City elector, country elector

The Electoral College, let’s stipulate, has no place in a modern election. What made sense for a set of states who considered themselves mostly sovereign in the 1780s, at a time when protecting the power of people who owned other humans for profit was a major feature of political negotiation, certainly does not make sense today. Even leaving aside the weak federation and the strong slavocracy that prevailed a quarter millennium ago, the people who framed the U.S. Constitution could hardly have imagined the political geography of a hemispheric empire, with 50 states of dramatically unequal size and urbanity. The time to reform or abolish the Electoral College (and the Senate) has long since arrived.

But let’s also note that there is a shred of a principle worth holding onto in the territorial apportionment of electoral power. The right wing’s objection that the voting power of cities will run roughshod over the interests of the heartland is almost entirely a cocktail of racist delusion and pure political cynicism. (After all, what we have is the exact opposite—a system where the voting power of the cities, along with a legitimate democratic majority, is itself run over roughshod.) Yet in a democratic system which retains protection for minorities, it is worth considering how and why we might want to reserve votes for specifically-defined geographic blocs.

This led me to think: what if we gave credit to the principle that the Electoral College should in fact protect rural voters by not simply tagging them on as a permanently-outvoted rump in a popular election, but designed a system which didn’t over-represent rural areas in respect to their actual numbers, as the current Electoral College does? What if, instead, we divided the population into equal-size sets, and arranged these sets so that the most urban voters voted together in a single bloc, followed by a bloc of the next-most-urban voters, all the way down to the most rural voters?

The “districts” in such a system would not be territorially-continuous areas, but instead segments of the population along an urban-rural spectrum. Voters in Manhattan and the Chicago Loop would share a “district” and an electoral vote, and so would voters in eastern Montana and northern Maine. The most rural slice of the population would thus be guaranteed an electoral vote—but so would the same number of voters in the most urban slice of the population.

I was curious what this would look like, so I ran some models based on the 2016 election. In these models, I’ve taken raw population density as a proxy for urbanity, which is a highly imperfect measure. (Los Angeles County, for instance, is less dense than Orange County, because the former includes a lot of uninhabited land area in mountains and desert—even though it is unquestionably more “urban” than the latter.) I’ve also run the analysis with counties as the geographic unit, which is also problematic: counties in the United States vary considerably in their makeup. But counties are one of the only units where we have good election data, and density is a crude but nevertheless usable index of urbanity.

The analysis is available in this Binder notebook for replicability and exploring. Unsurprisingly, the voting patterns of our density-sliced districts exhibit the pattern that has become a familiar part of political geography of the U.S.: a linear trend from Democrat-voting cities to Republican-voting rural areas. Here’s the result with 9 “districts” of equal size. In these tables, the districts are defined by their most-dense county (their upper bound on the density spectrum) and their least-dense county (their lower bound). Thus, in this example, District 1 includes all the counties whose densities range from New York County (Manhattan) to DuPage County (the inner suburbs of Chicago).

District Total Population Densest County Least Dense County Dem % Rep % Winner
1 37,299,101 New York County, New York DuPage County, Illinois 69.36 26.7 D
2 34,210,307 DeKalb County, Georgia Clayton County, Georgia 63.11 32.31 D
3 36,929,333 Middlesex County, Massachusetts Hillsborough County, Florida 59.26 35.89 D
4 35,956,241 Monmouth County, New Jersey Vanderburgh County, Indiana 51.63 43.23 D
5 32,504,452 Hampden County, Massachusetts Kalamazoo County, Michigan 46.64 48.25 R
6 38,893,609 Maricopa County, Arizona Saratoga County, New York 42.79 51.48 R
7 35,685,254 Hinds County, Mississippi Moore County, North Carolina 37.47 57.77 R
8 35,677,509 Lapeer County, Michigan Allen County, Kentucky 34.34 60.65 R
9 35,747,224 Bonneville County, Idaho Yukon-Koyukuk Census Area, Alaska 33.46 59.97 R


Most interestingly, though, Donald Trump still wins this version of the 2016 election, even though the constituencies are equally apportioned (or roughly so—because the rank order of counties has to be cut at distinct breakpoints, it’s impossible to make perfectly-equally sized districts, especially at the higher end of the spectrum, where each county may account for several million people). Here, he squeaks out a victory by winning just under a 2% margin in the middle-density District 5, which ranges in county densities from the Pioneer Valley of Massachusetts to Kalamazoo, Michigan. Of course, if this really were how electors were apportioned, the campaigns would have been fought differently, so it’s hard to make any assumptions about how voters might have behaved under such a system. But it still shows something interesting and important: that the Republicans have an advantage not only in rural counties, but in just enough counties that are exactly down the middle of the density spectrum for them to hold an advantage.

