How did the 2016 predictions compare?

[Princess Nyan Cat]

Hello, dear Atlasians.  Now that the Louisiana elections are finally complete, I make a rare visit once again to compare the accuracy of the multitude of election predictions.  For past election cycles, see...

2014:  https://uselectionatlas.org/FORUM/index.php?topic=208958.0
2012:  https://uselectionatlas.org/FORUM/index.php?topic=187008.0

Lots of new predictions were added this year, and as you'd expect from the shocking result, none did that great a job at the Presidential level.  Accuracy results for Senate, House and Governor races were more typical, though perhaps the victor in House elections may come as a surprise!

The "score" is a simple one based on whether the prediction picked the right party to win and does not attempt to assess strength of victory.  It is often difficult to quantify terms like Tilt, Lean, Likely, Moderate, Strong, Solid, and so on when many of the sites do not clearly define what those qualifications mean.  A prediction of "tossup" is automatically classified as wrong since the prognosticator was unwilling to make a stand one way or the other.

Enjoy!

[Princess Nyan Cat]

Presidential election - grades on at least 51 races (50 states + DC) and up to 56 (50 states + DC + congressional districts in Maine and Nebraska)

Many new prediction sources were added this year for a grand total of 32.  My past superstars Larry Sabato, Scott Elliott, and PredictWise were all at or near the top but Real Clear Politics and Electoral Map were also impressive.

92.9% - Election Projection (Scott Elliott) - 52/56 correct (2012: 50/51)
92.9% - Real Clear Politics - 52/56 correct (2012: 42/53)

92.2% - Electoral Map (Pivit) - 47/51 correct (2012: 50/51 using Intrade)

91.1% - Larry Sabato's Crystal Ball - 51/56 correct (2012: 52/54)
91.1% - New York Times - 51/56 correct (2012: 46/53)
91.1% - PredictWise - 51/56 correct
91.1% - Princess Nyan Cat =^..^= - 51/56 correct (2012: 55/56)

90.2% - CrossTab (G. Elliott Morris) - 46/51 correct
90.2% - Daily Kos - 46/51 correct
90.2% - DeSart & Holbrook Forecasting - 46/51 correct
90.2% - PolyVote (Andreas Graefe) - 46/51 correct

89.3% - 538 (Nate Silver) - 50/56 correct (2012: 55/56)
89.3% - Dave Leip's Atlas aggregate - 50/56 correct (2012: 55/56)
89.3% - PredictIt prediction market - 50/56 correct
89.3% - Princeton Election Consortium (Sam Wang) - 50/56 correct

88.2% - Electoral Vote (Andrew Tanenbaum) - 45/51 correct (2012: 49/51)
88.2% - People's Pundit Daily - 45/51 correct
88.2% - Rothenberg & Gonzalez Political Report - 45/51 correct (2012: 43/51)
88.2% - Washington Post - 45/51 correct (2012: 44/51)

87.5% - Cook Political Report - 49/56 correct (2012: 49/56)
87.5% - Huffington Post - 49/56 correct (2012: 47/51)
87.5% - National Public Radio - 49/56 correct

86.3% - Los Angeles Times - 44/51 correct
86.3% - Talking Points Memo - 44/51 correct (2012: 48/51)

83.9% - ABC News - 47/56 correct (2012: 43/51)
83.9% - CNN - 47/56 correct (2012: 43/51)
83.9% - Governing - 47/56 correct
83.9% - Race-4 - 47/56 correct (2012: 51/53)

82.4% - YouGov - 42/51 correct

82.1% - Associated Press - 46/56 correct

80.4% - NBC News - 45/56 correct

74.5% - Fox News - 38/51 correct (2012: 45/51)

70.6% - USA Today - 36/51 correct

Showing how poorly the experts did overall, every prediction was wrong in 3 states -- Michigan, Pennsylvania, and Wisconsin.

Only 4 were right in Florida -- Real Clear Politics, EVP-Scott Elliott, People's Pundit Daily, and Electoral Map.

Only 5 were right in North Carolina -- Real Clear Politics, EVP-Scott Elliott, DeSart & Holbrook, People's Pundit Daily, and Electoral Map.

[Princess Nyan Cat]

Senate elections - graded on 34 races

Includes a total of 16 prediction sources.

