Saturday, June 23, 2018

BTRTN: Can The Dems Take The House in November? An Updated, Analytic Look

Tom with our second look at the 2018 House midterms.

THE HEADLINE

The headline:  the outlook for the Democrats taking control of the House in the 2018 midterms remains strong, but hardly certain.  We still have many primaries to go, five more months of unexpected national and international events, the ongoing Trump madness, and, of course, 435 head-to-head campaigns.  But at this point, the Democrats are in an enviable position.

Our track record in predicting the House is quite strong.  The BTRTN record speaks for itself – we have been exceptionally close in each of the last four elections.

House Elections
BTRTN Prediction
Actual Outcome
2010
R + 58
R + 63
2012
D + 4
D + 8
2014
R + 10
R + 13
2016
D + 5
D + 6

Our predictions are driven largely by our proprietary BTRTN House Prediction Regression Model, which uses relevant predictive data since 1970 (with a nod to daughter Allie).  More on that later.

But to get to the answer (and we will return with a full explanation), this BTRTN snapshot indicates that if Election Day were today, the Democrats would pick up +38 seats – with a low-end of +28 and a high end of +50 -- and win control of the House.

NOTE:  The goal of this article is to provide an accurate portrayal of where the race to control the House stands today.  It is NOT a prediction of how Election Day will play out.  It is a “snapshot” of how the House would look if the election were held today. 

Image result for election 2018 buttonSETTING THE STAGE

The Republicans hold a huge advantage in the House today, with 235 seats to the Democrats’ 193, with seven vacancies.

The main question in this race, of course, is can the Democrats secure a net gain of +25 seats to move from 193 seats to 218, which would be just enough to win the majority (218 to 217 for the GOP)?

What would it mean if the Dems re-took the House?  It’s hard to understate the magnitude of such a coup.  The Trump domestic policy agenda – already difficult with a fractured GOP and the need for 60 votes on much legislation in the Senate -- would literally stop dead in its tracks.  Nancy Pelosi (or her successor as Speaker if she stepped down or was voted out) would control what bills came to the House floor.  Assuming the GOP held the Senate (which is likely), any proposed GOP Senate legislation would face only the dimmest prospect of securing Democratic support in the House; these bills, therefore, will never reach Trump’s desk for signature.

Furthermore, the House committees would be chaired by Democrats, who would control what the Committee did with its time – including what it would investigate and what bills it would consider.  Democrats could, for example, initiate all sorts of investigations into the Trump Administration, which would contain daily revelations, choking the airwaves and blogosphere.  Thus far GOP committees have been extremely partisan and supportive of Trump; the committees that have investigated Trump have not used the full powers at their disposal to get at “the truth” – they have not called all relevant witnesses, nor have they used their subpoena power to get at relevant documents.  They have accepted “executive privilege” as an excuse from members of the administration to refuse to answer questions, even when such privilege was not warranted.

And, the House could begin impeachment proceedings.  Mueller’s ultimate charges may not carry through to a Senate conviction – 67 Senate votes would be required, which would mean, say, anywhere from 15 to 22 GOP votes.  But the showdown could be put in motion and debated for months on end, with little oxygen left for anything else.  To be clear, Democrats seem to be leaning against impeaching Trump, fearing “overreach,” Senate conviction failure and a backlash.  Mueller would likely need a true “smoking gun” for the Dems to move, something so convincing that it would attract near-unanimous GOP support.  Otherwise, the prospect of running against Trump in 2020 has far greater appeal than impeachment without conviction, which could actually strengthen Trump.

If the Dems took the House but the GOP held the Senate, Trump would have only one clear area of domestic policy that he could affect without meaningful interference:  he could continue his dramatic reshaping of the Judiciary.  Only 51 Senate votes (including Pence, as needed) are required to approve a judge, even a Supreme Court justice, and the House is not involved in this process.  But the House could stop the rest of Trumpworld legislation, or negotiate hard for Dem-friendly terms.

RECENT HISTORY

It is conventional wisdom – born of fact -- that first-term Presidents have a tough time in the mid-terms.  Translating campaign poetry into governing prose is difficult work, and time and again the bloom quickly comes off the rose.  George W. Bush 43 was one of only two first-term presidents in the past 100 years to see his party pick up seats in the midterms, the other being FDR in 1934.  The GOP gained seats under Bush 43 on the strength of strong positive feelings about him (and his party) in the aftermath of 9/11, which had occurred 14 months earlier. 

