Reliance On Broken Polling Data Will Leave “Analysts” Stumped When Republicans Hold The House

Similar to the 2016 election, political pundits have been stating for months that the Democratic party is all but a sure thing to win enough seats to control the House of Representatives.  The Drudge Report recently featured the below headline highlighting FiveThirtyEight’s prediction:

We predicted that the Republican party would lose some seats but hold the House, even prior to the contentious Kavanaugh confirmation hearings.  Since then, the Republican party’s position has only strengthened, buoyed by non-stop campaigning by President Trump amid record-breaking crowds.  Still, this has not changed the tune of the political pundits (or the gamblers) who continue to predict a Democratic House and return of Speaker Pelosi.  

No one is more vocal about this then the aforementioned Nate Silver, who heads up FiveThirtyEight and presumably stands behind his calculations.  We will give him the benefit of the doubt and examine his “deluxe” forecast, instead of the “classic” cited above:

Well if he thinks Democrat House odds are 83%, he must love them at 67%, where he can currently bet them on PredictIt:

The major flaw in FiveThirtyEight’s forecasts is their near-total reliance on polling data, with little analysis given to the districts themselves, as stated in their methodology:

We’re adjusting poll results in three ways: Polls of registered voters or all adults are adjusted to a likely-voter basis; older polls are adjusted based on shifts in the generic congressional ballot since the poll was conducted; and polls are adjusted for house effects (the tendency for a firm’s polls to lean toward Democrats or Republicans). Polls with larger sample sizes and those conducted by higher-quality polling agencies are given more weight, as are more recent polls.

To counter this, FMShooter will examine three “close” districts in Florida, a swing state that commonly rates as “purple”, and where voting results frequently mimic the national mood in any given election year.  The most obvious example FiveThirtyEight’s data failure is in FL-27, the district containing Miami Beach and south Miami, which they are stating is all but certain to flip:

The district has been represented by Ileana Ros-Lehtinen since 2013, in spite of its heavy blue tilt.  Ros-Lehtinen had won fifteen elections, but is retiring this year.  Competing for the seat are Donna Shalala, Bill Clinton’s former HHS secretary, and Maria Salazar, a Telemundo news anchor.  The district is over 71% Hispanic, and this should be an easy flip for Democrats, except…

Shalala is a member of the heavily-disliked Clinton cabal, and does not speak Spanish in a heavily Hispanic district.  Salazar, meanwhile, has been on TV for years as an anchor, and is attempting to fill a longtime Republican seat.  While we do not expect the Republicans to win this seat, we certainly think the odds are much closer to 50/50 than the 80+% Silver’s analysts have assigned to the race.

FiveThirtyEight has another nearby district, FL-26, as a toss-up:

Two-term Republican incumbent Carlos Curbelo won both of his elections fairly handily, by three and twelve points respectively.  A district local, he has a relatively low Trump score according to FiveThirtyEight themselves, opposing Trump when it represented his district.

His opponent Debbie Mucarsel-Powell is a political unknown, whose father was killed in Ecuador, a country with far stricter gun laws than the US.  That of course hasn’t stopped her from making tighter gun control a linchpin of her platform, a foolish strategy in the “gunshine” state.

We’re not sure what has changed in the district that will lead Mucarsel-Powell to make up even three points, especially with President Trump’s work to carry the party on the national level.  With the Kavanaugh hearings being used by Curbelo against Mucarsel-Powell, we estimate the odds of Curbelo winning are far above the 54.6% (or 50-50 “classic” forecast) that FiveThirtyEight is projecting.  

But perhaps the most blatant area of FiveThirtyEight’s data failure is in FL-15:

Containing Lakeland and other Tampa Bay exurbs, the district has been a Republican stronghold for decades.  Even in 2008, when the national party was carried by former President Obama, it was lost by 15 points (when it was the 12th district).  Dennis Ross, the district’s current Representative, is retiring, and Florida State Representative Ross Spano is the GOP nominee to fill the seat.

The district is nearly 60% white, a demographic which heavily favors President Trump.  Ross’s opponent, Kristen Carlson, is somewhat of an unknown to us, running on what appears to be the party platform.  While most will attribute the national mood and a “blue wave” to the generic ballot swinging, the reliance on polls instead of solid analysis will more than likely leave Democratic voters disappointed with the results in FL-15.

Nate Silver would be wise to turn back the clock to the GA-06 special election in 2017, where tens of millions of dollars were spent by the DNC attempting to flip a solidly Republican seat.  Silver followed the polls, delivering this cringeworthy prediction:

If the above races follow our non polling-driven analysis, as opposed to a fancy website which relies on the generic ballot to forecast, it might be time for Silver to admit that his polling data is broken in the Trump political era. 

Note: The oddsmakers seem to also be following polls and data like Silver’s, delivering what we believe to be mispriced markets on betting websites.  See parts one, two, and three of our Midterm Election guide for Senate races that we like.  We plan on providing a final segment to the guide just before the election, updating one or two Senate races with new predictions.