Total Posts:4|Showing Posts:1-4
Jump to topic:

FiveThirtyEight (with a Pinch of Salt)

NHN
Posts: 624
Add as Friend
Challenge to a Debate
Send a Message
9/16/2016 2:19:42 PM
Posted: 2 months ago
Nate Silver, editor-in-chief of ESPN's election fun house blog, is at it again with his click-bait reporting. Voter uncertainties are doubly registered and the movability exaggerated to make the election seem more competitive.

In June 2012, Silver deliberately switched his logarithms to claim that Obama held a "tenuous advantage" ahead of Romney (http://fivethirtyeight.blogs.nytimes.com...). Feel free to revisit the extreme "jumps" in Obama's chances to win, especially as the October 3, 16, 18, and 22 debates drew near (http://fivethirtyeight.blogs.nytimes.com...).

This is of course done for the sake of commercial pressure (http://www.electoral-vote.com...), and you who avidly defend Silver should feel free to do so. His economic index logarithms bear some merit and put the biased house effects of pollsters like Gallup and Rasmussen to shame. (And, as the election nears the final weeks, he will remove the uncertainties and movability factors, as in 2012.)

But as you go along for the ride, don't pretend for a second that ESPN is seriously vested in the accuracy of election forecasts.
dylancatlow
Posts: 12,244
Add as Friend
Challenge to a Debate
Send a Message
9/16/2016 2:38:06 PM
Posted: 2 months ago
I've come to view Nate Silver's predictions as little more than a translation of poll data into win percentages. It doesn't appear that he's able to TRACK where poll numbers are headed all that well. Of course, it's slightly more nuanced than that (like, he claims to account for post-convention bias and overall economic conditions) but that's basically it.
dylancatlow
Posts: 12,244
Add as Friend
Challenge to a Debate
Send a Message
9/16/2016 2:46:29 PM
Posted: 2 months ago
In other words, he's like "candidate A is five points ahead of candidate B in the polls. In the past, a candidate five points ahead on election day has won x percent of the time, and a candidate five points ahead at this point in the race has typically ended up with X lead on election day". Now, simulating electoral college outcomes is way too complex for the average person to calculate, so I guess that's pretty useful.
NHN
Posts: 624
Add as Friend
Challenge to a Debate
Send a Message
9/16/2016 3:08:22 PM
Posted: 2 months ago
At 9/16/2016 2:46:29 PM, dylancatlow wrote:
In other words, he's like "candidate A is five points ahead of candidate B in the polls. In the past, a candidate five points ahead on election day has won x percent of the time, and a candidate five points ahead at this point in the race has typically ended up with X lead on election day".
That's basically the horrible Cook Political Report, which relies on reporting and "experts." At least Silver is an advanced statistician.

By the way, compare the main forecasters (http://www.nytimes.com...). There shouldn't be a difference to that degree between 538 and the Princeton Election Consortium.

Now, simulating electoral college outcomes is way too complex for the average person to calculate, so I guess that's pretty useful.
Absolutely. There is no way for anyone, who isn't a forecaster, to have access to either the data or the logarithms. Silver and ESPN use it for click-bait. The sharp turns of 2012 will likely be repeated as the debates take off on September 26.