94.4% - 538 (Nate Silver) - 32 correct (2014: 34/36, 2012: 31/33)
94.1% - Dave Leip's Atlas aggregate - 32 correct (2014: 35/36, 2012: 31/33)
94.1% - Larry Sabato's Crystal Ball - 32 correct (2014: 35/36, 2012: 31/33)
94.1% - New York Times - 32 correct (2014: 35/36, 2012: 24/33)
94.1% - PredictWise - 32 correct (2014: 35/36)
94.1% - Princess Nyan Cat =^..^= - 32 correct (2014: 35/36, 2012: 31/33)

91.2% - Daily Kos - 31 correct (2014: 35/36)
91.2% - Election Projection (Scott Elliott) - 31 correct (2014: 34/36, 2012: 31/33)
91.2% - Electoral Vote (Andrew Tanenbaum) - 31 correct (2014: 34/36, 2012: 29/33)
91.2% - Real Clear Politics - 31 correct (2014: 34/36, 2012: 26/33)

88.2% - Huffington Post - 30 correct (2014: 35/36, 2012: 28/33)
88.2% - Princeton Election Consortium (Sam Wang) - 30 correct (2014: 34/36)

85.3% - Talking Points Memo - 29 correct (2012: 28/33)

79.4% - Cook Political Report - 27 correct (2014: 27/36, 2012: 23/33)
79.4% - Rothenberg & Gonzalez Political Report - 27 correct (2014: 32/36, 2012: 29/33)
79.4% - Washington Post - 27 correct (2014: 35/36, 2012: 28/33)

73.5% - Ballotpedia - 25 correct

Every prediction was wrong for Pennsylvania and Wisconsin.

Dishonorable mention to Sam Wang, who I was tempted to grade 0% after we had to get his predictions from the New York Times because they were impossible to find on his own site.  When my husband asked about this, his post was deleted.  That's the second election something like this has happened.  From reviewing his blog, I get the impression he is especially arrogant and can't tolerate anything but fawning praise from his readers.

[Princess Nyan Cat]

House elections - graded on 435 races (number wrong listed instead of correct)

Includes a total of 9 prediction sources.  While the usual suspects did well, look at who is the Queen of the Heap!  Yay!

99.1% - Princess Nyan Cat =^..^= - 4 wrong (2014: 13, 2012: 11)

98.6% - PredictWise - 6 wrong (2014: 13)

98.4% - Larry Sabato's Crystal Ball - 7 wrong (2014: 10, 2012: 13)

98.2% - Election Projection (Scott Elliott) - 8 wrong (2014: 10, 2012: 12)

97.0% - Rothenberg & Gonzalez Political Report - 13 wrong (2014: 17, 2012: 24)

95.9% - Cook Political Report - 18 wrong (2014: 24, 2012: 31)

95.6% - Washington Post - 19 wrong (2014: 15, 2012: 26)

95.4% - Daily Kos - 20 wrong (2014: 21)

95.2% - Real Clear Politics - 21 wrong (2014: 31, 2012: 35)

94.7% - Ballotpedia - 23 wrong

Every prediction was wrong for these 2 districts: Minnesota-2, Texas-23

Only one prediction was right for these 2 districts: Maine-2 (PredictWise), Nebraska-2 (PredictWise)

[Princess Nyan Cat]

Governor elections - graded on 12 races

Includes a total of 10 prediction sources.  There's a lot of variety ranging from very, very good to very, very poor.  For the first time since I've been doing these comparisons, one source predicted every race correctly.  Congrats, PredictWise!  And maybe you should find a new line of work, Governing!

100.% - PredictWise - 12 correct (2014: 32/36)

91.7% - Election Projection (Scott Elliott) - 11 correct (2014: 32/36, 2012: 10/11)
91.7% - Princess Nyan Cat =^..^= - 11 correct (2014: 31/36, 2012: 10/11)

83.3% - Larry Sabato's Crystal Ball - 10 correct (2014: 31/36, 2012: 10/11)
83.3% - Rothenberg & Gonzalez Political Report - 10 correct (2014: 29/36, 2012: 9/11)

75.0% - Dave Leip's Atlas aggregate - 9 correct (2014: 31/36, 2012: 11/11)
75.0% - Washington Post - 9 correct (2012: 8/11)

50.0% - Cook Political Report - 6 correct (2014: 23/36, 2012: 8/11)
50.0% - Daily Kos - 6 correct (2014: 31/36)
50.0% - Real Clear Politics - 6 correct (2014: 32/36, 2012: 9/11)

41.7% - Governing - 5 correct (2014: 23/36)

Only two predictions were right for Indiana (PredictWise, Rothenberg)

[Princess Nyan Cat]

Overall Grade

Averaging the scores, here is how they all compare.  It's a little bit apples and oranges since some predict more races than others.  Nevertheless, PredictWise seems to emerge as the superstar prognosticator for the 2016 cycle!  And somehow cuddly little Nyan Cat outdid all the rest!

