Friday, November 8, 2013

Assessing the 2013 Actual vs. Expected Results and Exploring Why

Governor's races are so tricky.  Odd year races are even trickier.  Figuring out turnout in elections where there is no national decisions on the ballot is a bit of a guessing game.  Polls from reliable firms are harder to come by.  Hence, historically my margins of error have been greater in these races.  This year is no different.

Here is the rundown of actual versus projected results:
New Jersey Governor
Predicted Outcome: Christie +24.1%
Actual Result (preliminary): Christie +22.3%
Error in Projection: 1.8% bias for Christie

Virginia Governor
Predicted Outcome: McAullife +6.4%
Actual Result (preliminary): McAullife +2.4%
Error in Projection: 4.0% bias for McAullife

New York Mayor
Predicted Outcome: DeBlasio +41.2%
Actual Result (preliminary): DeBlasio +48.8%
Error in Projection: 7.6% bias for Lhota

First, the good news in the projection accuracy:
* All 3 races were called correctly
* The New Jersey race was actually very well called, considering the margin
* I was correct about Libertarian Robert Sarvis - his support did fade substantially versus the late polling, although he clocked in at 6.5%, above my projected 5-6% (but well below 9% or so that he was polling)

Now, the bad news:
* Like most, I dramatically overcalled the margin of Terry McAullife's victory in Virginia.  He simply underperformed virtually all of the polling.  A special call out to the Emerson College poll, which nailed the margin of the race, although it had Sarvis at a whacky 13%.  All of the other polls had McAullife ahead by far more.

So why did McAullife underperform?

Was it the scourge of Obamacare?  Republicans certainly think so and have been amping up the closing of the race as a precursor to a scorched earth campaign on Obamacare in 2014.  And they have some reason to believe.  Cuccinelli campaigned hard late on Obamacare and the news of the early failures of the website and of cancelled policies were happening right during the time when Cuccinelli was closing the gap.

A counterpoint to this would be depressed turnout.  Turnout for the 2013 Virginia Governor's race is estimated at 37%, down from 42% in the 2009 race, which may have formed the basis for a lot of poling models, and as in most elections, lower turnout tends to favor the Republican.  Still, 2009 was a historically high turn out race and 37% is not out of line for a more typical race.

My conclusion?  Voter turnout model error may be responsible for part of the result, but the Obamacare issue does loom large at the moment.  Whether it will work as an effective 2014 strategy for the GOP will depend largely on what happens in the next 12 months.  The story probably won't be about a broken website by then.  But whether it is about people paying more for policies that they don't want or having greater access, choice and value will decide whether Obamacare is an albatross to the Democrats or a boon.  We'll have to wait and see.

If you like this site, tell your friends.

No comments: