Simulating the Syria Resolution Vote in the Full Senate

Yesterday, the Senate Foreign Relations Committee voted 10-7 to approve President Obama’s request to conduct military strikes against Syria (one member of the committee–Ed Markey–courageously voted “present”).

With the Syrian resolution clearing the committee stage, it now heads to the full Senate. But the question is: Will the full Senate pass the resolution?  We will get an answer next week.  But in the mean time, I took a stab at this question by simulating the full chamber vote (spoiler alert, the estimated vote is 56 for – 44 against).

Now, there are some strong caveats that accompany this post.  Namely, this is a difficult question to answer with any certainty.  First, the committee vote didn’t fall along clearly identifiable lines.  For starters, the vote (somewhat) split the parties.  Two Democrats voted against the resolution (Tom Udall and Chris Murphy) while three Republicans supported it (Bob Corker, Jeff Flake, and John McCain).  Relatedly, the vote crossed ideological lines as well (but as with party, only “somewhat”).  What seemed to matter most are state demographics, reelection, and tenure.  For example, if we eyeball the data, the state normal vote–capturing how Obama performed in 2012–seems to have had a meaningful effect.  Perhaps most notably, Tom Udall is up for reelection and represents a state Obama narrowly won.  I’m positing that both factors are key to explaining yesterday’s vote.  Finally, the length of service in the Senate appears relevant.  Looking at the data, the average number of terms served by those opposed to the resolution is 1.3 while the average terms served by those in favor of the resolution is 4.5.

Nonetheless, the patterns are difficult to eyeball, hence a multivariate analysis is needed (see also Ed O’Keefe’s post at the Washington Post) .  Details on the methods are at the bottom.  Logit results in the table the below (1=vote for the resolution, 0=vote against).

Untitled

While it’s a little surprising to see most coefficients turn up statistically significant given the limited sample size, each one is in an intuitive direction.  (note: thanks to Brendan Nyhan for pointing out the issues with two clusters.  As he notes, however, this has no effect on the predicted probabilities).

The most substantively important factor is the interaction between the state normal vote and reelection.  The model predicts that a handful of vulnerable Democrats will oppose the resolution in the full Senate.  Indeed, and as Joshua Tucker argues, while national public opinion is unlikely to affect the decision to attack Syria, public sentiment may matter at the state- or district-level.  As an aside, this dynamic may explain whether (when?) the resolution fails in the House.  While the ideological and partisan dynamics of the vote are somewhat tenuous, House Republicans in conservative districts can be expected to vote no–in part–because they’re all running for reelection (unlike in the Senate, what with its staggered terms an all).  Fancy modeling aside, there appears to be tenuous support for the resolution–at best–among Republicans in the lower chamber.

But second, the number of terms served in the Senate seems to have had a substantively meaningful effect on the vote as well.  This could represent a number of things (most likely seems to be an establishment vs. outsider effect).  Thirdly, party mattered somewhat, with Democrats more likely to vote for the resolution than Republicans (once we control for other factors).  This is unsurprising.

Finally, and perhaps most importantly, I used the above model to simulate the vote in the full chamber.  Based on a senator’s predicted probability (table below, right column), I estimate that the resolution would pass 56-44.  While this is a simple majority, it is not a filibuster proof supermajority.  It’s unclear whether conservatives will filibuster the resolution.  However, Rand Paul walked back this possibility during yesterday’s markup.

Interestingly, the predicted probabilities show a number of Republicans joining Democrats and vice versa.  For example, the model predicts that Democrats Mark Pryor and and Mark Begich will vote “no” in the full chamber while Republicans Chuck Grassley and Orrin Hatch will vote “yes.”  Indeed, both Pryor and Begich are up for reelection in “red states.” In sum, the model predicts that 17 Republicans and 39 Democrats will vote for the resolution while 29 Republicans will join 15 Democrats in opposition.

Here’s the predicted probabilities.  The data are sorted based on a senator’s estimated probability of voting for the resolution in the full chamber (lowest at the top).