96.0% - PredictWise (2014: 94.4%, 2012: -)
94.0% - Princess Nyan Cat =^..^= (2014: 93.4%, 2012: 95.1%)
93.5% - Election Projection (Scott Elliott) (2014: 93.7%, 2012: 95.0%)
92.6% - New York Times (2014: 97.2%, 2012: 84.3%)
92.2% - Electoral Map (2014: 91.2%, 2012: 98.0%)
91.7% - 538 (Nate Silver) (2014: 90.3%, 2012: 96.1%)
91.7% - Larry Sabato's Crystal Ball (2014: 93.7%, 2012: 94.5%)
90.2% - CrossTab (G. Elliott Morris) (2014: -, 2012: -)
90.2% - DeSart & Holbrook Forecasting (2014: -, 2012: -)
90.2% - PolyVote (Andreas Graefe) (2014: -, 2012: -)

89.7% - Electoral Vote (Andrew Tanenbaum) (2014: 94.4%, 2012: 92.0%)
89.3% - PredictIt predicion market (2014: -, 2012: -)
88.8% - Princeton Election Consortium (Sam Wang) (2014: 94.4%, 2012: -)
88.2% - People's Pundit Daily (2014: 90.3%, 2012: -)
87.9% - Huffington Post (2014: 91.7%, 2012: 90.5%)
87.5% - National Public Radio (2014: -, 2012: -)
87.0% - Rothenberg & Gonzalez Political Report (2014: 88.5%, 2012: 87.1%)
86.3% - Los Angeles Times (2014: -, 2012: -)
86.1% - Dave Leip's Atlas aggregate (2014: 91.7%, 2012: 97.4%)
85.8% - Talking Points Memo (2014: -, 2012: 89.5%)
84.6% - Washington Post (2014: 96.9%, 2012: 84.5%)
84.1% - Ballotpedia (2014: -, 2012: -)
83.9% - ABC News (2014: -, 2012: 84.3%)
83.9% - Race4 (2014: -, 2012: 96.2%)
83.9% - CNN (2014: -, 2012: 84.3%)
82.4% - YouGov (2014: -, 2012: -)
82.3% - Real Clear Politics (2014: 92.1%, 2012: 91.5%)
82.1% - Associated Press (2014: -, 2012: -)
81.7% - Daily Kos (2014: 92.8%, 2012: -)
80.4% - NBC News (2014: -, 2012: -)

78.2% - Cook Political Report (2014: 77.8%, 2012: 80.7%)
74.5% - Fox News (2014: -, 2012: 88.2%)
70.6% - USA Today (2014: 79.1%, 2012: -)
62.8% - Governing (2014: -%, 2012: -)

That's it for now.  Maybe I'll see you again in another 2 years!

[Vosem]

Very interesting stuff, nyancat -- thank you for sharing! And I hadn't realized you were still an active poster.

[Princess Nyan Cat]

Thanks for the compliment, Vosem.  I don't post much being so busy with a career.  But I get the urge to toil over my my massive election prediction spreadsheet every couple years.

[Nym90]

It would be interesting to be able to take a deeper dive and calculate how accurate the predictions were not only in terms of predicting the winner but also in terms of predicting the vote share for each candidate/vote margin.

Fivethirtyeight made such numeric predictions, but unfortunately most predictors aren't bold enough to offer up numbers, only picks of winners and losers.

[DPKdebator]

I had yet to be a member of this site at the time of predictions, but my map was only two states off (NV and NH were wrong).

[DPKdebator]

I had yet to be a member of this site at the time of predictions, but my map was only two states off (NV and NH were wrong).

That's a really awesome prediction. Good on you!

Ironically, I was told by everyone (including teachers) it was too red.

[Thomas Jackson]

Presidential election - grades on at least 51 races (50 states + DC) and up to 56 (50 states + DC + congressional districts in Maine and Nebraska)

Many new prediction sources were added this year for a grand total of 32.  My past superstars Larry Sabato, Scott Elliott, and PredictWise were all at or near the top but Real Clear Politics and Electoral Map were also impressive.

