Elections USA: popular vote, rust belt and the Californian Effect

First things first, I have written another couple of posts in this blog in which I was explaining why I was sure Trump could not win the elections in USA. In a few words I thought that he was going to be perceived as so damaging for the economy, that by the day of the elections the many economic powerhouses in the US would have all supported Clinton. My reasoning was that populism as we are experiencing in Europe would have also been tamed at least in part thanks to Obama’s policies, which have not granted redistribution (he did not have the majority at the House to be fair) but they still are far better when compared with the austerity measures that have characterised the EU since the crisis in 2008. Well, I was wrong, so I need to cope with the results. One way I have been taught in Italy, is to analyse the data. If interested, you can download the raw data I collected from wikipedia on the elections from 2000 to 2016, HERE.

I have updated post a couple of times, with new images and data, to include results variation after they have become available. Last update, December 1st 2016.

The overall turnout (54%, 135.7 mln ballots) started circulating almost immediately, jointly with it the result of the “popular vote” which gave initially a small margin in favour of Clinton (initially estimated at 200’000, it is now close to 2.5 million). Despite a contained loss of votes in comparison with Obama 2012 (around 1 mln), she ended up losing the elections anyway. Possibly due to the initial results, which gave an estimated difference close to 5 mln votes in the comparison between Clinton 2016 and Obama 2012, many jumped to the conclusion that the election was lost either due to her being unable to attract voters or due to the usual “liberals” who decided to screw the world not showing up for the elections or voting for third parties. Bernie Sanders voters, millennials, somebody has to be blamed:


It is of course an oversimplification of what actually happened. First of all it turns out the turn out (!) is not as low as it might be considered when comparing the result with the elections involving Obama. Let’s go back a few more year and we see something different: the average across all candidates is 60.5 mln (median at 61.5) with Clinton above the average, and Bush 2000 as negative outlier. In a few words, Clinton is on the high end of the amount of votes you would expect from a winner of the presidential elections since year 2000.


Where is the problem? The distribution of the votes in USA is much more important than the total number. Clinton might have had ten millions votes more than Trump and she could still be defeated, if those votes were all coming from States she has won. This is why the total number of ballots is little informative and we can have a better grasp of what happened in a State by State comparison:


Initially this article pointed out that California, NY and WA, which were all won anyway, were responsible for most of the loss of votes in the comparison between Clinton 2016 and Obama 2012. Initial data pointed towards a combined loss of around 3 mln votes in these three States, making up most of the supposed 5 mln votes of difference with Obama (thus the “California effect” in the title, since it looked like Clinton was 2 mln votes “short” in California alone, in comparison with Obama). As you can see in the chart above, this is not really the case, as it was simply an effect of delayed ballot counting (why does it take so much time anyway?). Actually, Clinton managed to get more votes than Obama in California (+0.7 mln), Florida (+0.25 mln) and Texas (+0.55 mln).

Trump on the other side gained 1.25 mln votes in total (across all States) in comparison with Romney 2012, with the only significant negative comparison in California (-0.5 mln), which means that he has lost votes where it doesn’t hurt and he has gained in those States that required it in order to win the election. Among these, Florida (+0.45 mln) was a key one, since Trump would have lost the State should he have received the same ballots received by Romney in 2012. Yet, many argue the key of the election is elsewhere.

Before the elections, Michael Moore wrote a list of five reasons to explain why Trump could win the elections. The first point concerns the so called rust-belt: Pennsylvania, West Virginia, Ohio, Indiana, Michigan, Illinois, Iowa, and Wisconsin. is the industrialised area of the US where coal mines, steel production and car manufacturing has been damaged by ’90 liberalism and the competition that has pushed down salaries or triggered outsourcing. The chart for these states is basically self explanatory:


The eight states are represented in alphabetical order: Illinois, Indiana, Iowa , Michigan, Ohio , Pennsylvania, West Virginia, Wisconsin. Red bars represent the votes for the republican party in a comparison between Trump 2016 and Romney 2012. The blue bars represent the votes for the democratic party, in a comparison between Clinton 2016 and Obama 2012. Of these eight, Iowa (bar n. 3), Michigan (4), Ohio (5) , Pennsylvania (6), and Wisconsin (8) “changed colour” this year. So it is true Clinton did not manage to keep the democratic voters losing 1.53 mln votes in the eight States. Unfortunately it is also true that Trump managed to increase votes in comparison with Romney, gaining 0.8 mln in the same region. The result on these States implies some voters may have actually changed party. Honestly, I don’t see that as very likely: these might simply be new voters for Trump and lower turnout for Clinton. In any of these cases, considering the result, the next elections will see candidates that will try the same path as Trump, at the local or at the national level (a few weeks after writing this article, some extra data have been analysed by Slate, focusing on voters making $50000 per year or less).

It is difficult to establish a single cause for Clinton decrease and Trump success in these states. It is probably an effect of multiple converging factors. Among these, those highlighted by Micheal Moore are very likely to have played an important role. Probably, it is also necessary to add old fashion racism to the equation, So that the rural (white) America, which felt the loss of power with Obama (I am thinking north Florida), plus the impoverished (white) middle class (rust belt) screwed it up for everybody else. One last consideration for this specific elections: considering the small amount of votes that made the difference in so few states, I would be happy to blame the FBI as well. This hypothesis is also supported by the data showing the undecided voters who made up their minds in the last week, were mostly in favour of Trump.

PS this is also meant as a response to an article by an Italian Blogger I often read and appreciate.

Leave a Reply

Insert math as
Additional settings
Formula color
Text color
Type math using LaTeX
Nothing to preview
%d bloggers like this: