I get my analysis primarily from Martin Longman at Booman Tribune. He’s superb at explaining the nuts and bolts of how government works – and doesn’t – and currently has two pieces up looking at the hurdles and potential disasters waiting in September for Congress.
Cool, thanks.
You’re welcome. Another good blogger to follow is No More Mister Nice Blog. Steve is currently on a few days’ hiatus for personal stuff but he has guest writers who cover for him who are also excellent. Steve wades through the RW sewers and brings back reports so the rest of us don’t have to. 
Check out the last few weeks and see what you think.
Nate Silver for sure.
I think his 2016 forecast was his masterpiece, ironically so because he did end up getting it wrong in contrast to 2012 where he got almost everything right. However in 2012 the state polls were largely accurate so anyone could have got it mostly right. In 2016 the state polls were significantly off making it very difficult to make the right prediction. However Nate Silver was unique in figuring out that Hillary was vulnerable to a moderately sized polling error co-related across states and that is exactly what happened. It was especially impressive because he was probably rooting for Hillary and he got a lot of grief in the closing days before the election. It was a display of character as well as intellect.
I was also impressed by Nate Cohn, especially the quality of his analysis. He correctly identified Hillary’s weakness among WWC voters in the Midwest before the election.
As for Sam Wang, he is certainly a very smart person, but I think he was guilty of motivated reasoning and unable to identify and correct his own biases. I probably won’t be paying much attention to his election forecasts though I look forward to reading more from him on neuroscience which is after all his main career.
Scott Adams, and a youtuber called Styxhexenhammer. Both correctly called the last election. Both are interesting to listen to. However, these political gurus are usually a flash in the pan. Nine times out of ten they crash and burn. It’s best not to get too anamoured with any individual who gives political analysis or predictions. This election cycles wonderkid is often next election cycles laughing stock.
Scott Adams made every possible prediction in the last month or two of 2016. Any result would have confirmed at least one of his predictions.
This.
Based on the information and data that pollsters and election analysts had at the time, I’d say Nate Silver called it the right way. The available suggested that Hillary was a favorite to win, based on the history of prior elections and based on the accuracy of polling in the accurate. And at the risk of slip sliding into a debate I really don’t want to have again, Hillary Clinton did win in terms of popular vote. But as I recall, what Nate was trying to point was that the electoral vote itself made the final prediction about who would prevail much more complicated. This is where Sam Wang and the others were simply way off, assuming that national polling was predictive of the state-by-state results. He factored (correctly) that despite all of the data in hand, there was enough of a possibility that the final state-by-state results would lead to surprising outcome.
Why should anyone pay any attention to political prognosticators in the first place? They might as well be looking at goat entrails to read the future.
Some of them have very good track records, like Nate Silver.
In a sense, he has a point. What’s really the point of betting on the outcome? Why not just vote and see what happens?
But there probably is a real-world point in doing so somewhere, depending on what you do for a living. Besides, we’re just bored and like to bet for pride on SDMB.
But back to the topic at hand, I think that’s why some pundits started turning on Nate Silver last year. They didn’t like the fact that he was hedging. But sometimes, it’s perfectly logical to say “I don’t know - I don’t have enough good information.” If I were investing in markets, which one would you rather bet on the guy who calculates the unknown and advises you to spread your portfolio or the guy who promises to eat a bug and advises you to go all in on a particular candidate because he claims that he has infallible math?
Exactly. If the weather forecast said “30% chance rain” and it rained, I wouldn’t say they were wrong. I don’t think anyone else gave Trump this high a chance and had the analysis & data to back up that assertion.
Nate Silver correctly understood and incorporated the possibility of systematic error. The margin of error in polls are not just random noise; they include uncertainty in the assumptions and the corrections they are making. If those assumptions are off, all the results from the same pollster would be slightly off in the same direction. If multiple pollsters made the same incorrect assumption (which is actually likely, because they’re all working off of the same data set, namely the past few election results), they can all be off in the same direction. You can’t correct for systematic errors by taking the average of many polls.
I’ll pay more attention to Wang, just because I think Silver is an arrogant little prick. But I’ll also be tossing goat entrails on the ground and doing my own readings.
Wang was incredibly, hugely wrong. He has no credibility.
I never got that sense at all; I think he angered a lot of progressives during the campaign because he wouldn’t call the race for Clinton. In fact Ryan Grim the HuffPo gave Nate Silver the equivalent of a pitting for his refusal to do so.