Senator State Party Pr(Vote Yea)
Jim Risch Idaho R 0.00
Ron Johnson Wisconsin R 0.00
Rand Paul Kentucky R 0.00
Marco Rubio Florida R 0.00
John Barrasso Wyoming R 0.00
Tom Udall New Mexico D 0.00
Chris Murphy Connecticut D 0.00
Mike Enzi Wyoming R 0.00
Jim Inhofe Oklahoma R 0.00
Mike Johanns Nebraska R 0.00
Tim Johnson South Dakota D 0.00
Lamar Alexander Tennessee R 0.00
Pat Roberts Kansas R 0.01
Jeff Sessions Alabama R 0.01
Mark Pryor Arkansas D 0.01
Mitch McConnell Kentucky R 0.01
John Cornyn Texas R 0.01
Mark Begich Alaska D 0.01
Jay Rockefeller West Virginia D 0.02
Lindsey Graham South Carolina R 0.03
Saxby Chambliss Georgia R 0.05
Mary Landrieu Louisiana D 0.06
Thad Cochran Mississippi R 0.12
Deb Fischer Nebraska R 0.13
Mark Udall Colorado D 0.14
Ted Cruz Texas R 0.15
Tim Scott South Carolina R 0.16
Kay Hagan North Carolina D 0.20
Mike Lee Utah R 0.20
Max Baucus Montana D 0.21
Angus King Maine D 0.24
John Boozman Arkansas R 0.29
Jerry Moran Kansas R 0.30
John Hoeven North Dakota R 0.31
Carl Levin Michigan D 0.31
Jeffrey Chiesa New Jersey R 0.33
Roy Blunt Missouri R 0.36
Heidi Heitkamp North Dakota D 0.41
Rob Portman Ohio R 0.42
Mark Warner Virginia D 0.42
Pat Toomey Pennsylvania R 0.43
Kelly Ayotte New Hampshire R 0.43
Dean Heller Nevada R 0.43
Joe Donnelly Indiana D 0.46
Tammy Baldwin Wisconsin D 0.55
Martin Heinrich New Mexico D 0.55
Al Franken Minnesota D 0.56
Tom Coburn Oklahoma R 0.58
Roger Wicker Mississippi R 0.60
Elizabeth Warren Massachusetts D 0.63
Mark Kirk Illinois R 0.63
John Thune South Dakota R 0.64
David Vitter Louisiana R 0.65
Johnny Isakson Georgia R 0.69
Jeff Merkley Oregon D 0.70
Lisa Murkowski Alaska R 0.71
Mazie Hirono Hawaii D 0.72
Brian Schatz Hawaii D 0.72
Richard Burr North Carolina R 0.72
Mike Crapo Idaho R 0.73
Joe Manchin West Virginia D 0.76
Dan Coats Indiana R 0.78
Richard Blumenthal Connecticut D 0.81
Richard Shelby Alabama R 0.83
Ed Markey Massachusetts D 0.85
Sherrod Brown Ohio D 0.85
Michael Bennet Colorado D 0.86
Susan Collins Maine R 0.86
Jon Tester Montana D 0.86
Claire McCaskill Missouri D 0.87
Orrin Hatch Utah R 0.88
Bob Casey, Jr. Pennsylvania D 0.90
Amy Klobuchar Minnesota D 0.91
Kirsten Gillibrand New York D 0.91
Tom Harkin Iowa D 0.92
Sheldon Whitehouse Rhode Island D 0.93
Bernie Sanders Vermont D 0.94
Chuck Grassley Iowa R 0.95
Bill Nelson Florida D 0.95
Debbie Stabenow Michigan D 0.96
Maria Cantwell Washington D 0.96
Tom Carper Delaware D 0.97
Ron Wyden Oregon D 0.98
Chuck Schumer New York D 0.98
Patty Murray Washington D 0.98
Dianne Feinstein California D 0.98
Harry Reid Nevada D 0.98
Barbara Mikulski Maryland D 0.99
Patrick Leahy Vermont D 0.99
Jack Reed Rhode Island D 1.00
Jeff Flake Arizona R 1.00
Jeanne Shaheen New Hampshire D 1.00
Tim Kaine Virginia D 1.00
Bob Corker Tennessee R 1.00
John McCain Arizona R 1.00
Chris Coons Delaware D 1.00
Ben Cardin Maryland D 1.00
Bob Menendez New Jersey D 1.00
Dick Durbin Illinois D 1.00
Barbara Boxer California D 1.00

Methods:

The response is coded 1/0 (1 for, 0 against).  Normal Vote is the two-party vote for Obama in 2012 minus his national average.  Democrat is coded 1 for Democrats, 0 for Republicans.  Terms is a count of how long a senator has served in the Senate (logged).  The standard errors were clustered by party.  If a senator has a predicted probability of greater than 50%, they are estimated to vote “for” the resolution.  Any senator who voted for the resolution is the committee is assigned a probability of 1.0 while any senator who voted against the resolution in committee is assigned a probability of 0.0.

About these ads
This entry was posted in Filibuster, Legislative Politics, Senate. Bookmark the permalink.

4 Responses to Simulating the Syria Resolution Vote in the Full Senate

  1. MD says:

    A difficult vote to predict. But I applaud the effort; excellent post.

  2. Hanan Kolko says:

    How do you explain Bernie Sanders at 94% yes?

    • Jordan Ragusa says:

      The party id variable is coded as the party each senator caucuses with. Thus, as a Democrat, representing a liberal state, with a long tenure in the Senate, Sanders has a high probability of voting for the resolution according to the factors included in the model. That said, Sanders is certainly not a “typical” senator in this example and as a consequence the model will likely get his vote wrong. Thanks for the question.

  3. Hanan Kolko says:

    You’re welcome, and thanks for the thoughtful answer.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s