A recent news program discussion of gerrymandering got me interested in number crunching election data. There was a recent metric developed for legally establishing gerrymandering - Partisan Gerrymandering and the Efficiency Gap.
I may consider if that has any bearing here but it doesn’t seem immediately likely. My main interest will be the 2016 presidential election. Obviously in a presidential election gerrymandering isn’t a concern. Getting data on these things can be difficult. But through a variety of sources I was able to get presidential election results by county for ’04 to ’16. Having it at the county level might give large enough sample sizes so some statistics methods could make sense.
data04 <- read.csv("county04.csv")
data08 <- read.csv("county08.csv")
data12 <- read.csv("county12.csv")
data16 <- read.csv("county16.csv")Different sources seemed to slightly differ from other sources. But I believe the results are correct enough for my purposes. For example…
diff16 <- data16$dem.Count16 - data16$gop.Count16
sum(diff16)## [1] 1348693Shows a democratic popular vote win of a million plus. This page indicates that this should actually be 2,868,691. Out of 62,984,825 total votes I consider getting the correct result and this close as accurate enough.
There has been considerable discussion of how Trump found his way through to a Electoral college win despite this. Talk of Russian hacker interference and Wikileaks, also the Comey letter. Mainly, the conventional wisdom seems to be that the Democrats simply didn’t reach the voters that they had to have. What I am mainly would like to consider is what the numbers might tell us about how this result was reached.
To begin with I am going to ignore actual vote total and just count up counties flipping and compare 2016 to earlier years concerning the probabilities of counties flipping. I represent a county flipping parties as (1 = Rep->Dem,0 = no change, or -1 = Dem->Rep). See the flips.csv dataset.
From election to election…
## 2008##
## -1 0 1
## 47 2734 339## 2012##
## -1 0 1
## 199 2911 10## 2016##
## -1 0 1
## 218 2882 20It can be seen that there was a big shift in counties from Republican to Democrat in the 2008 election. This might of been ‘making a statement’ votes against a unpopular Bush presidency or it could of been the historic Obama candidacy. For whatever reason it was a major shift.
This trend did not continue. Already in the 2012 election many more counties were reverting to Republican. Was this what is known in statistics as Reversion to the mean?. Normally Republican tending counties voting more as usual? Or something else? Maybe a reversal that the Democrats should of already been aware of for the 2016 election?
In the 2016 election counties continued flipping to Republican, only even more so. The last two elections more than reversed the Democrat county gains from 2008.
The overall probability that a random county might flip…
## [1] 0.08899573So there has been about a 9% chance over the last 3 presidential elections that a random county will flip from it’s party vote for the prior election. This is actually higher than I thought it would be.
That is the percentage across the country with all things being equal. But they are not equal, some states are party base states while others are considered swing states. The probability a county might flip from state to state can be considerably different.
The probabilites of a county flipping by state…
## Top 10 highest probability party flipping states## states sdata
## 48 WI 0.3611111
## 30 NH 0.3333333
## 22 MI 0.2771084
## 12 IA 0.2289562
## 8 DE 0.2222222
## 14 IL 0.2156863
## 34 NY 0.2043011
## 23 MN 0.1954023
## 21 ME 0.1875000
## 47 WA 0.1538462## Top 10 lowest probability party flipping states## states sdata
## 50 WY 0.028985507
## 13 ID 0.022727273
## 18 LA 0.020833333
## 43 TX 0.019685039
## 16 KS 0.006349206
## 3 AZ 0.000000000
## 7 DC 0.000000000
## 11 HI 0.000000000
## 19 MA 0.000000000
## 36 OK 0.000000000Somewhat interestingly, in checking the Top 10 states Clinton seems to have taken 7 of them for a significant 80 to 32 point lead in electoral college votes. Maybe a little interesting is that five states had no counties flip at all in the last three elections. Again, states can be considerably different.