It even to the point where some were speculating that Silver might be intentionally using his star power to attract more right wing readers who were hoping to find any poll that might be favorable. There were all sorts of things written about Silver from the Left to the point where they became Sam Wang fanboys. Silver responded by saying “Facts or shut the fuck up”. Well we now know who was really right. And again, I think this is why people, particularly on the Left, don’t like Silver. Just as people on the right didn’t care for Silver in 2008 and 2012. He goes where the set of data takes him.
And as was already pointed out, Wang was way wrong. That doesn’t mean Wang can’t improve, but I think Nate has the best track record of calling races.
Not particularly relevant imo. Adams was predicting the election as best as he could. He was allowed a wobble or two. The prediction he held for the vast majority of the campaign(including on election eve) was a Trump win. He got his main prediction correct. He was also damn close to being correct with his prediction that the election would be as one sided for Trump as any Presidential election in recent history. Very, very few were predicting that. It was a prediction that was so close in coming true.
I think this is a copout – it’s easy to get a prediction right when you predict every outcome at some point.
He always predicts that the Republican will win. He predicted a Romney win, a McCain win and a Trump win. Next election he will predict a Trump win. He’s 1 for 3 and next election will be 1 for 4. That makes him a worse predictor than a coin. He also hedged the hell out of his Trump prediction. He also didn’t predict that Trump would lose the popular vote by such an enormous margin, which was the specific way in which the Trump election was an outlier.
Given that Wang was the one who spent months singing his own certainty and his own praises, and talking about how Silver was wrong because he was an amateur who didn’t understand math, I’m not sure I grasp how SILVER is the arrogant one.
538 is the best in the business.
Long and mathful post to follow…
When two (or more) models are predicting margins of victory there are better ways to compare them than counting who picked the most winners. For example, for Maine CD-2, RCP had R+0.5 and 538 had D+0.4. The actual result was R+10.3. Both predictions were pretty horrible and counting that as a hit for RCP and a miss for 538 doesn’t really capture the truth of the matter. Nate Silver was off by a mile and RCP was off by 5,270 feet. The truth is there just wasn’t very good polling data for Maine CD-2
One way to better compare such models is calculating root mean square error (RMSE).
I calculated RMSE for RCP, 538, and HuffPost for the states that they each predicted a margin for. RCP did not have margins for states where there was insufficient polling so it mostly limited the races to compare. HuffPost did not Maine CD-2 so that was not included much to others’ benefit based on the results reported above. Also I couldn’t find PEC margins. I believe they exist and would like to include them so if anyone knows where there are let me know.
Here goes…
State Actual RCP 538 HuffPost RCP RMS 538 RMS HP RMS
Iowa 9.41 3.0 2.9 3.2 41.09 42.38 38.56
Ohio 8.13 3.5 1.9 1.1 21.44 38.81 49.42
Georgia 5.13 4.8 4.0 2.7 0.11 1.28 5.90
North Carolina 3.66 1.0 -0.7 -1.5 7.08 19.01 26.63
Arizona 3.54 4.0 2.2 2.3 0.21 1.80 1.54
Florida 1.2 0.2 -0.7 -1.6 1.00 3.61 7.84
Wisconsin 0.77 -6.5 -5.3 -6.2 52.85 36.84 48.58
Pennsylvania 0.72 -1.9 -3.7 -4.5 6.86 19.54 27.25
Michigan 0.23 -3.4 -4.2 -6.7 13.18 19.62 48.02
New Hampshire -0.37 -0.6 -3.6 -3.3 0.05 10.43 8.58
Nevada -2.42 0.8 -1.2 -2 10.37 1.49 0.18
Maine -2.96 -4.5 -7.4 -9.2 2.37 19.71 38.94
Colorado -4.91 -2.9 -4.0 -5.2 4.04 0.83 0.08
Virginia -5.32 -5.0 -5.6 -5.5 0.10 0.08 0.03
New Mexico -8.22 -5.0 -5.8 -7.6 10.37 5.86 0.38
Oregon -10.98 -8.0 -9.2 -10.6 8.88 3.17 0.14
Total 11.25 14.03 18.88
RCP RMSE 538 RMSE HP RMSE
3.35 3.75 4.35
The RCP average still does the best here (lower is better), but I think this analysis shows that there’s a little more subtlety in the big picture than just counting wins and losses would indicate.
I don’t really follow any of them but when I do it is usually 538. Silver feels to me like an honest analyst.