Saturday, May 04, 2013

Why Does Nathan Silver's Interactive Map for 2012 Not Seem To Match A Prior NYTimes Post?



Nate Silver has published a handy interactive map for the impact of various minority groups in past/future elections.  Relying on that map, Byron York published the following article which claimed, using Nate's map, that "Romney would have had to win 73 percent of the Hispanic vote to prevail in 2012." I checked Byron's results against Nate's map, and it appears Byron is correctly utilizing it. In fact, from playing with the Map's numbers and holding everything else constant for the 2012 election, here is the minimum number of the Hispanic votes that the Map claims Obama was required to obtain to win the  following states:

Ohio - 25% and tossup at 24%
Virginia - 37% and tossup at 36% 
Pennsylvania - 29% and tossup at 28%
Colorado - 52%
Florida - 69%

These numbers though seem to conflict with a prior NYTimes Nov. 20, 2012 post, which utilized Exit Poll data reported by Edison Research.  That post said different numbers of Hispanic voters were required for the Dems to claim various states that they in fact carried.  Those states, and the number of Hispanic voters that it claimed Obama need to carry them, are as follows: 


Ohio - 22%
Virginia - 33%
Pennsylvania - 37%
Colorado - "just over 58%"
Florida - 58%

There is a pretty substantial difference with what the Map and the Nov. 20, 2012 article is reporting was needed, for some of these states.  Because I am not a stat professional and could easily be making a mistake I am not aware of I have posed the following questions to Nate Silver on his 538 blog and am hoping for a clarifying answer:
(a) Am I/Byron York using your interactive map correctly?;
(b) Am I interpreting your interactive map correctly?
(c) Was the Nov. 20, 2012 New York Times post incorrect?;
(d) Is your map incorrect?; and/or
(e) Any other explanation.   
For if it turns out that that there is a mistake in the Nate's interactive map, Byron and his allies might need to rethink their use of it to diminish the importance of the Hispanic vote in the 2012 election. 

2 comments:

raxelk said...

Hi Elliot, First of all, I really enjoy all your political analysis and am 100% rooting you on with immigration (and all the other social issues), even if I disagree on economics (All I'll say on that is the only pundit who has been nearly as right in all his predictions the last 5 years as Nate Silver has been Paul Krugman and I've done very well financially following both their predictions and thats not based on blindly following them, but on very deep dives into their analysis based on my extensive background in both economics and statistics)... As far as the discrepency you noticed in the nyt articles. as an amateur statistician, there is an easy explanation for this. Nate Silver, as usual, is right, and the prior Nov 20, 2012 article is wrong. The key is Nate is using the actual final election results (see https://docs.google.com/spreadsheet/lv?key=0AjYj9mXElO_QdHpla01oWE1jOFZRbnhJZkZpVFNKeVE&toomany=true#gid=19), whereas the original article used exit polling, which has a high margin of error (just ask Al Gore in 2000 -- all the exit polls predicted he would win). So its not at all surprising to see that degree of variation between the conclusions based on exit poll results and Nate's conclusions based on the actual data. Also, even based on actual data, results as of Nov 19 were substantially different than the final results when all votes were counted, which in some cases was not until the end of the year. So, for example, Nate, on Nov 19, had the winning margin in California as 20.9%, when in fact, after all votes were counted, it ended up being 23.12%. (see http://fivethirtyeight.blogs.nytimes.com/2012/11/19/the-2012-election-in-a-relative-sense/ So, the writers of the Nov 20 article probably did the best they could with the data they had, but it was just not very accurate data to begin with. Anyway, it is best to normally assume that in any statistical disagreement, Nate is probably right. I certainly have been making good (and easy) money the last couple national elections betting on Nate's analysis :-) ...

Unknown said...

I really like how your class timings of your blog. I enjoyed reading your blog and it is both instructional and interesting.Thanks!

lawyers in colorado springs