G. Elliott Morris Profile picture
Data journalist & US correspondent, @TheEconomist. Author of STRENGTH IN NUMBERS: How Polls Work and Why We Need Them: https://t.co/c8nxYdF0c0. Blog: https://t.co/PE6cfXm5sd

Mar 13, 2018, 28 tweets

TONIGHT is the much-hyped #pa18 race between @ConorLambPA and @Saccone4PA18. Who’s going to win? Predictions and results: [THREAD]


Before 8PM rolls around, I want to cover what the data say to expect in this contest. This thread will start off running through my predictions, will cover how I’m projecting the race, and will turn into a live blog with graphics when the votes start to come in.

There are multiple ways to predict an election. Polls, geographic data, demographic data, etc are all useful. I’m going to run through what all these measurements could predict in #pa18. Spoiler: it’s a mixed bag for both Lamb and Saccone.

1) The polling average says that Conor Lamb is favored to win this seat, has him up 2%. A high-quality @MonmouthPoll pegged the race at Lamb +6%. But there’s a 13% margin of error on average special elec polls! Anything from Lamb +15 to Saccone +11 wouldn’t surprise me.

2) The partisan lean of this district is pretty stinking red. At Trump+20, it’s redder than nearly 2/3rds of House districts. Plugged into my House forecasting model, this estimates that Saccone should win by ~11-12%, with a large 19% margin of error (yikes). #pa18

2.5) If we combine district partisanship with polls, we get a clearer signal & better predictions, but a slightly larger margin of error than for the average than polls alone. This method has Republican Rick Saccone up 4%, but assigns just a 63% probability to his victory.

3) If we model #pa18 based on what we know from state legislative special elections since 2016, we should expect to a swing from the partisan lean of the district of roughly 14 percentage points. That projects that Saccone would win by 6%, but gives Lamb a 43% shot at an upset.

4) If we combine all predictions for #pa18 (see graph), we get a rather competitive picture. The average prediction is Saccone +5% with an average Lamb win probability of about 40%. Clearly there is some wide disagreement between methods, & ultimately this feels like 🤷‍♂️.

Let me put my #PA18 thoughts this way: Polling very well may be picking up on something that the normal data don’t indicate, but a Dem pickup here would be the third largest swing in a federal special since 2016 and nearly twice the average. Possible? Hell yeah. Probable? Hmm…

Whatever the #pa18 prediction may be, we’re darn close to votes rolling in now. Here’s my set of county-level benchmarks (including a cool new calculation of the share of the electorate in each area) that combines presidential and down-ballot races into one reading.

B/c there can be large variation in partisan composition and reporting rate *within* counties, I’ve also made a table of benchmarks at the precinct level (since the full table is 592 rows, this is just the 20 precincts with the closest benchmark). #pa18

I’ll also be running a live precinct-level projection system tonight that will update the pictured table. Should be fun! Of course I’ll be on my blog’s chat room too (a perk for subscribers so I can pay web bills — go here: patreon.com/gelliottmorris) to send graphics quicker.

Tonight should be fun. Whether my punny takeaway is “Silence of the Lambs” or “Saccone of shame” depends on the voters!

Predictions and live results for #pa18: thecrosstab.com/2018/03/13/pa-…

Follow @DecisionDeskHQ’s JMilesColeman and @B_M_Finnigan for parallel live results

T-3 hrs!

I’m also keeping my eye on TN #SD14 special election tonight. The Democrat running won just 25% of the vote in the 2016 contest. Trump won by 44 in 2016

Results will be at this webpage starting at 7 CST:


We’re less an hour away from polls closing in #PA18 and tn #SD14 special elections. My live precinct-level projections of the PA race are at this spreadsheet:


Different methods produce different predictions for tonight’s special election in #PA18. Read about my take:


It’s election night! 🎉🎉🎉

Looking good, Charlie

Minutes away from votes in #PA18 and TN #SD14 special elections now. A reminder that the votes we get early won’t necessarily represent the outcome of the election, but it will give us ammo to make our prediction. Look at this projection map from #ALsen to see what I mean:

Whatever happens tonight, keep this in mind: it’s the AVERAGE of special election swing that is informative of November, not the outcome of any one (or two, or three) races. The trend is very bad for Republicans and either outcome tonight won’t change that

JUST IN: Democrat Conor Lamb beats Republican Rick Saccone in #PA18 by massive 0.0 percentage point margin. Final results:

@ConorLambPA: 0%
@Saccone4PA18: 0%
JEB! 100%

Democrat in TN down 30%, a 20% swing from 2016 if it holds.


I do my best :)

2 minutes til first votes. Over/under on finishong this tex mex?

Still only 5 precincts in, so can’t make a good projection at all yet, but Lamb is beating his precinct-by-precinct Allegheny benchmarks by 9%

GOP drop-off in vote share already pretty damning. Will see if it holds. If so… Lamb will win

Precincts reporting right now in #PA18 are heavy Clinton areas. Looking good for Lamb, but Saccone could benefit from nationalized environment. Doesn’t look like trend says that’s in the cards, however.

Note: Looks like my PA-18 thread broke here. Continue reading at this tweet:

Share this Scrolly Tale with your friends.

A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.

Keep scrolling