Two states commonly talked about as important swing states are Florida and Ohio. Their probabilities for county flipping are…
## Florida## states sdata
## 9 FL 0.04975124## Ohio## states sdata
## 35 OH 0.07954545Neither is very high, Florida is actually fairly low with less than 5%. But both states flipped Republican from the 2012 election for a total of 47 electoral college votes.
Although the election to election numbers seemed significant for showing trends for my purposes it might make more sense to consider the states that flipped in 2016 and then drill down into their counties to see why. Or, the county flipping analysis is just too simplistic and actual vote tallies and possible unusual turnout need to be taken into account.
All things being equal doesn’t apply to states and it doesn’t to counties either. The Demcrat base counties tend to be urban and more highly populated than Republican. The number of Republican counties greatly exceeds the number of Democrat ones.
Over the last four elections the percentage of counties voting Republican was…
## [1] 0.7867788With the above mentioned trend in county flipping after the 2016 election this was…
## [1] 0.8410256And Clinton won on popular vote. Clearly she had wasted votes, this is what the gerrymandering “efficiency gap” is based on. Maybe it is a measure that can apply without gerrymandering. One consequence of this would seem to be that when a Democrat county flips it means more votes lost, on average, than when a Republican county flips. All things and all counties not being equal.
Taking things to the state level since this is where the electoral college votes are. We might ask at this level, what states flipped between 2012 and 2016?
##
## -1 0
## 6 44You can see that six states flipped from Democrat to Republican, the rest all stayed the same, with no Republican states flipping to Democrat ones.
What were the six states that flipped Republican?
## State EV diff.12 diff.16
## 9 FL 29 74309 -112911
## 12 IA 6 91927 -147314
## 22 MI 16 449313 -10704
## 35 OH 18 166272 -446841
## 38 PA 20 309840 -44292
## 48 WI 10 213019 -22748These are the only six states. They decided the election. All other states had the same result as in 2012.
How many electoral votes did these states account for?
## [1] 99Romney had 206 in in 2012, adding in the 99 gives Trump a definite win. It probably could of taken a couple less than the six for the win, if you want to figure that out. Then you could possibly say the remaining were actually the only states that had mattered.
However, for what follows I will consider each of the six that flipped.
The Republican margin of victory in the state was…
## [1] "112911 (1.24%)"## Florida Voter Turnout## 2004 2008 2012 2016
## Dem 3583544 4282367 4235270 4485745
## Gop 3964522 4046219 4162081 4605515
## Tot 7548066 8328586 8397351 9091260I had originally thought there might be some way to differentiate between new voters turning out and voters switching parties in their voting. Here the number of Democrat voters increases as well as the Republican ones. It is difficult to say that they had voters changing parties. But there really is no way to know.
Turnout across the election years will still be provided as an indication of voter motivation in the election.
For seeing what happened in Florida I will first go back to the county level and see what the flipped counties look like just for this state.
##
## -1 0
## 4 63There are only four counties that flipped to Republican this election.
## flipped.fl diff12.fl diff16.fl
## 1 Jefferson 137 -393
## 2 Monroe 158 -2936
## 3 Pinellas 25774 -5419
## 4 St. Lucie 9667 -3436The vote swings for the four flipped counties account for only a portion of the Republican margin of victory. The rest has to be other counties where the Republican vote outperformed or the Democrat vote underperformed. It may not be statewide but it does have to involve more than the flipped counties.
Consider the average party results for the 2012 and 2016 elections for statewide.
## V1 mean.d12 mean.r12 mean.d16 mean.r16
## 1 FL 82043 73109 80656.25 83702.25So statewide the Democrat vote tallies are on average down while the Republican tallies are up on average. Is this really statewide or can some smaller subset of counties be identified that swung the election?
From the averages it seems that Republican gains were more a factor than Democrat losses in voter turnout. But I will consder both, which were the biggest county gainers and which had the largest losses?