92.9% - Election Projection (Scott Elliott) - 52/56 correct (2012: 50/51)
92.9% - Real Clear Politics - 52/56 correct (2012: 42/53)

92.2% - Electoral Map (Pivit) - 47/51 correct (2012: 50/51 using Intrade)

91.1% - Larry Sabato's Crystal Ball - 51/56 correct (2012: 52/54)
91.1% - New York Times - 51/56 correct (2012: 46/53)
91.1% - PredictWise - 51/56 correct
91.1% - Princess Nyan Cat =^..^= - 51/56 correct (2012: 55/56)

90.2% - CrossTab (G. Elliott Morris) - 46/51 correct
90.2% - Daily Kos - 46/51 correct
90.2% - DeSart & Holbrook Forecasting - 46/51 correct
90.2% - PolyVote (Andreas Graefe) - 46/51 correct

89.3% - 538 (Nate Silver) - 50/56 correct (2012: 55/56)
89.3% - Dave Leip's Atlas aggregate - 50/56 correct (2012: 55/56)
89.3% - PredictIt prediction market - 50/56 correct
89.3% - Princeton Election Consortium (Sam Wang) - 50/56 correct

88.2% - Electoral Vote (Andrew Tanenbaum) - 45/51 correct (2012: 49/51)
88.2% - People's Pundit Daily - 45/51 correct
88.2% - Rothenberg & Gonzalez Political Report - 45/51 correct (2012: 43/51)
88.2% - Washington Post - 45/51 correct (2012: 44/51)

87.5% - Cook Political Report - 49/56 correct (2012: 49/56)
87.5% - Huffington Post - 49/56 correct (2012: 47/51)
87.5% - National Public Radio - 49/56 correct

86.3% - Los Angeles Times - 44/51 correct
86.3% - Talking Points Memo - 44/51 correct (2012: 48/51)

83.9% - ABC News - 47/56 correct (2012: 43/51)
83.9% - CNN - 47/56 correct (2012: 43/51)
83.9% - Governing - 47/56 correct
83.9% - Race-4 - 47/56 correct (2012: 51/53)

82.4% - YouGov - 42/51 correct

82.1% - Associated Press - 46/56 correct

80.4% - NBC News - 45/56 correct

74.5% - Fox News - 38/51 correct (2012: 45/51)

70.6% - USA Today - 36/51 correct

Showing how poorly the experts did overall, every prediction was wrong in 3 states -- Michigan, Pennsylvania, and Wisconsin.

Only 4 were right in Florida -- Real Clear Politics, EVP-Scott Elliott, People's Pundit Daily, and Electoral Map.

Only 5 were right in North Carolina -- Real Clear Politics, EVP-Scott Elliott, DeSart & Holbrook, People's Pundit Daily, and Electoral Map.

http://politicalref.com/2016_politicalref_performance_analysis_electoral.php

[Lachi]

Presidential election - grades on at least 51 races (50 states + DC) and up to 56 (50 states + DC + congressional districts in Maine and Nebraska)

Many new prediction sources were added this year for a grand total of 32.  My past superstars Larry Sabato, Scott Elliott, and PredictWise were all at or near the top but Real Clear Politics and Electoral Map were also impressive.

92.9% - Election Projection (Scott Elliott) - 52/56 correct (2012: 50/51)
92.9% - Real Clear Politics - 52/56 correct (2012: 42/53)

92.2% - Electoral Map (Pivit) - 47/51 correct (2012: 50/51 using Intrade)

91.1% - Larry Sabato's Crystal Ball - 51/56 correct (2012: 52/54)
91.1% - New York Times - 51/56 correct (2012: 46/53)
91.1% - PredictWise - 51/56 correct
91.1% - Princess Nyan Cat =^..^= - 51/56 correct (2012: 55/56)

90.2% - CrossTab (G. Elliott Morris) - 46/51 correct
90.2% - Daily Kos - 46/51 correct
90.2% - DeSart & Holbrook Forecasting - 46/51 correct
90.2% - PolyVote (Andreas Graefe) - 46/51 correct

89.3% - 538 (Nate Silver) - 50/56 correct (2012: 55/56)
89.3% - Dave Leip's Atlas aggregate - 50/56 correct (2012: 55/56)
89.3% - PredictIt prediction market - 50/56 correct
89.3% - Princeton Election Consortium (Sam Wang) - 50/56 correct

