It occurred to me while looking at simulations of presidential elections based on past results PvEC [1], that for the situation that was mainly of interest to me, that being when the election was won in the Electoral College while losing on popular vote. That in this situation, votes had to be wasted in a way similar to Gerrymandering. Clearly if you win with less votes it’s the same consequence as Gerrymandering is supposed to provide. It could be interesting to look at gerrymandering in the context of presidential elections. I figured it would be fairly straightforward and easy. I had also previously looked at Gerrymandering, again if curious Gerrymandering [2].
For Gerrymandering a measure of this that is based on wasted votes is the ‘efficiency gap’. The original paper [3]. This was somewhat based on courts saying they had no good way to know if there was Gerrymandering. It gives a calculation you can do to get a number to rate the extent of Gerrymandering.
It had been some time since I looked at this and after a quick look back at what I had my initial assumption that I could directly apply it to presidential elections doesn’t appear to hold. There are assumptions made for the normal usage of the ‘efficiency gap’ that don’t apply in presidential elections. One is SMD - single member district, where each district has a single representative. Not true for the Electoral College. Every state has more than one representative. Another assumption is that each district is equal population. Actually, this is a legal requirement for state districts. But again, definitely not a valid assumption for states.
So, pretty much, use of the efficiency gap is out. It seems to me that a way to still base a similar measure would be to calculate wasted votes per electoral college representative. Where party A votes are greater than party B’s
(party A votes - party B votes) / number of state electoral college representatives
A - B is wasted votes and then the division gives you wasted votes per representative.
## Year V_Diff V_Ratio V_Distance GPE_d GPE_r Ratio Distance
## 1 1976 1683247 1.0429963 0.04299632 22374188 17188635 1.30168503 0.3016850
## 2 1980 -8416921 0.8082613 -0.19173872 3058604 38601397 0.07923557 -0.9207644
## 3 1984 -17677927 0.6753668 -0.32463324 1103842 50607118 0.02181199 -0.9781880
## 4 1988 -7073176 0.8552985 -0.14470149 9371246 37443416 0.25027754 -0.7497225
## 5 1992 5874946 1.1504372 0.15043718 31596464 11687821 2.70336647 1.7033665
## 6 1996 8080983 1.2055112 0.20551121 34499785 11416254 3.02198833 2.0219883
## 7 2000 543915 1.0107800 0.01077999 27586952 25713749 1.07284830 0.0728483
## 8 2004 -3000176 0.9516321 -0.04836787 30600143 32838423 0.93183960 -0.0681604
## 9 2008 9550193 1.1593071 0.15930709 50661078 19264584 2.62975206 1.6297521
## 10 2012 4982290 1.0817660 0.08176602 45048245 23241545 1.93826377 0.9382638
## 11 2016 2868691 1.0455457 0.04554575 31599230 38228001 0.82659906 -0.1734009
## 12 2020 7052762 1.0950300 0.09503002 50260890 33736634 1.48980157 0.4898016
V_Diff is Democrat votes - Republican votes.
V_Ratio is Democrat/Republican votes.
V_Distance Is votes ratio - 1.
GPE_d is gap per electoral Democrats
GPE_r is gap per electoral Republicans.
Ratio is GPE_d/GPE_r Distance is the
ratio - 1
Note that there is a double confirmation here. The 2000 and 2016 elections have the lowest vote differences. 2000 by far the closest. You can understand how it could be decided by a hanging chad or Reagan appointed Supreme Court. Actually 1976 was somewhat closer than the 2016 election but with Voting Age Population (VAP) that far back it probably would be considered less relative to the number of actual eligible voters. More on VAP in a following note on data.
The second confirmation is the Gap Per Elector (gpe) calculation. Yet again they are less for the years concerned, 2000+2016. Intuitively I guess this should of made sense. If the win is big there are more wasted votes than in close elections. Somewhat interesting maybe that George W Bush managed the two closest results. With Trump-Clinton third. This does shoot down the idea that there is some wasted vote gerrymandering thing going on. The result might be similar but the cause is clearly not.
I decided to change what I was looking at from considering something Gerrymandering-like to see if any other repeatable characteristics could be found in this sort of election scenario. Pretty much continuing what I had been doing in the Popular vs. Electoral College document.
A note on Turnout Data.