As is the case with any attempt to divide voters into sets and then aggregate their choices into winner-take-all blocs, the number of districts we draw has some surprising consequences. For instance, if we run the same election with 21 districts rather than 9, Hillary Clinton wins the election by edging out Trump in District 11:


District Total Population Densest County Least Dense County Dem % Rep % Winner
1 14,861,227 New York County, New York Essex County, New Jersey 80.64 16.47 D
2 15,912,247 Cook County, Illinois Winchester city, Virginia 63.17 32.35 D
3 14,800,056 Dallas County, Texas Newport News city, Virginia 61.62 34.28 D
4 16,320,929 Los Angeles County, California Westchester County, New York 64.97 30.27 D
5 15,597,047 Hennepin County, Minnesota Macomb County, Michigan 63.5 31.64 D
6 15,628,239 Norfolk County, Massachusetts Providence County, Rhode Island 58.99 36.55 D
7 15,318,996 Fairfield County, Connecticut Hillsborough County, Florida 57.05 37.61 D
8 15,117,251 Monmouth County, New Jersey Clarke County, Georgia 52.52 42.3 D
9 15,533,076 Bucks County, Pennsylvania Washington County, Oregon 50.7 44.46 D
10 15,060,023 Hamilton County, Indiana Genesee County, Michigan 48.68 45.9 D
11 15,417,696 Greenville County, South Carolina Cleveland County, Oklahoma 48.15 46.57 D
12 15,585,913 Sedgwick County, Kansas Leon County, Florida 46.06 49.04 R
13 15,316,510 Putnam County, New York Cumberland County, Maine 41.33 52.86 R
14 15,323,833 Erie County, Pennsylvania Saratoga County, New York 40.9 53.15 R
15 15,186,794 Hinds County, Mississippi Tompkins County, New York 39.57 55.79 R
16 15,404,525 Clackamas County, Oregon Boone County, Indiana 36.02 59.16 R
17 15,316,483 Belknap County, New Hampshire Barry County, Michigan 35.38 59.55 R
18 15,234,397 Fulton County, New York Putnam County, Indiana 35.73 59.36 R
19 15,336,889 Santa Fe County, New Mexico Preston County, West Virginia 32.32 62.59 R
20 15,320,870 Huntingdon County, Pennsylvania Edwards County, Illinois 32.91 62.26 R
21 15,310,029 Franklin County, Arkansas Yukon-Koyukuk Census Area, Alaska 33.65 58.71 R

But turn the dial back down to 17 districts, and Trump wins again, with 9 electoral votes to Clinton’s 8:


District Total Population Densest County Least Dense County Dem % Rep % Winner
1 20,084,946 New York County, New York Cook County, Illinois 78.83 17.9 D
2 19,239,594 Union County, New Jersey Fredericksburg city, Virginia 60.91 34.62 D
3 18,684,112 Passaic County, New Jersey Colonial Heights city, Virginia 63.99 31.35 D
4 18,614,150 Tarrant County, Texas Radford city, Virginia 63.41 31.88 D
5 18,562,096 Macomb County, Michigan Salt Lake County, Utah 57.21 37.4 D
6 19,226,969 Santa Clara County, California Hartford County, Connecticut 57.15 38.61 D
7 19,387,526 Travis County, Texas Erie County, New York 50.33 44.5 D
8 18,888,412 Galveston County, Texas Chatham County, Georgia 49.77 44.9 D
9 18,836,123 Hamilton County, Tennessee Manatee County, Florida 46.97 47.75 R
10 18,990,427 Marin County, California Douglas County, Colorado 44.67 49.78 R
11 18,936,579 Winnebago County, Wisconsin Hampshire County, Massachusetts 42.36 52.46 R
12 18,869,503 Benton County, Arkansas Sangamon County, Illinois 39.8 54.72 R
13 18,979,897 Windham County, Connecticut Albemarle County, Virginia 35.99 59.3 R
14 18,883,361 Cass County, Missouri Stokes County, North Carolina 36.74 58.04 R
15 18,908,564 Pike County, Pennsylvania Cherokee County, Oklahoma 33.37 61.7 R
16 18,909,925 Coshocton County, Ohio Buena Vista County, Iowa 32.27 62.89 R
17 18,900,846 Breckinridge County, Kentucky Yukon-Koyukuk Census Area, Alaska 33.69 58.97 R


And if we crank all the way up to 27 districts, we see a case where the partisan sort order isn’t perfectly cut down the middle. Here, Trump wins District 13 (by 0.2%), while Clinton edges him out in the just-a-bit-more-rural District 14 (by 0.6%).


District Total Population Densest County Least Dense County Dem % Rep % Winner
1 13,422,482 New York County, New York Richmond County, New York 80.69 16.43 D
2 10,840,874 Baltimore city, Maryland Bergen County, New Jersey 69.68 26.5 D
3 13,035,745 Orange County, California DuPage County, Illinois 59.33 35.73 D
4 8,274,429 DeKalb County, Georgia Newport News city, