Election Predictions: How Wrong (or Right) Were We?
We rated races for president, governor, state legislature and attorney general.
Politicians face voters on Election Day. In the succeeding days, political handicappers are the ones called to account.
With the ballot-counting (mostly) complete, how did we do?
First, some background. Governing offered ratings periodically during the 2016 campaign cycle for four types of contests: states in the Electoral College, the gubernatorial races, the state attorney general races and control of the state legislatures.
Ratings were based on interviews with political observers and a review of polling data. They were also based on experience: I’ve been rating the legislatures since the 2002 cycle, gubernatorial races since the 2006 cycle, and the Electoral College and the attorney general races since the 2008 cycle.
Each race and chamber was rated as either safe Republican, likely Republican, lean Republican, tossup, lean Democratic, likely Democratic or safe Democratic.
For the Electoral College and the gubernatorial and AG races, we added an additional step: Within each rating category, the states were rank-ordered so they could be viewed as a continuum between the seats most likely to go Republican (at the top) and the states most likely to go Democratic (at the bottom). The idea is to draw a line and find all the states above that line won by a Republican and all the seats below that line won by a Democrat.
What follows is a rundown of the accuracy of our final round of handicapping, published the day before the election.
There has been much anger at how some political number crunchers wrongly considered Hillary Clinton to be a virtual lock to win the presidency.
We’ll admit to being shocked at Donald Trump’s victory, like so many others. But this column is fundamentally about state politics -- not national politics -- so we have long used a system that focuses on the state-by-state results, not on predicting the ultimate presidential winner. Our system accepts that one party will ultimately be stronger than the other in winning competitive states, but our aim is to be accurate, to the extent we can, about the comparative likelihood of any given state going Republican or Democratic.
In 2008, our placement of states on the continuum was correct except for Missouri and one electoral vote for a congressional district in Nebraska. In 2012, we placed the line exactly right, with the one state in the tossup category that was ranked closest to lean Republican -- North Carolina -- going to Mitt Romney, and all other tossup states going to Barack Obama.
This year, we failed to correctly place three states on the continuum: Wisconsin, Michigan and Pennsylvania should have swapped places with New Hampshire, Nevada and Colorado. For the three cycles we’ve been handicapping the Electoral College, this is our biggest error yet. Like so many others, we underappreciated Trump’s ability to draw Rust Belt voters to the polls.
Here’s the complete list, with “misplaced” states in bold.
Safe Republican (86 electoral votes)
Alabama (9), Arkansas (6), Idaho (4), Kansas (6), Kentucky (8), Louisiana (8), Mississippi (6), Montana (3), Nebraska (4 of 5 electoral votes), North Dakota (3), Oklahoma (7), South Dakota (3), Tennessee (11), West Virginia (5) and Wyoming (3)
Likely Republican (71 electoral votes)
Texas (38), Alaska (3), South Carolina (9), Missouri (10), Indiana (11)
Lean Republican (29 electoral votes)
Utah (6), Nebraska (1 of 5 electoral votes), Georgia (16), Iowa (6)
Tossup (78 electoral votes)
Ohio (Trump, 18), Arizona (Trump, 11), Maine (Trump, 1 of 4 electoral votes), Florida (Trump, 29), North Carolina (Trump, 15), New Hampshire (Clinton, 4)
Lean Democratic (91 electoral votes)
Nevada (Clinton, 6), Colorado (Clinton, 9), Wisconsin (Trump, 10), Michigan (Trump, 16), Pennsylvania (Trump, 20), Maine (Clinton, 2 of 4 electoral votes), Virginia ( Clinton, 13), Minnesota (Clinton, 10), New Mexico (Clinton, 5)
Safe Democratic (183 electoral votes)
California (55), Connecticut (7), Delaware (3), District of Columbia (3), Hawaii (4), Illinois (20), Maine (1 of 4 electoral votes), Maryland (10), Massachusetts (11), New Jersey (14), New York (29), Oregon (7), Rhode Island (4), Vermont (3) and Washington state (12)
Our continuum for gubernatorial races was slightly more accurate. Ultimately, two seats were on the wrong side of the dividing line. (This assumes that the current leader in the North Carolina gubernatorial race, Democrat Roy Cooper, prevails in a recount.)
In our rankings, we placed two gubernatorial races that were won by Republicans -- Indiana and New Hampshire -- too far in the Democrats’ direction.
This is an improvement from our 2014 rankings, when we placed four races on the wrong side of the dividing line. But it’s not quite as good as 2012, when our continuum was perfect, and 2010, when we were off by just one gubernatorial contest.
Both 2014 and 2016 were notable for the large number of tossup races on the eve of the election: 12 in 2014 and seven in 2016. Having so many contests as tossup increases the likelihood of jumbled results when the election returns come in.
Here’s the complete list, with “misplaced” races in bold.
Utah, North Dakota
Missouri (Republican), Vermont (Republican), West Virginia (Democratic), Montana (Democratic), New Hampshire (Republican), North Carolina (Democratic), Indiana (Republican)
Finally, a category we nailed: We correctly constructed our continuum of races for attorney general seats. It was the third straight cycle we have correctly ordered the AG contests.
West Virginia (Republican), Missouri (Republican), North Carolina (Democratic)
Vermont, Oregon, Washington state
Because of the complexity of rating 98 chambers, we don’t rank-order them within each rating category. Instead, we judge our success based on whether we failed to peg any chambers that ultimately flipped party control as “competitive.” (“Competitive” races are either lean Republican, lean Democratic or tossup.)
We were quite accurate in this regard in 2016. The only chambers that changed partisan control that were not considered “competitive” had unusual quirks. In the formerly Republican-controlled Alaska House, a coalition of Democrats and moderate Republicans took control after the election. We didn’t predict that the GOP would lose control in that chamber -- we rated it likely Republican -- although we did at least cite in our final analysis the pivotal role of the group of rural Democrats known as the "bush caucus."
Similarly, our 2014 ratings were assailed by an unusual quirk: The West Virginia Senate flipped only due to post-election party switches.
Meanwhile, in the Delaware Senate -- which we had rated safe Democratic -- there’s a tie that should be short-lived. In the narrowly divided chamber, one Democrat lost a seat on Election Day, and another Democrat won the lieutenant governorship, creating a vacancy that left the chamber tied. But that deadlock is expected to disappear once a special election is held to fill the vacant seat -- and in any case, until then, the new Democratic lieutenant governor will be able to break ties in the chamber.
Beyond this pair of quirky results, our handicapping was on target for the chambers that flipped. Each was considered competitive in our final rankings.
Here’s the rundown of flipped chambers, along with their pre-election rating:
Kentucky House, Lean R; Iowa Senate, Tossup; Minnesota Senate, Lean D
Nevada Assembly, Lean D; Nevada Senate, Tossup; New Mexico House, Tossup
Connecticut Senate, Tossup