The chart below shows the last seven mid-term elections under first-term presidents, with a set of data that describes the political environment; the key data are the president’s approval rating and the closely linked “generic ballot.”  The generic ballot is a polling question that asks which party the respondent would support in a Congressional election – no candidate is specified, and that is what makes it “generic.”

Year
President (all first-term)
President's Party # of House Seats
President's Approval Rating (Gallup) Pre-Election
President's Party Gallup Generic Ballot Net
Actual Pres Party Seat Change
2010
Obama
257
45
-9
-63
2002
Bush 43
221
63
6
8
1994
Clinton
258
46
-7
-54
1990
Bush 41
167
58
-8
-8
1982
Reagan
192
42
-10
-26
1978
Carter
292
49
10
-15
1970
Nixon
192
58
-6
-12

The generic ballot is predictive of mid-term outcomes.  In each year, if the generic ballot was negative for the president’s party, which it was in five of the seven elections, that party lost ground.  The generic ballot favored the president’s party twice; once was the one time the president’s party actually picked up seats, in 2002 when Bush 43’s GOP gained +8 seats, and the other was in 1978, when the polls reflected post-Watergate/post-pardon blues for the GOP but Carter still lost seats at the ballot box.

But the generic ballot is not the only important variable; another one is how many seats the party in power holds.  Bill Clinton and Barack Obama had huge majorities at mid-term time (258 and 257 Democratic seats, respectively), and that, in combination with those negative generic ballots spreads, translated into huge losses, -54 seats for Clinton and -63 for Obama.  Ronald Reagan and George H. W. Bush also had large negative generic ballot gaps, but the GOP held far fewer seats, 192 under Reagan and 167 under Bush, so they lost only 26 and 8 seats, respectively.  This makes sense – the fewer seats the party has, the greater percentage of them that are “solid” seats, which provides a “natural floor” on how many seats the party might lose.

What does all this mean for Donald Trump and the GOP?  Trump’s current profile – just over five months from the mid-terms – closely resembles that of Clinton and Obama (see a subset of the chart, below).  His party is firmly in control with 235 seats, not quite as many as Clinton and Obama, but far more than Reagan or Bush 41.  He himself has a low approval rating, (using Gallup, for consistency with past presidents’ data) at 45%, exactly in line with the low levels that Obama (45%) and Clinton (46%) held at the time of their midterms. 

And Trump’s generic ballot is very negative, at -7, even with Clinton (also -7) and slightly better than Obama (-9).   This, of course, reflects the toxic political environment for the GOP, which includes Trump’s own low approval rating and the GOP’s legislative failures in certain important areas, including health care, gun control and immigration, plus the relatively low level of support for their one major accomplishment, the tax cut (four recent polls show that on average only 36% of Americans approve of the cuts, while 42% oppose them),

Pres Party House Seats
Pres Approval Gallup Pre-Election
Pres Party Gallup Generic Ballot Net
Pres Party Seat Change
Year
President
2018
Trump
235
45
-7
 ?
2010
Obama
257
45
-9
-63
1994
Clinton
258
46
-7
-54

Clearly, Trump’s situation is very similar to that of Clinton and Obama.  One might conclude, given that, he might lose as many as 50 seats.  And he might.  But our snapshot – which shows a range of a +28 to 50-seat loss – while dreadful for Trump is not quite as epic as the Clinton/Obama showing.  Why?

THE MODEL

We ran the data through our proprietary BTRTN House regression model, which was built using data from all off-year House elections since 1970 (the earliest year we had all the relevant data).  The model predicts, specifically, the number of seats the president’s party will win or lose in an off-year election.  There are five variables in the model:  the president’s term, the president’s party, the generic ballot, which party is in the majority, and the number of seats held by the president’s party.

The input on this data is as follows:  Trump is a first-term Republican, the GOP is the majority party with 235 seats, and the generic ballot right now has the Dems up by +7.  Given these inputs, the model predicts that the GOP would lose 47 seats in 2018 if today were Election Day.

This number is lower than Clinton’s and Obama’s because the GOP holds fewer seats now than the Dems held going into those elections and, in Obama’s case, the generic ballot is less negative under Trump than under Obama.

