Twitter | Search | |
Isaac Wood
As a former professional election prognosticator, here’s the three lenses through which I’m viewing the likely midterm election results. If you’re only looking at polls (or 538), you’re missing the full picture. (1/7)
Reply Retweet Like More
Isaac Wood Nov 5
Replying to @IsaacWood
One: Polls and poll-based forecasting models. I check polling averages of individual races and the generic House ballot on RealClearPolitics. I also keep an eye on the various 538 poll-based forecasts. (2/7)
Reply Retweet Like
Isaac Wood Nov 5
Replying to @kkondik @Redistrict
Two: Race-by-race predictions from teams with established track records. My old shop is the Crystal Ball at the Center for Politics where has kept up the great track record. The Cook Political Report’s predictions, led by , are also excellent. (3/7)
Reply Retweet Like
Isaac Wood Nov 5
Replying to @IsaacWood
Three: Macro-level political science modeling. These ignore individual races and look instead at factors like the party of the president, economic indicators, presidential approval, and the generic ballot. Here’s a summary of 2018 House predictions from the PS journal. (4/7)
Reply Retweet Like
Isaac Wood Nov 5
Replying to @IsaacWood
This year, there’s been some late hand-wringing about whether the conventional wisdom of a Democratic wave is right. If you’re interested in checking your assumptions, I’d recommend consulting all three categories of forecasts. (5/7)
Reply Retweet Like
Isaac Wood Nov 5
Replying to @AlanIAbramowitz
In 2016, some political science models had Trump winning (including the Time For Change model from ). That could have provided useful context to the Clinton-favorable predictions in other categories. (6/7)
Reply Retweet Like
Isaac Wood Nov 5
Replying to @IsaacWood
This year, though, the three signposts point roughly in the same direction, at least for the House: Democrats winning a majority in the House but picking up closer to the 23 seats they need than the 63 seats Republicans gained in their 2010 wave year. (7/7)
Reply Retweet Like
Paolo Lim Nov 5
Replying to @IsaacWood
But a problem I have with macro models is that it assumes all parts of the country move in the same direction even in differing magnitudes.
Reply Retweet Like
Paolo Lim Nov 5
Replying to @IsaacWood
But there is an inkling I have that POTUS and Rs may be doing well in more districts than we think but that this 'macro data' is obscuring that because it may be possible they may 'overrepresent' areas expected to go blue anyway.
Reply Retweet Like
Isaac Wood Nov 5
Replying to @paolo_lim
Absolutely. That’s why I recommend looking at a combination of micro (district-level polling and race-by-race ratings) and macro (generic ballot and political science models) indicators. This cycle I’ve actually seen an overemphasis of micro-level indicators so far.
Reply Retweet Like
Adriana Nov 5
Replying to @IsaacWood
Getting in first to say that I hope 23 is wildly under estimating the results.
Reply Retweet Like
Brendan Woodbury Nov 5
Didn’t the Time for Change model predict Trump winning the _popular_ vote in 2016?
Reply Retweet Like
Uhhh A Smoking Skeleton Nov 5
Replying to @IsaacWood
If so, because of gerrymandering. The overall vote percentage will absolutely blow 2010 out of the water.
Reply Retweet Like