Exceptionally Low Turnout Can Account For Polling Errors
At The American Prospect: Tweet In the home stretch, I wrote that midterm polling is far less accurate than in Presidential years. Today, in The...
A change to the banner
We’re trying out a change in the banner above. Previously, we showed a snapshot, i.e. what would happen in an election held today. Now, the only probability given is for Election Day. Why are we doing this?
The basic reason is that the election is not today. The day-to-day snapshot is extremely important because we use it to build a statistical model of the Senate race’s ups and downs. However, fluctuations in the probability tend to get people overexcited.
If you want to know about current conditions, focus on the seat count and the Meta-Margin. If you still want a current snapshot probability, you can still get it by clicking on the histogram at the right, which shows current conditions based on the most recent polls.
Topics:
I like it. I think it’s easier to focus on those two things. I’ve learned to look more closely at the meta-margin.
I think the meta margin is more revealing when looking for trend lines. Right now it appears the trend is toward the 50/50 outcome. When the trend line starts back up I’ll be more comfortable.
Podcast with Sam at
http://politicalwire.com/archives/2014/09/24/conversation_with_sam_wang.html
–bks
Any chance of putting an error bar on that Election Day probability? One that would represent an estimate of systematic uncertainties in your methodology?
I think of this as a potential teaching opportunity. I believe that, in general, overly precise numbers lead the innumerate to believe that we know more than we actually do.
It would also be an additional way to contrast PEC with 538, which quotes their probabilities to an absurd 3 significant digits.
I’ve thought about it. See this. However, the sigma is quite possibly higher than I thought; see this. That concerns me a bit. It does look like an asymmetric error bar.
If we get to a point where our senate predictions read “democrats have a 70% chance of taking the chamber but this statistic only has an 80% chance of being valid but the statistic on the statistic is 50% probable,” I’m going to start having flashbacks to Mrs. Nickels, the greatest math teacher who ever lived (Sorry, Sam) but also the most maniacal test-writer in the history of public education.
“Fluctuations in the probability tend to get people over excited.” Oh, how many times have I heard these words ring forth from statistician lips? Hum…actually…never…yay, math is exciting again!