You gotta love those Liberals at Public Policy Polling. They do their best to spin the numbers and influence public opinion with their polls. Many media outlets simply report the results without even bothering to let you know that the good folks at PPP are Democratic Pollsters.
That doesn’t mean that the polls are not valid, because the data is absolutely valid.
The interpretation of those numbers is where the term Democratic Pollster becomes relevant.
Most people assume that a poll simply calls people and reports the results. And while that is true, in order to make the poll actually mean something relevant to the topic being polled, especially in political polls, the results must be weighted.
So what is poll weighting?
I will try to make this as simple as possible in case a Democrat (or a Progressive) reads this. Logic and complex thought is simply beyond their abilities, bless their hearts!
If you ask 200 Democrats who they will vote for in the Virginia Senate race, Tim Kaine or George Allen, a very high percentage will say Tim Kaine.
And if you ask 200 Republicans the same question, George Allen will be the top answer.
But if you ask 200 people who are not a member of either party, the results will be entirely different.
So, how do you make any sense of this? Do we throw out the answers by Democrats and Republicans and only consider the Independents?
Well, that would inaccurate too because Republicans and Democrats will vote in the election.
So, why not just use statistics that tell us how many Republicans, Democrats and Independents there are in the state, and call them in the exact proportions?
Well, that’s one way to do it, but you would have to call a large number of people and not even use the data from a lot of them.
That is where weighting comes in. We know that currently , 39% of Virginians consider themselves Republicans and 36% are Democrats, leaving the remaining 25% as neither, usually called Independents.
So, the optimum sample of Virginians to give you a good idea of how the candidates might do if the election were held today, we would need to call 39% Republicans, because we know most – but not all – will pick Allen over Kaine. And we need to call 36% Democrats, most but not all picking Kaine. And 25% need to be neither party.
So how did this PPP poll do in correctly picking the right mix of Republicans, Democrats and Independents?
Well, they were perfectly on target for Democrats. Exactly 36% of the 600 people called said they were Democrats. So, they get an A+ for Democrats.
But they didn’t do so good with Republicans. Out of the 600 people they called, only 33% said they were Republicans, not the 39% we had hoped for. So the Republicans were under-counted by 6%.
Which also means that there were too many Independents. Instead of 25% Independents, they added the 6% that should have been Republicans! 31% of those called were Independents.
So, if you call the exact right number of Democrats (36%) and fewer Republicans (33% instead of 39%) you can expect a Democrat to do better than he actually will in the election, right?
So let’s “weight” these numbers so that they more accurately represent the voters in Virginia and see what this poll is really telling us.
Don’t worry about the math. I am using a calculator. And we will do a very simple weighting example so the Progressives can follow.
The PPP report said that they called 600 people. And we have the percentages of Democrats (36%), Republicans (33%) and Independents (31%).
Let’s change those percentages to real numbers. It makes it easier to see.
Of the 600 people called:
216 Were Democrats (600 times 0.36 if you want to check my math).
198 Were Republicans (600 times 0.33)
186 Were Independents (600 times 0.31)
Add those three up and we get 600. Good. The math is correct so far.
So when the PPP poll reported that Tim Kaine has 47% and George Allen has 42%, we can actually look at the numbers:
282 said Kaine (600 times 0.47)
252 said Allen (600 times 0.42)
That only totals 534, so 66 people were either clueless or were voting for someone else.
So, what do we know so far?
Well, we know most Democrats are going to pick Kaine and most Republicans are going to pick Allen.
We know that PPP called the right number of Democrats and too few Republicans.
And too many Independents.
So, the 282 people who picked Kaine had the correct number of Democrats and too many Independents.
The 252 people who picked Allen had too few Republicans and too many Independents.
So let’s do the correction, shall we? (Sorry Dems, this can get a bit complicated.)
Republicans were under-sampled by 6%. So let’s fix that.
6% of 600 is 36 people, most of whom would have picked Allen. So let’s add 36 to Allen’s total votes, which was 252. That gives us 288 votes for Allen compared to 282 for Kaine.
Which comes out to an Allen lead – 48% to 47%.
Now let’s look at the left wing spin in the poll results from PPP.
In four polls PPP has conducted this year on the Virginia Senate race,
Democratic former Gov. Tim Kaine has gained every time on George Allen, former
holder of the state’s other Senate seat and another ex-governor. The two were tied in February, then Kaine inched ahead by two points in May, to three points in July, and he now leads by five (47-42).
That would make you think that Kain is gaining ground. But if you look at the number of Democrats vs. Republicans called, every one of these polls has more Democrats than Republicans. And going by the last 2 elections, that is absolutely inaccurate. some of these polls had as many as 39% Democrats polled.
And then there is the “margin of error”. This is another calculation that means that the accuracy of the poll can be off by some calculated percentage. This PPP Poll’s MOE is 4% plus or minus which means that the unweighted results of the poll could be Allen 46% and Kaine 43% instead of the Kaine over Allen 47% to 42% as they reported.
PPP did the same thing in 2009 in the Virginia Governor’s race. At one point, PPP reported that Democrat Creigh Deeds had cut Bob McDonnel’s 14 point lead in half!
And the Progressives cheered at the wonderful news.
But PPP really fudged the numbers. Comparing the in September 2009) to their August poll, Deeds appeared to be mounting a comeback. What they didn’t tell you was that the August poll talked to more Republicans and the September poll talked to more Democrats. Here is what I wrote in 2009 about PPP:
Now, in looking at the internals in the poll, there are some glaring differences in the two sample groups. In the August 4 poll, they surveyed 35% Republicans and 32% Democrats. In the current poll, 38% were Democrats and 31% were Republicans. So, Republicans polled went from +3 to -7 – a swing of 10%. It seems that the only shift has been the in the sample. Deeds polls higher with women, and the latest poll surveyed more women (59%-41%) compared to the August 4 poll (55%-45%).
So, was Deeds really the comeback kid that PPP wanted to make the Democratic base think? Or were they just intentionally skewing the numbers to excite the Democratic base?
We all know that McDonnell beat Deeds by almost 18%.
My advice, take the PPP polls with a grain of salt. They are not usually what they appear.
If you really want an accurate poll, you need to look at Likely Voters, not simply Registered Voters as this poll does. They ask questions to determine how and when you have voted in the past. If you only vote in Presidential elections, then you are not very likely to vote in a state only race. And most polls of Likely Voters favor Republicans. Especially the last few years.
Roanoke College did a poll of Likely Voters for this race in September and it showed Allen leading by 3%. Rasmussen did one later in September and it showed Kaine up by 1%.
So, the bottom line in these polls is that this is a very tight race. I would call it even. Be careful of pollsters, particularly Democratic Pollsters like PPP. If you properly weight their results, the data is somewhat useful, but with a sample size of 600, none of these are going to be very accurate. I prefer a sample size of at least 1,000 and 1,500 is even better.
This poll called more Democrats than Republicans, which is not accurate at all and completely out of touch with the political landscape of Virginia. 30 more people in this Democrat skewed poll picked Kaine. The 5% “lead” is actually 30 people out of 600. This was done to give the left wing media the opportunity claim Kaine won the Associated Press debate.
The numbers have not moved. This race is dead even. Period.