Protected: Two ways to estimate primary outcomes without polls (transcript)
This content is password protected. To view it please enter your password below: Password:...
The Electoral Strike Zone (technical announcement)
Here at PEC, the likely range of electoral votes is set by assuming that opinion can drift in either direction, by an equal amount across states; and then converting that swing into electoral votes. This generates the red “strike zone” in the plots in the right sidebar. Today I make a small update.
When the previous relationship was established in June, few states were weak for Trump; indeed, in many states we were reliant on 2012 election results, in which Mitt Romney ran stronger than Trump is now. That meant that opinion swings toward Clinton would not change the electoral-vote (EV) estimate much. That led to an asymmetric strike zone. In fact, there was a whole region around 342 EV where multi-percentage-point polling swings led to no change at all. Full disclosure, it was not an ideal calibration curve.
Now that we have fresh polling data, it is possible to recalculate the relationship between margin and seats in the prior. Trump leads are fairly evenly distributed, including some states where he barely leads. The same is true for Clinton. Therefore a swing of opinion in either direction now leads to corresponding changes in EV.
Note that this change does not affect the November win probabilities at all. That probability treats a bare EV win and a blowout the same way. The exact number of electoral votes does not matter.
P.S. Yes, this aspect of the calculation – the conversion of Meta-Margin for purposes of November EV estimation – was not automated! It will be soon – one of several outstanding issues. I realize that it would have been more satisfying to see the change unfold gradually.
Topics:
Sam, I am a bit puzzled by your comment.
Certainly, each state has its own median and “meta-margin” needed to move it to one camp or another. I had always assumed that the “strike zone” was derived from overall national shifts in polling numbers assumed in the meta-margin shift plot, which gave rise to the asymmetric behaviour before today.
said another way, the variance in “national” poling gives us the variance final meta margin results and gives us the variance in EV. So no adjustment would have been required — it was already all in the data, so to speak.
Looking around at other aggregation sites where the state-by-state margin is a bit easier to get at, it still looks to be the case that a shift toward Trump would push a lot more states his way than the same shift toward Clinton. Is this not the case?
To repeat myself…it was already in the data, but the data changed.
Isn’t this calculated automatically? If you vary all the state margins fit today by the same amount (namely, the top (and, then, the bottom) of the standard deviation band on meta margin projection for election day) and then run the EV calculation on the higher margin numbers and then the lower margin numbers, doesn’t that show the correct standards deviation band for EVs?
That’s what I had assumed you were doing all along. This post makes it seem like you’re just eyeballing how symmetrical the EV deviation might be.
It is calculated using the same algorithms, but not afresh every day.
“Opinion can drift in either direction, by an equal amount across states” – is this a testable assumption? I would expect some variation in standard deviation across states.
Yes, I posted about this several months ago.
Comments are closed for a while.