Looking at the Top 10 gains for each party…
## Top 10 Democrat county gains## names diff.d16
## 43 Miami-Dade 82230
## 48 Orange 55352
## 6 Broward 39526
## 50 Palm Beach 21993
## 28 Hillsborough 20264
## 49 Osceola 18086
## 35 Lee 14645
## 11 Collier 9263
## 57 Seminole 9176
## 58 St. Johns 7882## Top 10 Republican county gains## names diff.r16
## 35 Lee 37029
## 51 Pasco 29608
## 53 Polk 25650
## 52 Pinellas 25554
## 64 Volusia 25290
## 50 Palm Beach 23505
## 5 Brevard 22350
## 40 Manatee 16065
## 28 Hillsborough 16024
## 6 Broward 14789You can see that for the first three counties the Democrats did very well. This would be in their more urban, highly populated base. After that the Republican gains start being greater, this begins with Pinellas which you might remember was one of the counties that actually flipped between 2012 and 2016.
In table form…
## Democrat gains greater than Republican gains##
## FALSE TRUE
## 3 64Counties where the Republicans out-gained the Democrats number 64 to 3.
If the gains are totaled up…
## Total Democrat gains between 2012 and 2016## [1] 250475## Total Republican gains between 2012 and 2016## [1] 443434Both parties gained votes in 2016. You can’t really even say that the Democrats underperformed. It’s just that the Replubicans very considerably outperformed across the state. There is no key subset of counties for the Democrats to address. They need statewide improvement or less Republican improvement.
The Republican margin of victory in the state was…
## [1] "147314 (10.13%)"## Iowa Voter Turnout## 2004 2008 2012 2016
## Dem 741898 828940 816429 650790
## Gop 751957 682379 727928 798923
## Tot 1493855 1511319 1544357 1449713Flipped counties…
##
## -1 0
## 31 68Many counties actually flipped here, almost half.
## flipped.ia diff12.ia diff16.ia
## 1 Allamakee 281 -1663
## 2 Boone 917 -1941
## 3 Bremer 327 -1850
## 4 Buchanan 1460 -1538
## 5 Cedar 421 -1697
## 6 Cerro Gordo 3120 -1743
## 7 Chickasaw 716 -1475
## 8 Clarke 63 -1243
## 9 Clayton 622 -2073
## 10 Clinton 5700 -1170
## 11 Des Moines 3743 -1301
## 12 Dubuque 7335 -610
## 13 Fayette 1225 -1925
## 14 Floyd 1208 -1194
## 15 Howard 974 -937
## 16 Jackson 1706 -1984
## 17 Jasper 1372 -3448
## 18 Jefferson 1343 -36
## 19 Jones 802 -1939
## 20 Lee 2641 -2567
## 21 Louisa 33 -1417
## 22 Marshall 1674 -1502
## 23 Mitchell 188 -1301
## 24 Muscatine 3061 -1265
## 25 Poweshiek 898 -664
## 26 Tama 661 -1774
## 27 Union 229 -1601
## 28 Wapello 1839 -3119
## 29 Webster 1005 -3756
## 30 Winneshiek 1627 -94
## 31 Worth 600 -920I don’t know if I have the data to indicate whether this result is more unusual or that the state favored the Democrat Barack Obama as it did. But based only on the data I have, while not having the most electoral votes, it is clear that for this election the Democrat candidate underperformed badly.
The Republican margin of victory in the state was…
## [1] "10704 (0.24%)"Very close.
## Michigan Voter Turnout## 2004 2008 2012 2016
## Dem 2479183 2872579 2561911 2268193
## Gop 2313746 2048639 2112673 2279805
## Tot 4792929 4921218 4674584 4547998Flipped counties…
##
## -1 0
## 12 71And which ones…
## flipped.mi diff12.mi diff16.mi
## 1 Bay 3062 -6686
## 2 Calhoun 932 -7335
## 3 Eaton 1719 -3074
## 4 Gogebic 614 -1094
## 5 Isabella 2238 -934
## 6 Lake 265 -1220
## 7 Macomb 16096 -48351
## 8 Manistee 731 -1936
## 9 Monroe 717 -16396
## 10 Saginaw 11656 -1074
## 11 Shiawassee 1235 -6685
## 12 Van Buren 148 -4632Leaning Republican.
The overall averages…
## V1 mean.d12 mean.r12 mean.d16 mean.r16
## 1 MI 36871.08 33586.67 30406.08 38690.83More of a Democrat loss than Florida but not as much in Republican gains. But both again favor the Republicans.
