For election nerds: correlations
       |           

Welcome, Guest. Please login or register.
Did you miss your activation email?
March 18, 2024, 11:08:03 PM
News: Election Simulator 2.0 Released. Senate/Gubernatorial maps, proportional electoral votes, and more - Read more

  Talk Elections
  Presidential Elections - Analysis and Discussion
  U.S. Presidential Election Results (Moderator: Dereich)
  For election nerds: correlations
« previous next »
Pages: [1]
Author Topic: For election nerds: correlations  (Read 2650 times)
buritobr
YaBB God
*****
Posts: 3,591


Show only this user's posts in this thread
« on: March 05, 2016, 05:43:07 PM »

I think everyone here have learned correlation in the statistics class at school. The closer to +1 means very positive correlated, the closer to zero means no correlation, the closer to -1 means very negative correlated.
An election is realigning (concerning the geography) when the correlation to the previous election is low.

Here we can see the correlation of the D-R margin by state to the previous election
2012   0.9822
2008   0.9431
2004   0.9762
2000   0.9563
1996   0.9299
1992   0.8956
1988   0.9347
1984   0.8667
1980   0.8921
1976   0.4157
1972   0.6859
1968   0.4240
1964   -0.0803
1960   0.5389
1956   0.8354
1952   0.5750
1948   0.5409
1944   0.6160
1940   0.9135
1936   0.9422
1932   0.7720
1928   0.8100
1924   0.9689
1920   0.9138

We can see that until 1940 (except 1928) and after 1980, there were few changes of the margins by state in comparison to the previous election. Between 1944 and 1980, there was a mess. The most realigning election took place in 1964, the only one in which the correlation to the previous election is negative.

Here we can see the correlation of each election to 2012
2008   0.9822
2004   0.9521
2000   0.9196
1996   0.8543
1992   0.7887
1988   0.7383
1984   0.7599
1980   0.6276
1976   0.4297
1972   0.7173
1968   0.6897
1964   0.5610
1960   0.2666
1956   -0.2439
1952   -0.0920
1948   -0.2941
1944   -0.2139
1940   -0.2552
1936   -0.3759
1932   -0.4150
1928   -0.1670
1924   -0.3650
1920   -0.3269
1916   -0.4139

Since 1960, every election has positive correlation to 2012. The correlation has been growing since 1960, except in the accident of 1976-1980. Until 1936, there was a heavy negative correlation to modern elections, except the accident of 1928.


Other interesting facts

Correlation 1980 2012
Considering Reagan + Anderson only one candidate: 0.4993
Considering Carter + Anderson only one candidate: 0.7327 (close to a modern election)

Correlation 1968 2012
Considering Humphrey + Wallace only one candidate: 0.2388
Considering Nixon + Wallace only one candidate: 0.7383 (close to a modern election)

Correlation 1956 1964: -0.6044 (the most negative correlation)

Logged
SingingAnalyst
mathstatman
YaBB God
*****
Posts: 3,639
United States


Show only this user's posts in this thread
« Reply #1 on: March 05, 2016, 05:53:53 PM »
« Edited: March 05, 2016, 05:55:29 PM by mathstatman »

Great work.

I'm surprised the 1976-2012 correlation is not negative, given the large number of Ford-Obama and Carter-Romney states.

Is every state + DC weighted equally? DC is of course an outlier.

I predict the 2012-2016 correlation to be around 0.98--similar to the 2008-12 correlation and one of the highest in US history.

I would love to see the same analysis done on counties.
Logged
buritobr
YaBB God
*****
Posts: 3,591


Show only this user's posts in this thread
« Reply #2 on: March 06, 2016, 08:13:06 AM »

I understand the positive correlation between 1976 and 2012. There are almost equal four groups of states: Carter Obama, Carter Romney, Ford Obama, Ford Romney. But states which had big margins for different parties are rare: Arkansas, Alabama, West Virginia in one side and Vermont in the other side. Big margins for Carter and Obama in DC, Massachusetts, Rhode Island, Maryland and Minnesota, and big margins for Ford and Romney in the Plains and in the Rocky Mountains could have made the difference. In the Ford Obama states in the Northeast, Midwest and Pacific, Ford won by small margins (except Vermont and New Hampshire). But before the calculation, I expected a smaller correlation.
The biggest surprise for me was 1988: it is slightly less correlated with 2012 than 1984 is correlated with 2012. Maybe, because Dukakis overperformed in the Plains and in the Rocky Mountains for a democrat. Many posts in this forum told that 1988 was the beggining of the maps of the modern era. Actually, 1984 was closer to the 21th century, but it was submerged by the huge R national margin.
Logged
kcguy
Jr. Member
***
Posts: 1,031
Romania


