Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
Nate Silver, the statistician behind the site fivethirtyeight, became famous for his excellent prediction of the 2012 Presidential election. His prediction model -- which currently has Hillary with a roughly 81% chance of victory -- is widely cited.
However, a lot of people -- typically those who don't like his predictions -- point out that he was wildly wrong about Trump's chances in the primary. This point is correct, and in fact Silver maintained until relatively late in the game that Trump had as little as a 5% chance of winning the nomination. I think this critique is a bit overblown, and there are a few specific reasons why predicting primary results is fundamentally different than predicting general election results.
First, primaries are inherently less predictable than general elections because a major factor is when various candidates drop out of the race. The 2016 Republican primary started with sixteen candidates, and there's a good chance that Trump would not have won had everyone but Jeb or Rubio dropped out close to the beginning. However, as we all know, that isn't what happened. Many candidates lingered, and it was possible for Trump to win many states despite only having roughly 25% support. Part of predicting the winner of the Republican nomination was predicting the future behavior of the 6-8 politicians still in the race. That is an inherently difficult prediction.
Second, polling for the general election is far more thorough and sophisticated than it was for the primary. More polling dollars are being devoted to fewer questions than they were in the primary, which increases the likelihood that the polls are accurate. Silver's model is partially based on polling results, so this improvement in accuracy for the polls improves the accuracy of his model.
There are probably additional reasons, but I think these two reasons in particular give us good reason to look past Silver's inaccurate prediction about Trump in the primary.
Nate Silver, the statistician behind the site fivethirtyeight, became famous for his excellent prediction of the 2012 Presidential election. His prediction model -- which currently has Hillary with a roughly 81% chance of victory -- is widely cited.
However, a lot of people -- typically those who don't like his predictions -- point out that he was wildly wrong about Trump's chances in the primary. This point is correct, and in fact Silver maintained until relatively late in the game that Trump had as little as a 5% chance of winning the nomination. I think this critique is a bit overblown, and there are a few specific reasons why predicting primary results is fundamentally different than predicting general election results.
First, primaries are inherently less predictable than general elections because a major factor is when various candidates drop out of the race. The 2016 Republican primary started with sixteen candidates, and there's a good chance that Trump would not have won had everyone but Jeb or Rubio dropped out close to the beginning. However, as we all know, that isn't what happened. Many candidates lingered, and it was possible for Trump to win many states despite only having roughly 25% support. Part of predicting the winner of the Republican nomination was predicting the future behavior of the 6-8 politicians still in the race. That is an inherently difficult prediction.
Second, polling for the general election is far more thorough and sophisticated than it was for the primary. More polling dollars are being devoted to fewer questions than they were in the primary, which increases the likelihood that the polls are accurate. Silver's model is partially based on polling results, so this improvement in accuracy for the polls improves the accuracy of his model.
There are probably additional reasons, but I think these two reasons in particular give us good reason to look past Silver's inaccurate prediction about Trump in the primary.
All bets are off the with regards to current polling data. Trump had already started cutting into the HC lead before Fridays bombshell. A bombshell that might well cost HC the election. Right now Silver like everyone else is just speculating about the repercussions.
All bets are off the with regards to current polling data. Fridays bombshell might well have been a major game changer. Right now Silver like everyone else is simply speculating about the repercussions.
We won't know until later in the week.
That is true. But my point wasn't that Silver's current number is perfectly accurate. It was simply that his inaccuracy during the primaries is not a good reason to dismiss him in the general.
About polling in general. If the polling data this election turns out to be way off my guess something will have to change. I was reading about a pollster who said today with call blocking or those not accepting recognizable caller ID it's all but impossible to secure a reliable sample of the target population.
About polling in general. If the polling data this election turns out to be way off my guess something will have to change. I was reading about a pollster who said today with call blocking or those not accepting recognizable caller ID it's all but impossible to secure a reliable sample of the target population.
This might be a valid concern. Do you remember who the pollster was?
I will be surprised if the polls are off more than a few points, though. In 2012, I believe the RCP average was off by about three points. That's within the realm of credibility for me. Currently, HRC is up by about 4.6 points in the RCP average. If the election were held today, I would say an HRC win of at least a point or two would validate the polls. A Trump win of any kind would invalidate them in my book.
About polling in general. If the polling data this election turns out to be way off my guess something will have to change. I was reading about a pollster who said today with call blocking or those not accepting recognizable caller ID it's all but impossible to secure a reliable sample of the target population.
To determine if there is a problem you can't just look at the last polls before an election.
When a poll like The Washington Post poll shows a 10 point swing in a relatively small window, something is wrong.
This might be a valid concern. Do you remember who the pollster was?
I will be surprised if the polls are off more than a few points, though. In 2012, I believe the RCP average was off by about three points. That's within the realm of credibility for me. Currently, HRC is up by about 4.6 points in the RCP average. If the election were held today, I would say an HRC win of at least a point or two would validate the polls. A Trump win of any kind would invalidate them in my book.
The RCP average was about 3 points more friendly to Romney than the actual as you mentioned. The actual state polling tended to be more accurate, which is a big part of the reason Nate's models put a much heavier weight on the state polls, even when extrapolating the national numbers.
When a poll like The Washington Post poll shows a 10 point swing in a relatively small window, something is wrong.
If it's a tracking poll than big swings are expected. I take all tracking polls with a grain of salt and really try to ignore them if possible.
Tracking polls track the exact same people(the same individuals) time after time and keep getting their input over different dates. If some of those being tracked don't respond on a specific date(but respond other dates), then the poll can swing wildly. One of the tracking polls has 1 black male that sometimes doesn't respond and it has caused the poll to shift 8 points because he's weighted heavily in the poll.
About polling in general. If the polling data this election turns out to be way off my guess something will have to change. I was reading about a pollster who said today with call blocking or those not accepting recognizable caller ID it's all but impossible to secure a reliable sample of the target population.
Exit polls have been shown to skew Democratic since Republicans don't like talking to the media. I suspect Republicans are also more likely to screen their calls.
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.
Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.