The Top 10 gains for each party…
## Top 10 Democrat county gains## names diff.d16
## 81 Washtenaw 7234
## 41 Kent 4007
## 70 Ottawa 3208
## 33 Ingham 363
## 45 Leelanau 198
## 28 Grand Traverse 89
## 42 Keweenaw -55
## 24 Emmet -253
## 48 Luce -270
## 66 Ontonagon -410## Top 10 Republican county gains## names diff.r16
## 50 Macomb 32693
## 82 Wayne 15322
## 25 Genesee 12367
## 77 St. Clair 9991
## 58 Monroe 7662
## 44 Lapeer 6312
## 47 Livingston 5665
## 61 Muskegon 5080
## 3 Allegan 4789
## 9 Bay 4592The Democrats managed gains in only six counties. They didn’t do as well as in Florida in hanging onto their base, even losing ground in Macomb county which appears to be part of metro Detroit. A more high population Democrat type county.
The loss seems statewide again. Although more due to the Democrat’s not doing as well as they might of rather than large scale Republican outperforming. This state could easily of gone Democrat. Macomb county alone would of given it to them if they had hung onto it.
The Republican margin of victory in the state was…
## [1] "446841 (8.54%)"A big Republican win.
## Ohio Voter Turnout## 2004 2008 2012 2016
## Dem 2739952 2940044 2697260 2317001
## Gop 2858727 2677820 2593779 2771984
## Tot 5598679 5617864 5291039 5088985Flipped counties…
##
## -1 0
## 9 79Which ones…
## flipped.oh diff12.oh diff16.oh
## 1 Ashtabula 5074 -7564
## 2 Erie 4314 -3609
## 3 Lorain 20017 -388
## 4 Montgomery 7795 -3105
## 5 Ottawa 891 -4253
## 6 Portage 3617 -7515
## 7 Sandusky 544 -6312
## 8 Trumbull 21901 -6022
## 9 Wood 2599 -5294The overall averages…
## V1 mean.d12 mean.r12 mean.d16 mean.r16
## 1 MI 45044.56 37627.67 36822.67 41718.44A large Democrat decline and moderate Republican gains.
The Top 10 gains for each party…
## Top 10 Democrat county gains## names diff.d16
## 25 Franklin 10307
## 21 Delaware 3453
## 83 Warren 1129
## 38 Holmes -802
## 27 Gallia -854
## 61 Noble -881
## 31 Hamilton -921
## 80 Union -1016
## 82 Vinton -1050
## 58 Morgan -1061## Top 10 Republican county gains## names diff.r16
## 50 Mahoning 11106
## 78 Trumbull 10607
## 76 Stark 9387
## 47 Lorain 7251
## 15 Columbiana 6308
## 48 Lucas 6002
## 45 Licking 5742
## 43 Lake 5477
## 4 Ashtabula 4865
## 60 Muskingum 4762It was a little difficult for me to understand how the numbers I’m getting accounted for the large Republican margin of victory. But with almost every county statewide shifting more Republican, and each Republican gain representing a Democrat loss, it seems that it does.