Show only this user's posts in this thread
« Reply #3 on: March 06, 2016, 12:04:41 PM »

As a self-proclaimed nerd, I can say that this thread is really cool.
Logged
Asian Nazi
d32123
Sr. Member
****
Posts: 2,523
China


Show only this user's posts in this thread
« Reply #4 on: March 06, 2016, 12:27:51 PM »

Nice thread.

To me, this reinforces a point I've been making around here for a while.  Realignment isn't something that happens over the course of just one election, it's a long process that shapes out over generations.
Logged
Nym90
nym90
Atlas Icon
*****
Posts: 16,260
United States


Political Matrix
E: -5.55, S: -2.96

P P P
Show only this user's posts in this thread
« Reply #5 on: March 06, 2016, 06:32:20 PM »

Well done. It would be fascinating to see a table showing the correlation between all elections in your dataset.
Logged
SingingAnalyst
mathstatman
YaBB God
*****
Posts: 3,639
United States


Show only this user's posts in this thread
« Reply #6 on: March 06, 2016, 06:36:05 PM »

I understand the positive correlation between 1976 and 2012. There are almost equal four groups of states: Carter Obama, Carter Romney, Ford Obama, Ford Romney. But states which had big margins for different parties are rare: Arkansas, Alabama, West Virginia in one side and Vermont in the other side. Big margins for Carter and Obama in DC, Massachusetts, Rhode Island, Maryland and Minnesota, and big margins for Ford and Romney in the Plains and in the Rocky Mountains could have made the difference. In the Ford Obama states in the Northeast, Midwest and Pacific, Ford won by small margins (except Vermont and New Hampshire). But before the calculation, I expected a smaller correlation.
The biggest surprise for me was 1988: it is slightly less correlated with 2012 than 1984 is correlated with 2012. Maybe, because Dukakis overperformed in the Plains and in the Rocky Mountains for a democrat. Many posts in this forum told that 1988 was the beggining of the maps of the modern era. Actually, 1984 was closer to the 21th century, but it was submerged by the huge R national margin.
Good point, and I think 1968 was the beginning of modern political alignments (1968 even had a primordial 4-point gender gap, with women leaning more to Humphrey). There were some exceptions (such as the fact that Catholics now vote near the national average whereas in 1968 if only Catholics had voted, Humphrey would have won easily), but the geographic markers were there in 1968 probably as much as in 1988 or 1992 even.
Logged
SingingAnalyst
mathstatman
YaBB God
*****
Posts: 3,639
United States


Show only this user's posts in this thread
« Reply #7 on: March 06, 2016, 06:38:41 PM »

Well done. It would be fascinating to see a table showing the correlation between all elections in your dataset.
I would think the correlation between 1908 and 2008 would be close to -1.
Logged
buritobr
YaBB God
*****
Posts: 3,591


Show only this user's posts in this thread
« Reply #8 on: March 06, 2016, 08:16:26 PM »

Well done. It would be fascinating to see a table showing the correlation between all elections in your dataset.
I would think the correlation between 1908 and 2008 would be close to -1.

I stopped in the elections in which the map is complete, and I ignored 1912 because there were too many candidates. But I can do further.
I don't think 1908 and 2008 are close to -1. Bryan won Colorado and Nevada, Taft won Utah, Bryan won Virginia, Taft won West Virginia.
Logged
buritobr
YaBB God
*****
Posts: 3,591


Show only this user's posts in this thread
« Reply #9 on: March 06, 2016, 08:22:38 PM »

Nice thread.

To me, this reinforces a point I've been making around here for a while.  Realignment isn't something that happens over the course of just one election, it's a long process that shapes out over generations.

