Election 2016 Forecast / Turnout and Result: “Likely Voter” Indicators that worked in 2008 and what it means for 2016
There are many stages to poll analysis. And, in each stage, an incorrect assumption can produce poor or misleading overall results. One of these stages is calculating Likely Voters.
Many polls include questions that seemingly have nothing to do with the candidates. They might seem more general. Some of these deal with determining if that individual is a Likely Voter and should/not be included in the poll results. Some basic questions might include:
The basic idea is that by looking at such responses you can more or less tell how likely it is that this respondent will actually vote in the election.
The 2016 election will not likely be a normal election in terms of models and expected turnout. On the Democrat side, we have the first female candidate for a major party which could produce, if relying too heavily on historical norms, one-off errors in forecasting. On the Republican side, we have the first ever non-politician / non-military running for president who has attracted, according to most reports, a following that includes many people who normally do not vote. On top of this we have two 3rd party candidates that have gotten much more attention than they did in 2012 when both topped the ticket for their respective parties.
There is no perfect precedent for the 2016 election given the number of unique factors. However, 2008 had some of the same first-ever groundbreaking elements to it. This was the election year which included Obama as the first non-white to top a major party ticket. There was a considerably high level of interest in the election from people who normally do not vote. This general level of excitement about the election and the inclusion of many people into the political system who had not fully participated in the past made 2008 a break-out year.
It should come as no surprise then that the metrics traditionally used to determine ‘Likely Voters’ did not work that well in 2008. Overall, it looked like McCain’s supporters were more enthusiastic and had the potential for a fairly large turnout. In reality, it was Obama’s supporters who surprised most by turning out in larger numbers and breaking many previous norms and records. It is of interest for us to look at which metrics worked in 2008 as there could be a similar outcome in 2016.
PEW Research has done an excellent job compiling 2008 data for analysis. This report shows the ‘Likely Voter’ indicators going into the 2008 election. The following table shows each general indicator and which side the indicator determined a higher likelihood
Table 1: ‘Likely Voter’ Indicators from 2008 US Presidential Election
| ‘Likely Voter’ Indicator | Candidate with Better Results |
| Voted in last election | McCain |
| “Always Vote” | McCain |
| Voted before in this precinct | McCain |
| Absolutely certain will vote | McCain |
| A great deal of interest in politics | McCain |
| Know where people go to vote | McCain |
| Given quite a lot of thought to election | Leans Obama |
| Following campaign news very closely | Leans Obama |
| Already voted (in this current election, early voting) | Leans Obama |
| Plan to vote early | Obama |
Source: PEW Research
Looking at this data, you would assume McCain would have the better turnout. By traditional metrics, his supporters were seen as much more likely to vote as they were more interested in politics, voted more frequently, knew where to go to vote, and even stated that they were certain to vote. In reality, Obama’s supporters might not have been traditional voters but they showed an unusual level of interest in this particular election, certainly more than you would expect for individuals who do not normally vote. It was this difference between following this election and what their modus operandi was that was the key indicator.
For 2016, we should look for similar indicators. As stated multiple other times, Republicans during 2016 seem to have higher enthusiasm concerning this election. The situation could be similar to Democrats in 2008 in that there are apparently many Trump supporters who are very interested in this election who do not normally vote. This is not exactly the same as 2008, but it is very similar.
Many analysts in 2008 were reluctant to accept indicators that pointed towards higher voter turnout for demographics that normally did not have historically high voter turnout. Specifically, we can point to minorities and the youth vote. These groups posed unusually high turnout and shocked many analysts. Then, in 2012, the assumption was that such groups could not go higher, but many did. For instance, in 2012, Black Women became the demographic with the highest voter turnout ratio. This would have been unthinkable if basing your analysis on historical precedent.
The main point here is that we are hearing very similar stories in 2016. Many political pundits acknowledge that Trump is disproportionately attracting a demographic that statistically does not normally vote. They are generally categorized as white males without college degrees. Unfortunately, this analysis often times degenerates into referring to them as uneducated angry white males. This is reminiscent of 2008 commentary that many of those supporting Obama belonged to demographics that normally did not vote in high percentages. However, in both situations, there was a relatively high propensity to ‘follow the election closely’ and to give ‘a lot of thought to this election’.
My impression is that political analysts need to learn a lesson from 2008 – namely, that in a non-traditional election year interest in this election is more powerful than historical voting behavior in order to forecast ‘Likely Voters’.
Looking at Gallup data for interest in the election, it clearly seems like Trump has an edge. The question asked was:
“How closely are you following the news about the 2016 presidential election campaign — very closely, somewhat closely, not too closely or not at all?”
The results of those following ‘very closely’ appear in the following table.
Table 2: Percent Following the Election ‘Very Closely’
| Republicans/Lean Rep% | Democrats/Lean Dem% |
Republican – Democrat |
|
|
Jan 15-16, 2016 |
40% | 30% |
10% |
|
Feb 15-16, 2016 |
44% | 33% |
11% |
|
Mar 16-17, 2016 |
49% | 38% |
11% |
|
Apr 15-17, 2016 |
48% | 34% |
14% |
|
May 13-15, 2016 |
47% | 39% |
8% |
| Jun 14-15, 2016 | 47% | 40% |
7% |
|
Jul 15-16, 2016 |
47% | 42% |
5% |
| Aug 15-16, 2016 | 46% | 36% |
10% |
Source: Gallup
This implied enthusiasm gap favoring Republicans has been fairly consistent throughout this election cycle. From the beginning of the primary debates we saw that the Republicans had an edge here. If conditions continue and if patterns from 2008 remain relevant, it seems like having high levels of interest in this election will be more important than historical voting records.
The other main indicator that tended to work very well in 2008 was early voting. We should start to receive early voting data soon and this will provide further insights into voter turnout indications for 2016.