The Republican margin of victory in the state was…
## [1] "44292 (0.75%)"## Pennsylvania Voter Turnout## 2004 2008 2012 2016
## Dem 2938095 3276363 2907448 2844705
## Gop 2793847 2655885 2619583 2912941
## Tot 5731942 5932248 5527031 5757646Flipped counties…
##
## -1 0 1
## 3 62 2A couple actually flipped to Democrat.
Which ones flipped…
## flipped.pa diff12.pa diff16.pa
## 1 Centre -20 1456
## 2 Chester -1048 24606
## 3 Erie 19034 -2348
## 4 Luzerne 6005 -26054
## 5 Northampton 5772 -5448The overall averages…
## V1 mean.d12 mean.r12 mean.d16 mean.r16
## 1 PA 70006.60 64058.00 69790.60 71348.20The Democrats lost hardly anything at all but the Republicans posted decent gains.
The Top 10 gains for each party…
## Top 10 Democrat county gains## names diff.d16
## 46 Montgomery 23502
## 15 Chester 17956
## 2 Allegheny 14866
## 9 Bucks 6263
## 23 Delaware 4364
## 51 Philadelphia 3518
## 36 Lancaster 2958
## 14 Centre 2878
## 48 Northampton 922
## 21 Cumberland 299## Top 10 Republican county gains## names diff.r16
## 40 Luzerne 19539
## 67 York 15357
## 51 Philadelphia 13578
## 35 Lackawanna 13372
## 65 Westmoreland 12998
## 6 Berks 12237
## 48 Northampton 12142
## 54 Schuylkill 11781
## 25 Erie 11066
## 26 Fayette 8543The Democrats did well in the top spots. These appear from Wikipedia to be somewhat typical base counties for them, highly populated and in the case of Chester high income. But getting past the first three places the Republican gains again take over in being better than the Democrat ones. Actually, even Philidelphia county itself shows superior Republican gains so Democrats were not entirely successful in preserving their base.
The Republican margin of victory in the state was…
## [1] "22748 (0.82%)"Again, pretty close.
## Wisconsin Voter Turnout## 2004 2008 2012 2016
## Dem 1489504 1677211 1613950 1382210
## Gop 1478120 1262393 1408746 1409467
## Tot 2967624 2939604 3022696 2791677Flipped counties…
##
## -1 0
## 22 50Which ones flipped…
## flipped.wi diff12.wi diff16.wi
## 1 Adams 894 -2203
## 2 Buffalo 217 -1518
## 3 Columbia 4144 -635
## 4 Crawford 1558 -418
## 5 Door 1229 -558
## 6 Dunn 1093 -2462
## 7 Forest 249 -1204
## 8 Grant 3319 -2300
## 9 Jackson 1397 -1086
## 10 Juneau 837 -3088
## 11 Kenosha 9896 -255
## 12 Lafayette 1221 -689
## 13 Lincoln 99 -3030
## 14 Marquette 354 -1904
## 15 Pepin 82 -883
## 16 Price 3 -1891
## 17 Racine 3714 -4114
## 18 Richland 1389 -444
## 19 Sawyer 46 -1779
## 20 Trempealeau 1898 -1725
## 21 Vernon 2096 -643
## 22 Winnebago 3337 -6393The overall averages…
## V1 mean.d12 mean.r12 mean.d16 mean.r16
## 1 WI 12237.05 10461.05 9504.32 11287.14Fairly small both ways, but Democrat losses and Republican gains.
The Top 10 gains for each party…
## Top 10 Democrat county gains## names diff.d16
## 13 Dane 2117
## 68 Waukesha 1582
## 46 Ozaukee 1092
## 40 Menominee -188
## 19 Florence -286
## 26 Iron -510
## 47 Pepin -530
## 21 Forest -838
## 7 Burnett -1033
## 6 Buffalo -1038## Top 10 Republican county gains## names diff.r16
## 45 Outagamie 4285
## 56 Sauk 3042
## 43 Oconto 2564
## 9 Chippewa 2541
## 5 Brown 2454
## 37 Marathon 2442
## 38 Marinette 2386
## 3 Barron 2182
## 22 Grant 2092
## 69 Waupaca 2024The familiar pattern of very few Democrat gains but consistent Republican ones. It wasn’t a big margin of victory but no county stands out here as one that might of swung the state.
Overall voter turnout.