But “D+47” is not the final answer.  We have to make an adjustment based on the impact of a more recent factor:  gerrymandering.

THE ADJUSTMENT

In 2016, the GOP won the “popular vote” in the House (that is, the sum of all the individual 435 elections) by a 49%/48% margin.  But astonishingly, they came away with a decisive majority of seats, winning 55% of the seats on only 49% of the votes, thereby taking a far greater number of seats than they “deserved.”  Such is the effect of the epidemic of gerrymandering now in action.  A one-point gap should have more reasonably translated into a 220 (GOP) to 215 (Dem) House composition, or a +5 GOP seat lead instead of the +47 they actually achieved.

Since 2010, the GOP has done an astonishing job of winning state houses and controlling state Senates and Assemblies, and thereby effecting redistricting schemes –“gerrymandering” -- that favor their candidates.

You can see from the chart below that, in 2010, when a relatively unpopular Barack Obama got crushed in his first mid-terms, there were 82 “close” elections – elections decided by a margin of 10 or fewer percentage points – which was just under 20% of all the races.  But the number of close elections has steadily dwindled since then, and in the last go-round in 2016, there were only 35 such races – 8%. Clearly gerrymandering is a factor.


House Elections Decided  By 10 Points or Less
Year
#
%
2010
82
19%
2012
68
16%
2014
51
12%
2016
35
8%

The GOP lost only 17 races by 10 percentage points or less in 2016, which means that even if the Dems flipped all of them in 2018, that would not be enough to gain control (as mentioned, they need to win a net +25).  There are another 18 races that the GOP lost by 11 to 15 percentage points, and the Dems would have to take than a few of those as well.

But gerrymandering works both ways.  Courts have ruled against the status quo in Pennsylvania, resulting in a redistricting scheme that is clearly advantageous to the Dems, and should result in a pick-up of up to as many as six seats in Pennsylvania.

Our model uses data from the 2010 era forward, and so it incorporates some of the effects of recent GOP gerrymandering.  While it is safe to assume that some sort of discount is in order, on judgment we are taking a conservative posture of reducing the expected gain by -20%.

THE SNAPSHOT

Hence, while our model has predicted the Dems will pick up 47 seats, we are reducing that by 20%.  If the election were held today, the Dems would pick up +38 seats, with a likely range of +28 to +50.

Right now the various rating services show 113 races that have the slightest chance of being "in play."  The GOP currently holds 89 of them.  This is where the Dems will have to pick up these seats.

The takeaway from this exercise is three-fold:  1) if this political environment persists, the Dems have a solid shot of re-taking the House; 2)  keep a close eye on the “generic ballot” which is the strongest predictor of actual performance, adjusted for gerrymandering effects (any Dem lead at all in the generic ballot, even by a slim margin, puts the House in play); and 3) at the end of the day, elections are still won and lost based on the popularity of incumbents (and whether they run again – 37 GOP incumbents reps are not, for various reasons, versus only 18 Dems), the strength of the challengers, they money they raise and the effectiveness of their campaigns. 


Don’t take any of that for granted. If the Dems want to re-take the House, they have to take what they have done in the high-profile New Jersey, Virginia and Alabama wins in 2017, and bring that level of talent, resource, commitment and energy to a national scale in 2018.



5 comments:

  1. If the American people put the Left back in control of America, it's end will be assured. The Constitution will be ripped apart. Your taxes will go up, your military will be neutered. We will become a 2rd world power of nothing.

    ReplyDelete
    Replies
    1. We're already a third-rate power, by Trump's choice and by his actions. We are a laughingstock right now.

      Delete
    2. Get back to the fry station.

      Delete
  2. "The generic ballot is predictive of mid-term outcomes. In each year, if the generic ballot was negative for the president’s party, which it was in six of the seven elections, that party lost ground. The one time the generic ballot favored the president’s party was, in fact, the one time the president’s party actually picked up seats, in 2002 when Bush 43’s GOP gained +8 seats." Not according to your chart. I think you have a typo in the table that appears above this - there are only five negatives shown for the generic poll, not six. I suspect that the Carter 1978 year should be a -10 not a 10.

    ReplyDelete
  3. Good catch. The data is correct; I have corrected the words to reflect the proper interpretation. Thanks.

    ReplyDelete

Leave a comment