It's time for Bughouse's quadrennial election cycle the-electoral-college-is-terrible post! (definitely made at least a few in 2016/2020). This time I bring a tool and some actual numbers, not just philosophizing on why it's bad.
I remember coming across
this* years ago (and being horrified) and wondering if I could play around with maps myself and more extremely than this.
Well apparently you can!
this is an interesting tool, if you have lots of time on your hands. If you want to have fun you could try to gerrymander hilarious new state lines under which the popular vote winner wins every state. I was less nefarious and just played around with totally fictional and arbitrary state boundaries, only respecting the following rules:
1) DC still a separate district and have 50 other total states (i.e. I wanted my math to produce a 538 EV map)
2) Alaska and Hawaii stay their own things bc not contiguous
3) I thought it would be fun to do a better job of grouping counties up to states by using a geographic commonality like a long stretch of the coast or border or surrounding lake michigan, etc.
4) I wanted to make sure that there was a good range of populations, not aiming for each state to all be ~5-10M people, but at the same time to be less dramatic than currently. No CA behemoths.
Everything else was up for grabs. 4 color map below to show where the states are, this is not the political lean of the fictional state. (DC is still a district with 3 EVs and AK and HI are left as is, see #1 and #2 above, just not included in the screenshot.)
Even a map with somewhat less electoral college distortion because there's no states as large as CA in my map (largest pop is ~20M, about half of CA's) and very few super small states (only 2 states, AK and a state I called "the desert" in northern NV, under 1M, vs 5ish states in reality under 1M) you still end up with plenty of popular vote to electoral vote distortions:
| Year | Electoral Votes(R) | Electoral Votes(D) | Real World Votes (R) | Real World Votes (D) | Shift caused by random map |
|---|
| 2004 | 300 | 238 | 286 | 252 | R+14 EVs |
| 2008 | 204 | 334 | 173 | 365 | R+31 EVs |
| 2012 | 269 | 269 | 206 | 332 | R+63 EVs |
| 2016 | 282 | 256 | 306 | 232 | D+24 EVs |
| 2020 | 243 | 295 | 232 | 306 | R+11 EVs |
This is not remotely to say that the current EV map/real world state boundaries is bad for Rs and good for Ds (that seems rather unlikely given that in the real world there's a good chance for Ds to win the popular vote but lose electoral college every close election under the modern map, but the reverse is close to impossible...) I just may have drawn a map that would have historically been
even more favorable for R voters than the real world map.
Really the more important point to get out of these numbers instead is the following:
1) State lines are ultimately arbitrary and have arbitrary effects on the electoral college and will pretty much always swing the result one way or another by ~1-5 states
2) Unless you're very consciously trying to gerrymander to avoid the result (which I absolutely was not here, nor are the real world state lines in any way "intentional" for modern day elections), you will inevitably have a few states that are much closer than others and these 1-5 states will therefore swing a close election, hence why the counts are normally swinging by ~10-30 electoral votes. For example in my 2020 map the state that's "south texas" is decided by 0.11% for Rs. The state that makes up most of real world CO and WY is a 0.72% D win in 2012 and a 0.3% R win in 2016. The state covering Minnesota+ in 2004 would have been decided by literally 3 votes.
3) The absolute "error" caused by the randomness here is ~29 EVs per election, or ~5% of the total EVs. If we assume absolutely no bias anywhere in the system, truly arbitary state lines drawn with no purpose in mind and that have nothing to do with generally well-distributed populations, this is basically like saying "an election that is a tie in the popular vote would be expected on average to come out 298-240 one way or the other. Half the time it might come out closer than that (i.e. closer to tied popular vote reality), half the time it might come out farther than that (i.e. farther from reality). Really this confidence interval would be shifted rightward, since the map seems to have a +R bias, but I'm oversimplifying here because this is just one random map. A real analysis of this, if we analogize to state level district gerrymandering, would require generating many thousands of random maps and assessing the impact overall. One map doesn't really prove anything but it's illustrative of possible effects of having individual states voting in an electoral college, in general. That's all I'm after here.
4) Likewise, an election where someone wins by ~<5% of the vote they could still lose a decent portion of the time if the map happens to not be favorable to them (as we already know). On the flip side, elections that are actually pretty close by popular vote can be distorted by an electoral college vote. Obama only beat Romney by 4% in 2012. Under a different random map that could absolutely come out tied. Instead it was almost fair to call it a landslide because Obama happened to win FL by <1% and the only close state he lost at all was NC. He won every other state that was within a 7% margin. A random map it turns out would actually have been less kind to 2012 Obama and the election could have even been tied with an apparently unfavorable map I made up.
or tl;dr
1) swing states are inevitable and arbitrary
2) swing states have an inherent distorting effect on the popular vote
hence:
3) where state lines happen to be arbitrarily drawn has the potential to swing literally any close election within ~5% of the popular vote, give or take.
4) please for the love of god abolish the electoral college, it's an abomination to democracy.
*side note from the 2020 article
Article reads:
"(No, that’s not a typo. While I moved the counties from Washington to Idaho, the apportionment formula is such that North Carolina picks up the extra House seat.)"
To learn more, watch this to learn how the math we use for congressional apportionment, and hence he electoral college, is itself a mess!
therefore an alternative way to look at things is that some of that ~29 EV movement is not due to swing states but due to the randomness of apportionment. nothing about NC changes but it gets 1 more or fewer seat and EV. This could definitely add up to a few EVs per census cycle, and is yet another reason why further extending the problem of apportionment (which is at least to some extent needed for congress!) to a context where it's just not needed. use a popular vote for president and the apportionment issues go away.