## Overall Voter Turnout## 2004 2008 2012 2016
## Dem 58789456 69373764 62228082 62411041
## Gop 61711414 59756255 58772655 61062348
## Tot 120500870 129130019 121000737 1234733892016 was the second best of the four elections for both parties. Democrats better than Republican, again they had the popular vote edge.
Swing states are still of interest to me. Again, Florida and Ohio over the years have been frequently discussed as swing states and both are included above. In our list of six critical states for 2016 compared to the 2012 election. We found this list in hindsight. But how well can you tell for 2020 who the critical battleground states will be?
It seemed to me that are some similarites here to finance. In finance risk is measured as volatility. How extremely do prices, or whatever, swing up and down? I did some checking for anything related to elections and volatility. I came across the Pedersen Index. But this mainly seems concerned with volatility between multiple parties in European elections. Not regional volatility in a two party system. Not useful for my purposes.
Somewhat interesting again is efficiency gap for gerrymandering almost seems to relate. Looking at this Washington Post article, specifically towards the bottom, the graphic where it says “Some swing states have low scores”. You will notice that four of our six states of interest are in the bottom 10 and Pennsylvania and Iowa aren’t that much higher. This I assume is not even based on presidential election data.
After some thought it seemed to me that what I am interested in is the probability that a state will swing. With ‘P’ indicating probability that seemed like it should be…
P(swing) = P(vote moves right way)*P(vote moves enough to swing)
With…
P(vote moves the right way) = # of times voted for that party/times voted
P(vote moves enough) = # of times vote moved enough/times voted
There again seems to be a question of how is sample size worked to give meaningful statistics. At the state or individual county level there might not be enough data. Possibly it can be summed across counties somehow for the state? But then don’t number of votes have to also be accounted for, something like Effect size. How much does each county contribute to the possibility that the state flips?
To begin with I will consider something a little simpler, party margin of victory as a percentage as was shown previously for the six states. This relates to the probabilities above as one measure of the distance that needs to be moved for a state to flip. The less distance needed to move the more probable a flip. It can be seen standalone as a measure of how competitive the state is. The smaller the percentage, the closer to dead even, the more it is competitive. Possibly, a rough metric of how likely a swing state it is in itself.
## State Percentage
## 22 MI -0.24
## 38 PA -0.75
## 48 WI -0.82
## 9 FL -1.24
## 27 NC -3.81
## 10 GA -5.32
## 35 OH -8.54
## 43 TX -9.43
## 12 IA -10.13
## 40 SC -14.92Four of our six critical states head the list. The other two make the Top 10 list. Not too bad a metric itself, and very simple. All negative meaning Republican now, potentially swinging to Democrat.
Again, the data probably isn’t adequate to realistically compute probabilities at this level but it should be reasonably easy to do. Since the Democrats won two and the Republicans two of the last four elections, the probabilites that a state’s votes go in the correct direction should be about 50-50. For the ‘competitive’ states this should, hopefully, be close to correct. Then, we will use the actual state level vote swings.
## State Probs
## 1 UT 44.444
## 2 FL 33.333
## 3 MI 33.333
## 4 PA 33.333
## 5 WI 33.333
## 6 AZ 22.222
## 7 GA 22.222
## 8 IA 22.222
## 9 IN 11.111
## 10 NC 11.111Which doesn’t look much like our list of six critical swing states, I would still believe this is the inadequacy of the data at this level.
But more sophisticated simulation approaches could possibly compenstate for the lack of data. A Monte Carlo simulation can be attempted. For this one I will randomly choose county votes from the four elections. Usual I think is to select random numbers from assumed or known probabilitly distributions.
