analysing the US election result, from Oxford, England
Seth Flaxman (Oxford), Dougal J. Sutherland (UCL), Yu-Xiang Wang (CMU), and Yee Whye Teh (Oxford), published on arXiv this morning an analysis of the US election, in what they called most appropriately a post-mortem. Using ecological inference already employed after Obama’s re-election. And producing graphs like the following one:
Xi'an's Og 
November 15, 2016 at 10:32 pm
This is a very interesting analysis. I like the data breakouts. It appears the chart with the yellow circles doesn’t match up with the data breakouts. For instance, the white/no degree bubble implies approximately 55% participation and about 85% vote for Trump. However, in the data breakouts, white shows 60% vote for Trump, some college at 54% and high-school or less at 53% vote for Trump — each far below 85%. Also, the participation for those groups are 54%, 49% and 25% — well below the 60%. Could you add in the supporting data used to populate the chart?
November 16, 2016 at 12:07 pm
Note that I am only copying what I read in this report: the data and the tool are available as supplementary material as listed in the paper. You should check with the paper or the authors for further information.
November 15, 2016 at 6:31 pm
I’m looking forward to a reviewed version of this work. It looks interesting, but still seems to be in draft form as others have noted. That last graph describing the model fit is poorly/hardly explained.
Thanks for sharing this with us.
November 15, 2016 at 7:03 pm
Thank you. It indeed looks somewhat rushed.
November 15, 2016 at 2:37 pm
Sad to say, the graph above tells this:
1) presidential election was just a “white affaire”
2) someone seemed to have spoken in favour of (and heard by) non-black working class, and it was not the Dem candidate
November 15, 2016 at 4:22 pm
I would not guess that anyone has spoken in favour of working classes from this graph. Or from this campaign…
November 15, 2016 at 4:50 am
actually, the polls at the National level were pretty much spot on. what was “wrong” was the urban/rural vote in mixed states, i.e. Mid-West and Mountain West. whether the rural vote in those states was legit or fiddled we will likely never know.
November 14, 2016 at 11:50 pm
There are some obvious mistakes in the article. First one I noticed is that the caption for Table 2 is the same as for Table 1. More egregiously, the participation rate for Asians (w or w/o degree) are both under 20% in the Figure 2, but listed as 39% in the table. I’m not even sure that Simpson’s Paradox can explain that one.
November 14, 2016 at 11:42 pm
Those numbers for Whites with and without college degrees look way more polarized than I recall from Election night.
November 15, 2016 at 8:21 pm
I agree. I doubt very much that 85% of whites w/o a college degree went for Trump, especially considering that college students are in that group.
November 14, 2016 at 4:17 pm
Looks like they were all wrong.
November 14, 2016 at 5:52 pm
This is not prediction, but postdiction, how can “they” be wrong?
November 14, 2016 at 11:29 pm
You mean the “common sense” degree? 8-)