach diminished substantively. This prompted a greater and greater need to poll to determine feedback on the smaller less defined politics as well as to support it. Then came the internet.

Near instantaneous access to virtually all recorded facts and data pertaining to any specific issue and political conundrum over centuries of society were at the literal fingertips of the public at large. Innovations like Twitter gave immediate feedback on news and events. Social media allowed for access to the globe like no other medium, prompted by the desire and interest of masses of people. Youtube ensured that potentially every word, campaign promise, and event were available to absolute confirmation of the truth. Thus, predictably, the American public (like most in 1st world nations) became addicted to videos of kittens, and social media brawls over reality television programming.

The necessity of polls is even more a matter of importance in a world of near instant response than ever before. Voter apathy has risen to levels normally only seen at times of worldwide depression and war (1916 – 1928, 1940 – 1944). That is before the increase in immediate access with a concurrent decrease in trust in Government (below 50% on domestic issues since 2004, international issues since 2003).

This backdrop being said, there is no lack of polls available to the public and politicos. Direction of the nation (right track below 50% since before 2009) is but one example. Rassmussen and Gallup are leaders in an industry devoted to constant polling updates an a series of categories. During an election year additional specific polls are created as well. But the differences in these polls can vary widely, depending on the day taken; the question asked; what is included in the poll; and of course who (and how many) are asked.

One of the biggest differences is the last criteria – who is asked and how many people. The size of the poll affects accuracy of the poll. The more people asked the more likely that accurate findings are possible. The time to accurately poll even a congressional district is logistically impractical for the demands of a near instantaneous society. An example is the New York 22nd Congressional District, which contains over 720,000 residents. Some 147,000 are Republicans and 138,000 are Democrats, giving the District a R+9 rating. Which is normally considered a safe haven for Republicans (and the same for Democrats in an equivalent District).

The Republican and Democrat NY-22 candidates
An August 14-16, 2016 poll by the DCCC on behalf of Legislator Kim Myers was based on 400 respondents, or 0.14% of the total registered voters, excluding any Party affiliation except Democrat and Republican. The NRCC on behalf of Assemblywoman Claudia Tenney likewise did a poll of 400 respondents in August 23-25, 2016.

Based on current statistical models, a mere 400 people are enough to claim an accurate poll sample. With the difference of a week between these 2 polls, based on the same number of voters, the expectation would be that a similar answer would be determined. That is not the case.

The internal poll by the DCCC determined the race was tied at 35% (140 people each), with a third candidate – Martin Babinec – getting 21% (84 people). The margin of error is 4 – 9%. But perhaps most critically, the poll targeted Likely Voters. Without knowing the questions asked, which were not released to the public, that is a critical factor.

By comparison, the NRCC poll targeted 400 general voters. That poll showed Martin Babinec at 23% (92 people), Kim Myers at 27% (108 people), and Claudia Tenney at 33% (132 people). The margin of error was again 4 – 9%. Thus like in the current 2016 presidential race, the polling data shows conflicting information.

How a question is asked is a big deal. But since both polls do not provide that, evaluation can only be done by the general facts that are clear. That is the number of people, the trend, and who was asked. The trend can be seen in our review of the NY Presidential primary, as well as primaries across the nation. Democrats showed up at record low levels, Republican at highs. Both presidential candidates have had political gaffes and negatives released since the respective Conventions, but if the national trend on turnout remains accurate then polls indicating a Republican win seem more credible.

Who was asked is harder, but perhaps more valuable. Polls targeting Likely Voters are based on Board of Elections data collected by the State, and available to all campaigns. Data can be compiled based on screening for individuals who have consistently voted (though the actual vote is not known) over the past 4 – 8 years or more. This would indicate people who will turn up to vote again in November. Much of politics is geared towards these individuals, who are a far smaller subset of all registered voters. These individuals, often older and less connected to the internet and social media, are also often the most likely to vote solely on partisan lines. Thus, it would be expected that a Likely Voter poll will favor the trend of the region the poll is taken in.