## Monte Carlo## State flipped.dg flipped.gd P EV
## 1 MI 0 50 1.00 16
## 2 PA 0 50 1.00 20
## 3 WI 0 50 1.00 10
## 4 IA 0 46 0.92 6
## 5 FL 0 16 0.32 29
## 6 OH 0 12 0.24 18
## 7 NV 9 0 0.18 6
## 8 VA 8 0 0.16 13
## 9 CO 1 0 0.02 9
## 10 AL 0 0 0.00 9Where for a simulated 2020 run…
flipped.dg is the number of times the state flipped democrat to gop
flipped.gd is the number of times the state flipped gop to democrat
P is the states probability of flipping in 2020.
EV is the state’s electoral votes.
Monte Carlo appears to do a better job of highlighting what seem to be probable swing states. Notice where our six critical states turn up. The likely swing states don’t change much.
The certain probabilities it gets do not seem realistic. I sampled using actual county results from the last four elections. Two of those were the Obama wins in ’08 and ’12, these results probably reflect his strength in these states for those elections vs. the GOP ones in ’04 and ’16. This could indicate some Democrat bias in this sampling approach - at least for those states. Given the ’16 result for Iowa, in my opinion, I would say it’s probability of flipping back to Democrat shouldn’t be so high either. Normally, a Monte Carlo simulation will involve thousands of iterations. I didn’t do that, fifty took some time and seemed sufficient for illustrative purposes.
Given the limitations of the simulation I think it does roughly convey which are the likely swing states. Also, I think it gives a correct indication that flipping overall, for most states, isn’t that likely. Less than 10 states show any chance of flipping at all.
Given that, although I had thought of doing additional simulations I think for now this will do.
Throughout this I have considered what were the Swing states for the 2016 election. The impression maybe being that they are the only possible Swing states. Some backtesting can be done on this.
Consider Swing states for the prior two elections…
## [1] "2004-2008"##
## 0 1
## 41 9## State EV diff.04 diff.08
## 5 CO 9 -99523 215004
## 9 FL 29 -380978 236148
## 12 IA 6 -10059 146561
## 15 IN 11 -510427 28391
## 27 NC 15 -435317 14177
## 32 NM 5 -5988 125590
## 33 NV 6 -21500 120909
## 35 OH 18 -118775 262224
## 45 VA 13 -262217 234527## [1] "2008 Competitiveness..."## State Percentage
## 24 MO -0.14
## 26 MT -2.47
## 10 GA -5.26
## 41 SD -8.59
## 3 AZ -8.63
## 28 ND -8.87
## 40 SC -9.09
## 43 TX -11.87
## 25 MS -13.28
## 49 WV -13.35## [1] "2008-2012"##
## -1 0
## 2 48## State EV diff.08 diff.12
## 15 IN 11 28391 -267656
## 27 NC 15 14177 -92004## [1] "2012 Competitiveness..."## State Percentage
## 27 NC -2.07
## 10 GA -7.91
## 3 AZ -9.23
## 24 MO -9.56
## 15 IN -10.40
## 40 SC -10.62
## 25 MS -11.60
## 26 MT -14.07
## 43 TX -16.02
## 18 LA -17.49Neither of these appears to be as competitive as the 2016 election. There was some consideration of this on the NOVA broadcast “Prediction By The Numbers” that I just saw. 2008 seemed to be an election where the voters generally shifted Democrat as they generally seemed to go Republican in 2016. Maybe a little farther than they normally would as the two swings in 2012 seem to be regression to the mean sort with Indiana and North Carolina returning to more normal voting patterns. Although, given it’s placement on both the 2012 and 2016 competitive lists, as well as two actual flips, North Carolina does seem a genuine swing state.
FL, IA and OH show as flipping in the 2008 election. I still think possibly this was an unusual swing for Iowa and the 2016 election may of been regression to the mean there. MI, PA and WI do not appear, looking at past results they all would probably have been considered Democrat base states having all been Democrat since 1988. Wisconsin even then. Iowa as well actually. It appears to be more Democrat base than I would of thought. Possibly it will turn back Democrat in a regression to the mean as well, despite having had almost half of it’s counties flip.
The Trump presidency has been somewhat unique. It could be hard to say what will follow it. A big swing to Democrat election seems possible right now.