In this case swing states are determined by actual results rather than by simulations as when I had looked at this before. So they are real historical, not made up imaginary numbers. The criteria I used was the difference between Democrat and Republican states divided by the sum of votes cast. I then order them by this ‘closeness’ and select the top 15.
Considering turnout for the years with Electoral College victories with less popular votes.
The overall Democrat turnout average 28.117%
The overall Republican turnout average 27.849%
2000
Democrat turnout 25.121%
Republican turnout 27.301%
2016
Democrat turnout 27.191%
Republican turnout 29.395%
2020 for contrast
2020
Democrat turnout 31.362%
Republican turnout 31.468%
How far the year’s turnout is from the mean turnout across all years
Average distance Dem 2000 -1.742
Average distance Rep 2000 -2.209
Average distance Dem 2016 -0.485
Average distance Rep 2016 0.085
Average distance Dem 2020 1.878
Average distance Rep 2020 2.5
## State D Dist D Turn R Dist R Turn Winner EC
## 10 FL -3.677 23.9% -2.737 23.9% R 25
## 33 NM -3.745 22.7% -0.882 22.7% D 5
## 49 WI -2.624 32.4% 0.357 32.3% D 11
## 13 IA -3.487 30% -1.647 29.8% D 7
## 38 OR -2.998 29.7% 1.048 29.4% D 7
## 31 NH -4.042 30% -0.841 30.9% R 4
## 24 MN -2.826 34.2% -0.497 32.4% D 10
## 25 MO -2.072 27.2% -1.993 29.1% R 11
## 36 OH -4.309 26.7% -1.683 28.7% R 21
## 34 NV -4.568 19.3% -1.285 20.7% R 4
## 43 TN -0.524 23.1% -2.765 25% R 11
## 39 PA -3.085 27.3% -2.338 25.1% D 23
## 23 MI -2.931 30.2% -1.519 27.2% D 18
## 22 ME -2.597 35.1% 0.531 31.5% D 4
## 3 AR -1.351 21.7% -2.241 24.3% R 6
Summary
Democrat winners 85
Democrat distance -44.8
Republican winners 82
Republican distance -18.5
——————————–
Average Turnout % Difference - 2000
1-5: 0.131
6-10: 1.591
11-15: 2.68
## State D Dist D Turn R Dist R Turn Winner EC
## 23 MI -2.216 30.9% 2.366 31.1% R 16
## 31 NH 0.601 34.7% 2.716 34.4% D 4
## 39 PA -0.140 30.3% 3.317 30.7% R 20
## 49 WI -2.219 32.8% 1.452 33.4% R 10
## 10 FL 0.523 28.1% 2.163 28.8% R 29
## 24 MN -1.705 35.3% 1.182 34.1% D 10
## 34 NV 1.514 25.3% 2.033 24.1% D 6
## 22 ME -1.918 35.8% 2.653 33.6% D 4
## 4 AZ 1.056 23.5% 1.419 25.4% R 11
## 28 NC 2.168 29% 2.805 31.2% R 15
## 11 GA 1.221 24.9% 1.528 27.7% R 16
## 6 CO 3.595 34% 0.767 30.6% D 9
## 46 VA 3.836 32.1% 0.462 28.7% D 13
## 36 OH -3.104 27.9% 2.713 33.1% R 18
## 33 NM 0.975 27.4% -0.802 22.8% D 5
Summary
Democrat winners 51
Democrat distance 4.185
Republican winners 135
Republican distance 26.77
——————————–
Average Turnout % Difference - 2016
1-5: 0.424
6-10: 1.758
11-15: 3.916
Taking the last election as a more typical election year to contrast
with these.