My interpretation
The map until 1936 was stable (except the accident of 1928). There is no big difference between 1920 and 1936 concerning the distribuition of votes by state. The difference between these years was only the national margin. Then, there was a long realignment, which started in 1940 and finished in 2000, when the map became stable again. The biggest jump too place between 1956 and 1964. There were some backward movements in this period. In 1956, the map moved to the past, although the candidates were the same of the previous election. The other backward movement took place in 1976/1980.
Logged
Del Tachi
Republican95
Atlas Icon
*****
Posts: 17,665
United States


Political Matrix
E: 0.52, S: 1.46

P P P

Show only this user's posts in this thread
« Reply #10 on: March 07, 2016, 05:13:44 AM »

We can see that until 1940 (except 1928) and after 1980, there were few changes of the margins by state in comparison to the previous election. Between 1944 and 1980, there was a mess. The most realigning election took place in 1964, the only one in which the correlation to the previous election is negative.


South in y'all's mouth.
Logged
buritobr
YaBB God
*****
Posts: 3,591


Show only this user's posts in this thread
« Reply #11 on: March 07, 2016, 06:20:56 PM »

Correlations

1896/2012: -0.5710
1896/2004: -0.6073
Logged
buritobr
YaBB God
*****
Posts: 3,591


Show only this user's posts in this thread
« Reply #12 on: March 08, 2016, 09:53:47 PM »

I think the 2016 Democratic Primaries will have some correlation with the general election of 1976. Hillary Clinton is winning many states won by Jimmy Carter, and Bernie Sanders is winning many states won by Gerald Ford.
Logged
mianfei
Jr. Member
***
Posts: 322
Show only this user's posts in this thread
« Reply #13 on: February 08, 2018, 07:03:22 AM »

Going back further (excluding 1912 and 1892 where third-party candidacies and the absence from the ballot of major parties makes things very difficult) for correlations between 2012 and older elections:

2012/1908-0.355483058
2012/1904-0.234931391
2012/1900-0.373803112
2012/1896-0.570924893
2012/1888-0.478979666
2012/1884-0.463004505
2012/1880-0.467247571

The relatively low 1904 correlation is concentrated entirely in two regions:

  • The four mid-Atlantic states of New York, New Jersey, Maryland and Delaware that were more D than the nation in both 1904 and 2012
  • The Mountain and Plains states of Idaho, Utah, Wyoming, North Dakota, South Dakota, Nebraska and Kansas that were much more Republican than the nation in both elections.

For adjacent election correlations (doing 1896 and 1888 due to the 1892 third party candidacy:

1908/1904+0.952293518
1904/1900+0.89919504
1900/1896+0.827327068
1896/1888+0.661423031
1888/1884+0.947724405
1884/1880+0.896881201

It might be noted that the 1896-to-1888 correlation is really lower than the figure given here, since the seven states admitted between 1888 and 1896 all voted for Bryan yet except for the special case of Utah had leaned Republican at the territorial level.
Logged
Annatar
Jr. Member
***
Posts: 968
Australia


Show only this user's posts in this thread
« Reply #14 on: February 09, 2018, 06:48:39 AM »

Going back further (excluding 1912 and 1892 where third-party candidacies and the absence from the ballot of major parties makes things very difficult) for correlations between 2012 and older elections:

2012/1908-0.355483058
2012/1904-0.234931391
2012/1900-0.373803112
2012/1896-0.570924893
2012/1888-0.478979666
2012/1884-0.463004505
2012/1880-0.467247571

The relatively low 1904 correlation is concentrated entirely in two regions:

  • The four mid-Atlantic states of New York, New Jersey, Maryland and Delaware that were more D than the nation in both 1904 and 2012
  • The Mountain and Plains states of Idaho, Utah, Wyoming, North Dakota, South Dakota, Nebraska and Kansas that were much more Republican than the nation in both elections.

For adjacent election correlations (doing 1896 and 1888 due to the 1892 third party candidacy:

1908/1904+0.952293518
1904/1900+0.89919504
1900/1896+0.827327068
1896/1888+0.661423031
1888/1884+0.947724405
1884/1880+0.896881201

It might be noted that the 1896-to-1888 correlation is really lower than the figure given here, since the seven states admitted between 1888 and 1896 all voted for Bryan yet except for the special case of Utah had leaned Republican at the territorial level.


Thanks for the analysis, how do the correlations look compared to 2016, they would be pretty similar I imagine as the change from 2012 to 2016 overall was quite marginal.
Logged
Pages: [1]  
« previous next »
Jump to:  


Login with username, password and session length

Terms of Service - DMCA Agent and Policy - Privacy Policy and Cookies

Powered by SMF 1.1.21 | SMF © 2015, Simple Machines

Page created in 0.051 seconds with 11 queries.