Investor’s Business Daily bills themselves as the most accurate pollster in the past elections. And this story backs up their claim. And I like tracking polls that “weight” their data to make it more realistic and diminish the error due to over or under sampling various demographics. And this IBD poll is weighted. Which is a good thing.
“Weighting” a poll involves a complicated formula intended to make up for errors in a random sample poll. In other words, if you call too many Republicans and not enough Democrats, the poll will be skewed more towards Republicans and it will not give an accurate representation of the opinions of the voters. Polling companies use data from past elections, census data, voter registration and things like enthusiasm to calculate some of the Demographics. We know the percentage of the population by race and a person’s age, income, education can be asked as poll questions. So many pollsters look at the 2008 and 2012 Presidential elections and use that data and today’s electorate to attempt to arrive at a reasonable profile for the 2016 election. And they “weight” the poll accordingly.
You have probably heard people talking about how polls are skewed towards Democrats in numbers that people do not think is reasonable. In 2008, Democrats outnumbered Republicans at the polls by +7%. Which more Democrats than Republicans turned out to vote for Obama. In 2012 that number was +6%, slightly lower, but in Obama’s favor.
Most people think that Hillary Clinton will not see as many people coming to the polls next week as Obama got in past elections. There is far less excitement and enthusiasm for Hillary than Obama had. What the exact turnout will be is only a guess until after the election. So that is the Holy Grail pollsters try to predict.
Some polls have come out predicting that Hillary will more than double the Democrat’s advantage over Republicans. And these polls, of course, show a much larger lead for Hillary than other polls. My personal opinion is that Hillary will have an advantage with more Democrats turning out than Republicans, but not as much as Obama saw in 2012. So a +3 or +4 Democrat advantage seems reasonable to me.
The IBD poll at hand over-samples Democrats by +7 or +8. It averages +7.3% more Democrats than Republicans. I think that is a bit high, but as long as the poll stays in a narrow range it is very useful to see movement in the election. Like me, if you think the Democrat advantage should be lower, then Trump is actually a couple of points higher than they show. But the trend as far as who is rising and who is falling will be pretty accurate, though the final results may be higher or lower than their actual results.
And IBD was doing a good job of keeping their weighting steady right up until the latest poll released today.
The big question on everyone’s mind is how the news from FBI Director Comey about reopening the Clinton Email case is going to effect the election. The Clinton spinners are out in force trying to minimize the fallout. They are suggesting that this is nothing new, that the email case was already “cooked” into her numbers, and a lot of screaming and whining about this unprecedented release of bad information this close to an election. I would submit that what is unprecedented is a Secretary of State completely ignoring our laws on handling classified data and not even having an email address on the State Department servers.
So what did IBD do that I find so egregious? With all of the attention being given to polls sampling too many Democrats and too few Republicans, IBD took another course that will also favor Hillary in this most recent poll. The pro Democrat skew was still +7, but they lowered the number of Independents in the sample. A recent WikiLeaks email from Hillary friend and ally John Podesta has the playbook the Democrats use to “rig” the polls. So what I found makes me very suspicious of the sudden change in weighting for this poll.
Why does this matter? Well, look at the graph below:
As you can see, there are two groups of lines, 3 at the top and 3 at the bottom. IBD began providing the weighting on the poll taken 10/20 through the one taken yesterday. Prior to that, the numbers were not available.
Let’s look at the top three lines first. These are the numbers of Democrats (in Blue) Republicans (in Red) and Independents (Brown). For the first 10 days, the weighted poll sampled 37% Democrats with only one day dipping the number of Democrats to 26. Independents started out at 34% and on 10/25 they dropped the Independents to 32% for the first 10 polls. And the Republicans (Red) started out at 29% and rose to 30% the first 10 days.
But look at the poll taken on 10/30 and released today. We went from a relatively flat sample for 10 days and suddenly the number of Republicans and Democrats rose 3% each, keeping the skew the same at +7, but the number of Independents dropped 5%. Why would they do that?
As I said earlier, the real value in a tracking poll is that the data samples are steady after weighting, or pretty much so. And the poll taken Sunday, after the Clinton Email case was reopened is anything but steady.
Now let’s look at the lower three lines. the bottom line (green) shows Hillary’s lead after each poll. the first 3 days, Trump was leading by 1 or 2 points (shown as negative numbers). On 10/23 IBD said it was a tie. And from 10/24 until 10/28 Hillary went up steadily and peaked at a 4 point lead. On the 29th her lead dropped to 2 and the 30th (released today) it is down to just a 1 point lead over Trump.
The black line in the middle is the pro Democrat skew. It started at +8, went to +6 for one day and then remained at +7.
The top line (yellow/gold) is the advantage Trump is seeing with Independents. Trump went from +11 with Independents (31% to 43%) up to +16 and then started a pretty steep drop in Independent support for Trump. He was still +5 among Independents at the lowest point, but that was a big drop.
The reason for the drop was Trump’s statement at the final debate that he would wait until after the election to decide if he will accept the results of the election. And the Democrats trotted out another unsubstantiated claim by a woman Trump said something to her she didn’t like a long time ago. Both were hammered by the media and Independents tend to pay attention. And as Independent’s support dropped for Trump, Hillary’s lead (the Green line) rose.
And then, suddenly on 10/29 Independents came flooding back to Trump and he went from +7 on Friday to +12 on Saturday among Independents. And on Sunday, Independents were back at +16 support for Trump. And during this period that the Independents were returning to Trump, Hillary dropped from a 4 point lead to just 1 point.
But is that really a 1 point lead for Hillary? Or did IBD change the math by increasing the number of Democrats and Republicans equally at the expense of the number of Independents polled? Remember this is a weighted poll, so the pollsters at IBD changed their formula. And whether it was intentional to buffer the effects of the reopening of the Clinton Email FBI case, or something else I cannot say. But I would imagine that the dramatic uptick in Independents for Trump raised a lot of eyebrows.
And if you look back to 10/21 when Trump enjoyed a +16 advantage among Independents, he had a 2 point lead over Hillary. Now, with that same 16 point advantage, he is behind by 1 point?
The only ratio that changed was the weighted sample of Independents. Lowering the percentage of Independents by 5 points who were +16 Trump (32 to 48) dilutes the numbers for Trump.
I personally find this movable weighting makes the IBD poll of no value to me anymore. And I really find myself questioning the fairness and intent of the poll. And I can think of no valid statistical reason to suddenly drop the sample size of Independents at a time when Trump is seeing a surge in Independent support. Except to help Hillary.
Until now, I have had no reason to question IBD’s poll as far as actually measuring public opinion rather than attempting to influence opinion. I think that their pro Democrat skew is a bit high, but as long as the weighting keeps things constant, this has been a very useful poll in detecting trends.
But after seeing this mathematical shenanigans, the IBD poll has just lost a fan in me.