88.2% - Electoral Vote (Andrew Tanenbaum) - 45/51 correct (2012: 49/51)
88.2% - People's Pundit Daily - 45/51 correct
88.2% - Rothenberg & Gonzalez Political Report - 45/51 correct (2012: 43/51)
88.2% - Washington Post - 45/51 correct (2012: 44/51)

87.5% - Cook Political Report - 49/56 correct (2012: 49/56)
87.5% - Huffington Post - 49/56 correct (2012: 47/51)
87.5% - National Public Radio - 49/56 correct

86.3% - Los Angeles Times - 44/51 correct
86.3% - Talking Points Memo - 44/51 correct (2012: 48/51)

83.9% - ABC News - 47/56 correct (2012: 43/51)
83.9% - CNN - 47/56 correct (2012: 43/51)
83.9% - Governing - 47/56 correct
83.9% - Race-4 - 47/56 correct (2012: 51/53)

82.4% - YouGov - 42/51 correct

82.1% - Associated Press - 46/56 correct

80.4% - NBC News - 45/56 correct

74.5% - Fox News - 38/51 correct (2012: 45/51)

70.6% - USA Today - 36/51 correct

Showing how poorly the experts did overall, every prediction was wrong in 3 states -- Michigan, Pennsylvania, and Wisconsin.

Only 4 were right in Florida -- Real Clear Politics, EVP-Scott Elliott, People's Pundit Daily, and Electoral Map.

Only 5 were right in North Carolina -- Real Clear Politics, EVP-Scott Elliott, DeSart & Holbrook, People's Pundit Daily, and Electoral Map.

http://politicalref.com/2016_politicalref_performance_analysis_electoral.php

His vote totals are well off for a lot of states however.

[Plankton5165]

I got 40 out of 51. (78.4%)

However, I actually predicted a Trump landslide.

The states I predicted that Trump would win that Clinton won were: Oregon, Nevada, New Mexico, Colorado, Minnesota, Virginia, Delaware, New Jersey, Vermont, and New Hampshire. I also predicted he would win the statewide EVs in Maine.

I especially predicted Nevada for Trump. I was projecting the states myself, and at 10 PM, projected Nevada would go to Trump. I was highly shocked when Nevada went to Hillary Clinton.

Trump became the first Republican candidate to win the White House without carrying Colorado since 1908. Nevada since 1908, and Virginia since 1924.

Every single EV which was won by Trump was considered Safe R in my prediction. (except for Nebraska's 2nd congressional district)

[Senator-elect Spark]

I got 45 out of 50, (90%).

MI, WI, PA, CO, & NH were wrong.

I had a feeling Trump was going to win MI or PA but wasn't going to be bold. NH was a wild-card pick and I thought CO disliked Hillary.

[peterthlee]

I got way wrong - had 9 states and 1 CD (ME-02) wrong. Moreover, percentage predictions are a mess.

[Pessimistic Antineutrino]

I got 46 out of 51, or 90% roughly.

MI, WI, PA, FL and NC were wrong, as was ME-02 (this is the one I'm kicking myself over the most, actually).

Senate I got 32 out of 24 (94%)

I got PA and WI wrong only. Sticking to conventional wisdom wasn't terrible here.

House I got 10 seats wrong for a total of 425 out of 435 (96%). Granted about 90% of those were no brainers.

I missed: CO-06, FL-07, FL-18, FL-26, MI-01, MN-02, NE-02, NJ-05, NY-19, NY-22, and TX-23.

[Keep cool-idge]

While I was not member I had to maps a toss up map and a actual map

For the toss up map I had Nevada Michigan North Carolina Ohio New Hampshire and Florida as toss ups
For my actual map I had trump winning out of those Ohio North Carolina Florida I projected 278-260 Hillary

The only states wrong were Pennsylvania Michigan and Wisconsin

[DPKdebator]

I predicted the correct result in 49 out of 51 states (96%), though I didn't have an Atlas account at the time of the election so it's not linked to my account.

The two states I got wrong were Nevada and New Hampshire, which I said would go for Trump.

[America Needs a 13-6 Progressive SCOTUS]

I got 7 states wrong:

Pennsylvania
Michigan
Wisconsin
Iowa
North Carolina
Florida
Arizona