Thus the question of who was asked is important. How many were Democrats versus Republican or Conservative or Libertarians? Were voters in this poll (again using the NY-22 as an example) located in Broome County – the only region where Democrats outnumber Republicans in the District? Did it include significant numbers of people in Oneida (the 2nd largest concentration of Democrats)? How old were the respondents? This information would reveal a lot about the poll and therefore its accuracy, but that was not provided to the public.

A general voter poll, hits a far broader section of people. In many election years this is considered less useful beyond determining penetration of political messages as these voters cannot be assured to show up in November. In 2014, as an example, in New York State only 28.8% of voters were involved in the election process. That would primarily be Likely Voters. But a Presidential Election brings in 1-time voters. This is especially true in the 2016 race. It was the support of 1-time voters that swelled Republican primary polls for Donald Trump and provided much of the attention in the Democrat turnout for Sen. Bernie Sanders. In context, a general poll in 2016 could be more likely to consider the increase in voters and the direction of 1-time voters.

But there is another piece of information that is critical that should be addressed. Undecided voters. A figure that is often ignored as candidates promote polls showing support and minimize those that do not. According to the DCCC poll only 9% of voters are unsure or undecided on a candidate. The NRCC poll reflects 17% undecided. Just looking at Broome County alone, New York State data shows 20% of voters did not indicate a Party affiliation. An additional 7.8% were affiliated with Parties other than Democrat or Republican. For the entire NY-22, about 8.5% of the District belong to Parties other than Democrat or Republican, and 21% (106,845 people) have no Party affiliation at all.

Thus, it would appear, that the DCCC poll ignored in large part any person not affiliated with a political Party, and/or affiliated with any Party beyond the two majors. The NRCC poll, appears to more accurately reflect the composition of the NY-22, as of April 2016. In a non-presidential election year, the DCCC may be more accurate. But in 2016, given the presidential candidates and the trends to date, any poll not taking into account 1-time non-Party affiliated voters is likely incorrect to some degree.

This example gives some insight to why there are so many 2016 Presidential Election polls. The manner in which the questions are being asked, when the public is asked, and critically who is or is not asked generates the range of results that are common at this time. As the NY-22 general election polls show, ignoring voters that either have no affiliation or are affiliated with Parties other than the 2 majors skews the figures.

What may be the best way for the public at large to view political polls is to look at the overall trend of several polls, if available. Once more using the NY-22 as an example, this time considering the Republican Primary, 3 polls were released at least in part to the public. Each poll disclosed different data, and conclusions. Our analysis of the trend of the polls, in combination with news announcements from the campaigns, and related political data led to our conclusion. Yet, as we said at the time,

“Trends in polls are not always how voters vote. Negative ads do affect voters – which is why they have gained popularity – but they also hurt the candidate that uses them. Candidate performances in debates can sway voters on critical issues, like the Time Warner Cable News debate on the Trans Pacific Partnership question. Finally, last minute endorsements, like the NRA, can shift critical groups of voters in a single announcement.”

Thus, no voter nor pundit should put their faith in any single or group of polls. As much as candidates and their surrogates may use a particular poll as a part of their campaigning, it is ultimately little more than data at best, and marketing at worst. Nothing should replace a concerned voter reviewing all data and records important to their decision for making a vote. With the access to the internet being so immersive in daily life, there is no time in history better for voters to know what matters beyond a 30 second soundbite or commercial.

Rating 3.00 out of 5

1 Comment

  1. Claudia Tenney stands heads and shoulders above the other two candidates. Neither Babinec nor Myers can hold a candle to Claudia on any issue! Neither of them really care about the citizens or have a record of working for the good of the people of this District 22 like Claudia Tenney has!

1 Trackback / Pingback

  1. Playing fast and loose with internal polls | VASS political blog

Thank you for lending your voice. We appreciate hearing what you have to say.