Princeton Election Consortium

Innovations in democracy since 2004

Sep 18: Biden 352 EV (D+5.3% from toss-up), Senate 52 D, 48 R (D+3.2%), House control D+2.0%
Moneyball states: President NV AZ PA, Senate MT KS AK, Legislatures KS TX NC

Senate outlook: 45-50 Democratic/Independent seats (median=47)

November 2nd, 2018, 10:47am by Sam Wang

If systematic polling error is no larger than past midterm elections (median of 3 percentage points, as high as 5 points), control of the Senate appears to be a difficult lift for the Democrats. The polling error would have to be at least 5.0 points (that’s how the Meta-Margin is defined).

They have a much better shot at achieving a 50-50 tie: in that case, the error only has to be 2 percentage points. More on that below – but first, the aggregated data in all its glory.

Because the Meta-Margin has been stable for about two weeks, I created a table showing polls with an end date of October 19 or later. Poll medians are given with standard error of the mean (SEM).

Then the lead is converted to units of sigma by combining the SEM and median historic error (3 points) into a single quantity, sigma. Sigma therefore includes not just sampling error but also systematic error, i.e. how far off pollsters are about who ends up voting. Conceptually, a lead of 2.0 sigma approximately corresponds to the margin of error in a single poll.

State Leader Median +/- SEM (n) # sigma lead
MT Tester D+8.0 ± 2.1% (3) 2.2 sigma
NJ Menendez D+5.0 ± 1.0% (3) 1.6 sigma
FL Nelson D+2.0 ± 0.7% (17) 0.6 sigma
NV Rosen D+2.0 ± 3.1% (8) 0.5 sigma
AZ McSally R+1.0 ± 1.3 (9) 0.3 sigma
MO Hawley R+1.5 ± 1.2 (4) 0.5 sigma
IN Braun R+2.0 ± 1.7 (6) 0.6 sigma
TX Cruz R+4.5 ± 1.0 (6) 1.4 sigma
TN Blackburn R+6.0 ± 0.9 (6) 1.9 sigma
ND Cramer R+9.0 ± 2.3 (3) 2.4 sigma

Within the 2-sigma margin of error are a whopping eight races. This range includes everything from a Menendez loss in New Jersey to wins in Texas and Tennessee for O’Rourke and Bredesen. However, these extreme possibilities don’t seem realistic. Voter turnout should be high, making it a little easier for pollsters to gauge likely voters.

Seems to me that a more realistic range of uncertain races is plus or minus 3 percentage points: Florida, Nevada, Arizona, Missouri, and Indiana. In that case the range of outcomes is 45 to 50 Democratic/Independent seats. The median outcome is 47 D/I seats, a net change of two seats. Good thing for the Republicans they are running in such red states this year.

A 50-50 tie for control of the Senate is possible if Democrats win all of these close races. That’s not crazy at all because close races tend to break the same way – usually in the direction of the winning party. If that outside possibility came to pass, I wonder whether there would be power-sharing in the Senate, as there was in 2001, or whether the Republican caucus would use Vice-President Pence as the tiebreaker to maintain control.

For Democrats, the ActBlue link at left has been updated to reflect these close races. For Republicans, the NRSC page is still a good place to give.

Tags: 2018 Election · Senate

9 Comments so far ↓

  • Larry Guy

    Good to see you back with insightful analytics!

  • Andy

    “If that outside possibility came to pass, I wonder whether there would be power-sharing in the Senate, as there was in 2001, or whether the Republican caucus would use Vice-President Pence as the tiebreaker to maintain control.”

    I think we all know the answer to this.

    • ArcticStones

      Yes, I think we do know the answer. In the current Senate, McConnell & Co have been perfectly willing to depend on Pence’s tie-breaking vote, rather than steer a more moderate course and seek compromise with the Democrats.

      In addition, the Blue Slip tradition for federal judges has quite simply been buried.

  • Tony Roberts

    Just two weeks ago, you all but promised Sinema and Rosen would win. When will you learn to stop leading your following over the cliff? This is becoming a joke! Why don’t you just keep your trap shut and let the numbers speak like Nate Silver does instead of toying with peoples’ emotions, even if unwittingly? I could give a rat’s bleep about your bug-eating promotions. What I care about, and I think I speak for many, is not being fed false hope!

    • Sam Wang

      Wow, hi, Long time.

      You’re doing this all wrong. The idea is to find races where your effort makes a difference. If you’re just sitting around looking for reassurance, you’re not doing your job fully as a citizen. Go make a difference in the world.


      Tony, you don’t speak for me. If you don’t like Sam’s stuff, you can always stop visiting the site instead of just being rude.

      I don’t know about false hope, but I haven’t given up on anyone from Nelson to Beto in Sam’s list. Even if we don’t get the Senate back it would be nice for the country to have someone like Beto win. I just made a small donation to his campaign. Not much else I can do from Wisconsin.

    • Joe Bland

      Elections are a moving target and only generally predictable. You should have read the fine print.

      BTW, Sam, I thoroughly enjoyed your recent CITP lecture. Great stuff!

      Sacto Joe

  • Marconi Darwin

    The way it is shaping up, Democrats would be lucky not to have fewer than 48 seats in the Senate. And what that means is that Mitch McConnell would be likely to be the Senate Majority leader till at least 2022.

    Sure, in 2020 there are more Republican seats in play, but just about every one of them is in a deep red states with margins of + 5 points or more.

    There are two in play: Susan Collins and Cory Gardner. Even if both of them flip (unlikely) and Doug Jones holds on to ihis AL seat (unlikely), after the 2020 elections, Mitch will have 50 Republicans if Drumpf wins again.

    And he looks likely given there is not a Democrat in sight capable of challenging him.

Leave a Reply to Andy (Cancel)