Hashemite's Random Numbers and Stats about French elections

[Hash]

The greatness of the OVF on Geoclip has made me waste tons of time coming up with really random numbers and stats about French elections, mostly comparing election results with other elections or with demographic indicators.

I love calculating the correlation coefficient between two variables. That's the only thing I remember about Grade 11 math. Some of the results are really fascinating. If you don't know the correlation coefficient, you should look it up on Wiki.

Before I track down some of my more interesting stats, here are some interesting correlations:

Sarko (R1 2012) vs. median HH income: 0.41 (I think it's stronger of a correlation than in 2007)
Le Pen (R1 2012) vs. % of ouvriers: 0.466
Le Pen (R1 2012) vs. % of salariat modeste (ouvriers+employees): 0.475
Le Pen (R1 2007) vs. ouvriers: 0.471
But... Le Pen (R1 2002) vs. ouvriers: 0.237 (!)

White and null, R2 2012 vs Le Pen R1 2012: 0.58 (yeah...)
Evolution of abstention between both rounds, 2012 vs. Le Pen R1 2012: -0.49 (as expected: negative correlation between higher turnout in the runoff and Le Pen results)

And a little amusing one for kicks:
Sarko R1 2012 vs. Boutin R1 2002: 0.37

I'm working on some very cool stats about the 2005 referendum. I also picked up some very interesting (and myth-squashing) numbers about the FN vote in 88 and the 90s which I should put up. Eventually, I'll get around to doing something about the Catholic vs. Protestant vote in rural Alsace.

[Antonio the Sixth]

Correlation coefficients are a great thing. Shame I've more or less forgotten how to work them out.

Looking forward to what you come with.

[Hash]

Correlation coefficients are a great thing. Shame I've more or less forgotten how to work them out.

I've forgotten how to work them out to, but =CORREL(x,x) on Excel hasn't

[Antonio the Sixth]

Correlation coefficients are a great thing. Shame I've more or less forgotten how to work them out.

I've forgotten how to work them out to, but =CORREL(x,x) on Excel hasn't

Great, I'll try to experiment as well.

[minionofmidas]

Evolution of abstention between both rounds, 2012 vs. Le Pen R1 2012: -0.49 (as expected: negative correlation between higher turnout in the runoff and Le Pen results)

Turnout rose modestly in the Ile-de-France, and massively in the overseas possessions, and was effectively flat everywhere else, so, yeah. Probably no causation here.

[Hash]

Various correlations with 2012 candidates (first round)

Melenchon vs. median HH income: -0.27
Joly vs. median HH income: 0.19
Hollande vs. median HH income: -0.14
Bayrou vs. median HH income: 0.29

Joly vs. CPIS: 0.33
Joly vs. uni diploma/degree: 0.48
Joly vs. ouvriers: -0.47

[Antonio the Sixth]

Joly vs. uni diploma/degree: 0.48

Expected, but wow. That's the single strongest coefficient so far.

[Hash]

Epic bump for this graph:

2012 vote by self-defined ideology/political leanings. For the first time, Ipsos offered a different 'centre' and 'neither left nor right' (ni-ni) category which makes this all very interesting. The FN's curve is of particular interest.

[Franknburger]

Looking forward for what you will be coming up with (as always).

One technical remark: It would be nice if you also posted the R² values alongside with the coefficients.

For non-statisticians: R² measures how much of the variance is explained  by the regression. For example, if you have a regression between income and a party vote, a R² value of 0.56 would say that 56% of the votes can be explained by income.
For political-demographical analysis, R² values in the 50s / low 60s are pretty good. Anything above that is quite unusual and very remarkable. 35-50 mean it is a contributing, but not the only/ main factor, Below 35, we tend to get into the range of statistical noise.

[Antonio the Sixth]

Very interesting, though not exactly surprising.

[big bad fab]

(I know it's not about elections, but the other "odds and ends" or "trash" or "dustbin" thread is too old to bump)

Here are some (few, I must say) interesting (and sometimes hilarious: guess where there are many practitioners of shooting ? ) maps on sport membership in France:
http://www.sports.gouv.fr/IMG/pdf/atlas.pdf

Is basket-ball a MRP sport ?

[big bad fab]

http://www.ifop.com/media/pressdocument/607-1-document_file.pdf

A quite interesting analysis of differences between the "leftish northern" FN and the "rightish southern" FN.

[Hash]

http://www.ifop.com/media/pressdocument/607-1-document_file.pdf

A quite interesting analysis of differences between the "leftish northern" FN and the "rightish southern" FN.

Yeah, I saw that analysis. It's pretty interesting and it provides some interesting data, but at the same time it's a reiteration of what has been said in academic literature for quite some time now (specifically Nonna Mayer's 'Ces francais qui votent Le Pen') and what I've been saying as well.

[Diouf]

http://www.ifop.com/media/pressdocument/607-1-document_file.pdf

A quite interesting analysis of differences between the "leftish northern" FN and the "rightish southern" FN.

Can you translate the main points of the analysis?

[Leftbehind]

Yep, share the wealth!

[Hash]

Bump: For anyone interested by the topic, the Fondation Jean Jaures (which always publishes excellent and intelligent analyses on various topics, including some of the best election analysis - besides my blog ) has published a book, freely downloadable online, on religions in France. Even if you don't speak French, there are tons of unique maps in there.

http://www.jean-jaures.org/Publications/Essais/La-religion-devoilee

[joevsimp]

is correlation coefficient the same thing as R squared?

[Antonio the Sixth]

is correlation coefficient the same thing as R squared?

R-square is basically the coefficient of correlation squared.