Posted October 28, 2016on:
Pull on your high top waders. Tie on your chullo. Carbo-load. We are about to be hit by a polling tsunami – an experience similar to being an over-caffeinated mole in a free Whack-a-mole game.
Each day can bring sixty new polls. It is just an avalanche of data overload. How do we make sense out of it all?
Let me help.
First, Hillary is going to win the election. She knows it. Her people know it. The data nerds know it. And most importantly, Trump and his people know it.
The only “unknown” is by how much. That is important. The control of the Senate and the working majority of the House hang in the balance. So it is vitally important to vote.
Second, if you are interested in polling keep on reading. If not, just skip to the “Comments” and tell me how much you enjoyed this post.
The public polling we are used to seeing is a lagging indicator of attitudes. To explain, let’s say we were going to conduct a survey. The initial survey would be designed, reviewed, tested, redesigned, re-reviewed, and retested before it was ever fielded for the first time.
But before it was fielded for the first time, the sample would have to be constructed. This means the sample’s configuration for the phone numbers would have to be designed, purchased, delivered, and processed from a data supply house.
Once these and many other steps are completed, the survey is fielded. Typically, it takes three days of calling to achieve a sample of 1,040 responses. For most reputable survey companies, calls are only made from 6:00 – 9:00 p.m. prevailing time. In other words, calls stop in the Eastern Time zone just as they are starting in the Pacific Time zone.
Here’s the kicker – in 2012 the survey completion rate was only 9%. Simply put, it takes ten calls to get one complete response. So it isn’t uncommon to buy sample of 20,000 phone numbers because you over-sample in order to make sure you have fully captured the correct demographics required by the methodology.
Once a sufficient sample is captured, the raw data is then scrubbed by data analysts. This can take from a few hours to a few days depending upon the sophistication of the pollster.
Only at this point is the data run for analysis. Once the data run is complete, it is analyzed by the scientists and the cross-tabs are reviewed. A report is then written. Polls routinely generated by the news networks are more systematic, but are still on a weekly turn-around schedule.
This is from the poll taker standpoint, here’s what happens from our standpoint as consumers of news events. When we experience news reports, we have an initial reaction. We take it in, run it against our experiential knowledge, figure out if it makes sense, and then run it against our personal set of biases – ethics, morals, values, philosophies, etc. Then we forget most of it. Then we listen to acquaintances or pundits tell us how to feel about what we heard. We forget some of the opinion, appropriate some as our own, and then go on about our lives.
A few days later, we get a call from a pollster who asks us what we remember and how intensely we feel about that memory.
Because of all this, there is a lag time of about a week to ten days before opinions are captured in public polls. In the throes of a campaign, this lag time is compressed because the questions are abbreviated and the samples are smaller. Campaigns have polls in the field virtually constantly so they can register and react to events on a daily basis.
So what was the point in explaining all that? To lay the framework for the back story of the rather surprising level of candor captured in this Bloomberg piece about the Trump data shop.
Almost every public and private metric suggests Trump is headed for a loss, possibly an epic one. His frustrated demeanor on the campaign trail suggests he knows it…Several things jump out. Despite Trump’s claim that he doesn’t believe the polls, his San Antonio research team spends $100,000 a week on surveys (apart from polls commissioned out of Trump Tower) and has sophisticated models that run daily simulations of the election. The results mirror those of the more reliable public forecasters—in other words, Trump’s staff knows he’s losing. Badly. “Nate Silver’s results have been similar to ours,” says Parscale, referring to the polling analyst and his predictions at FiveThirtyEight, “except they lag by a week or two because he’s relying on public polls.”
The article refers to the “dark art of voter suppression”. Honestly, it isn’t so much voter suppression as “voter dissuasion”. Drumpf is attempting to dissuade young white liberals, women, and African-Americans from coming out to vote for Hillary. As a weaponized political science ploy this has only been around for about thirty-years. No one knows if it will work in a national election since it has never been tried on such a scale.
Some people believe it could have a reverse effect and drive more Hillary voters to the polls. In any event, the Trump data team is being managed by a guy who was an inexpensive tech journeyman known to the Trump family for his cut-rate webpage design work. Literally, this is his first time at the campaign rodeo. If that wasn’t enough, the data team works in an office festooned with “inspirational quotes from Sheriff Joe Arpaio and evangelical leader Jerry Falwell Jr.”
The clamorous Yam likes to point to the Brexit vote as proof that polling can be wrong. As with most things the Yam says, this would be true if it wasn’t false. The Brexit polling was correct. As many polls had it passing as had it failing in a close election.
In any event, if everything goes perfectly for the Orange Marmalade Skin Puppet and he wins all the toss-up states, he will still be short of 270 electoral votes. To analogize: The ambulatory Pumpkin Navel Lint has to catch lightning in a bottle, turn it loose, and catch it again while blindfolded to stop Hillary from being the 45th President of these United States.
Ain’t gonna happen.
What’s on your mind today?
70 Responses to "Catching lightning…"
Comments are closed.