## State D Dist D Turn R Dist R Turn Winner EC
## 11 GA 6.643 30.3% 4.006020 30.2% D 16
## 4 AZ 6.821 29.3% 5.153084 29.1% D 11
## 49 WI 1.131 36.2% 3.802201 35.7% D 10
## 39 PA 3.939 34.4% 6.138101 33.5% D 20
## 28 NC 5.910 32.7% 5.162468 33.6% R 15
## 34 NV 5.525 29.4% 5.921058 27.9% D 6
## 23 MI 3.052 36.2% 5.498248 34.2% D 16
## 10 FL 2.906 30.5% 5.980624 32.6% R 29
## 44 TX 4.559 24.2% 3.092049 27.1% R 38
## 24 MN 3.140 40.1% 1.737189 34.7% D 10
## 31 NH 4.721 38.8% 1.695988 33.4% D 4
## 36 OH -1.243 29.8% 4.651024 35% R 18
## 13 IA -1.733 31.8% 6.099737 37.5% R 6
## 22 ME 2.839 40.6% 2.695263 33.6% D 4
## 46 VA 8.285 36.6% 1.512680 29.7% D 13
Summary
Democrat winners 110
Democrat distance 56.5
Republican winners 106
Republican distance 63.15
——————————–
Average Turnout % Difference - 2020
1-5: 0.499
6-10: 2.776
11-15: 6.052
Not entirely typical as 2020 was the highest turnout year according to
my web page source. With a nationwide average of 61.4%, the next highest
year was 2008 with 56.9%. I have heard mention that with the move of
Harris to the top of the ticket there is Democratic enthusiasm like
there was in 2008. How does Democratic turnout compare for those two
years?
2008 Obama turnout 30.8%.
2020 Biden turnout 31.4%.
As already noted 2020 was a record turnout year. For whatever motivated voters, Biden’s turnout actually exceeded Obama’s. Harris of course was also on the 2020 ticket, and sometimes the talk is that she could put together a ‘coalition’ similar to Obama’s.
Note that in 2020 the Democrats were much closer in turnout than in 2000, 2016. This allowed them to win the swing states in this election. Notice that all the northern “blue wall” states fall into the swing category and were swept in 2020 by Biden - Pennsylvania, Michigan, Wisconsin and Minnesota. However, they are still swing states, not necessarily what you would associate with a ‘wall’ for upcoming elections.
There was a fair amount of talk about some of these 2020 swing states post-election. Georgia with Orders from the DA [4], Arizona of the Cyber Ninjas [6], Notice in these two states the Democrats had their best turnout gains. Also, Wisconsin of the election recount [7]. Pennsylvania had the best Republican turnout gain but they did not succeed in winning the state.
Also somewhat discussed, and it does turn up in the 2020 top 10, is the possiblity that Texas might be becoming a swing state. A bit the other way I’m originally from Minnesota which I always thought of as solid blue. Now though that appears just below Texas as a swing state. I’m not sure how at risk it is but Harris did just pick Minnesota Governor Tim Walz as her VP running mate.
So, possibly not the deepest insight but it seems swing states and turnout are important. Turnout in swing states probably doubly [8]so. Given this I find it strange that a candidate would say they don’t need votes [9]
Again, these elections happen once every four years, it is difficult to call anything ‘typical’. In considering this I thought it might be useful to consider the swing state turnout differences for all the elections from 2000 on.
## Year 1_5 6_10 11_15
## tdiff 2000 0.1311604 1.590966 2.680234
## X 2004 0.6980677 2.082592 3.697898
## X.1 2008 10.7027048 16.885724 11.887326
## X.2 2012 9.4495818 14.396069 10.566800
## X.3 2016 0.4241536 1.757940 3.915853
## X.4 2020 0.4987496 2.775959 6.052002
Given the first three elections considered (2000,2016,2020) it seemed possible that the 2020 results meant there was a polarization of the electorate, possibly trending. Looking at the above what do you think it might say about ‘polarization’? I was planning to do interpretation elsewhere at this time.
Elections can have gerrymandering-like results. Not due to imbalances in vote gaps but to being very close and the mix of turnout, swing states and maybe luck.
Everything I’ve done for turnout is based on Voter Turnout Ranking of States [10]. This has apparently been done by the state of New Hampshire. It was the only place I found historical turnout numbers by state. A few places seemed to have state turnout for the 2020 election but nothing prior to that. With the percentages given and having the Democrat and Republican total votes it is possible to determine what the state voting age population should be. The following shows that by election year.
This indicates that voting age population (VAP) is increasing by election with 2016 for some reason being an outlier where the population decreased. I can think of no reason this should be true, so an outlier. I don’t know how these populations were determined. Probably from census data but that, of course, is only done every 10 years. Maybe interpolating intermediate results somehow. With the VAP determined then the party turnout can be determined.
Taking the mean across the years a relative distance from that mean can be determined. The strictly increasing value of the VAP is problematic for this. I tried a little bit to normalize the years but the results didn’t look correct so now I don’t adjust. For very early election years like 2000 this means the turnout would be less more than it should be and for 2020 more. This should be true for both parties though, so I think using this for relative comparisons of the parties should have some validity.