A Tale of Two Polls
I wrote a piece a couple of days ago on understanding the Presidential Polls. Far too often, polls attempt to sway opinion rather than measure it.and trust me, I am constantly on the look out for that type of bias.
One of the things the polls like to do is show their favorite candidate ahead and the lead climbing between polls. But often, nothing has really changed, or their candidate (usually the Democrat) has actually fallen, but they sample more and more Democrats to make it appear that the opposite is true.
Let’s look at two brand new polls by CNN and CBS.
New CNN Poll
There is a brand new CNN poll out this morning that shows Donald Trump with a 5 point lead over Hillary if you add in the other candidates running. This poll was taken July 22-24 2016. Which means that this poll was taken completely following the Republican Convention, so it should show the convention “bounce”. Interestingly, there has been no convention bounce that lasted beyond a few days since 2000.
The CNN Poll shows Donald Trump with a10 point bounce going from Clinton +5 to Trump +5.
I am always suspicious of polls and like to look at the numbers behind the polls.
What I was looking for is if CNN is being consistent in their polls, or are they planning to drive some narrative? Did they sample more Republicans this time to show a Trump “bounce” planning to oversample Democrats after the Democratic convention to show a bounce for Hillary?
The answer is no! I looked at the last 3 CNN Polls and all three are amazingly consistent. While many polls are using turnout models predicting the turnout for Hillary will be better than the turnout for Barack Obama his last two elections, CNN is not. Their polls are almost evenly split between Republicans and Democrats with the GOP respondents a point or so ahead of the Democrats. Most polls, regardless of bias, are showing Republicans are far more enthusiastic this time and Democrats are not excited at all. So predicting a turnout model of slightly more Democrats – up to 4 or 5 points more, is not too unreasonable an assumption. CNN seems to be predicting a pretty even split.
We won’t know the exact D to R ratio until after the election, so no matter who is polling, there is always going to be a certain amount of guesswork.
Why I love this poll
What I really like about this CNN Poll series is that the ratio of Democrats and Democrat leaning Independents to Republicans and Republican leaning Independents has been consistent in (at least) the last 3 polls. Which means that while we may not know precisely the exact turnout on November 8th, this poll should be amazing accurate at tracking the trends. What I mean by that is if a poll samples 40% Democrats to 30% Republicans this week and then next week shows Democrats pulling ahead by 5 points over the previous poll, but they sampled 43% Democrats and 28% Republicans this time, are we really seeing an increase in Democrat support? Probably not. Had the samples been the same ratio, there would likely have been no movement at all.
But the CNN Polls have been a constant. Clinton had a +4 lead in the mid June poll and a +5 lead in the mid July poll. And that has now, after the GOP Convention, gone the other way. We now see a +5 point lead for Donald Trump. That is a 10 point swing towards Donald Trump. And the sample ratio of Dems and Republicans have been constant. So while we can’t be confident that if the election were held today Hillary would get 39% and Trump would get 44%, what we can say with a fair amount of confidence is that Donald Trump gained 10 points on Hillary in a week.
The CBS Poll – Not So Much Apples to Apples
So while the CNN Poll has kept a constant R to D ratio through at least 3 polls, the last poll being post GOP Convention, CBS has not kept things the same.
A poll conducted by CBS after the GOP Convention shows Trump with a 1 point lead. A poll they took July 8-12, prior to the Convention shows a Tie. So the lead story for the CBS Poll is Trump only got a 1 point bounce from the Convention.
But when we look at the weighted numbers, after CBS “corrected” for sampling errors, the post Convention poll sampled Democrats 4% more than Republicans and Trump leads by 1%. But in the prior poll earlier in the month, the “weighted” numbers favored Democrats by +6 points and showed a tie. And the one prior to that taken last month had a +7 Democrat advantage and Hillary was up by +6.
So the CBS Poll shows Trump moving from a 7 point underdog in June to a tie in July to a 1 point lead in the post Convention poll. That would seem to be an 8 point “bounce” since June, but none of these polls used the same sample ratio of D’s to R’s. So instead of an 8 point bounce in a month, Trump is probably looking at just a 5 point bounce.
But the real news in the CBS poll is that CBS itself is adjusting their Republican to Democrat ratio, moving them closer to 50-50 with each poll.
CNN is already expecting a 50-50 ratio.
It should be noted that both of these polls are of Registered Voters and not Likely Voters. LV’s tend to show the Republican candidate a bit higher. So that is good news for Donald Trump.
I will have to look at some other polls to see if they are creeping closer to the reality that Hillary is not going to inspire the type of turnout that Barack Obama had in 2008 and 2012. These pollsters watch one another and they all want to be remembered for nailing the election in the end, even the partisan players. Trending towards Trump this early is significant. CNN is looking at an even GOP and Democrat turnout which would probably mean a Trump landslide victory because the Independents are breaking for Trump and the trend has usually been that undecideds move to the party out of power the last few days.
Both the numbers and the trends of both the polls and the pollsters themselves are looking good for a GOP victory.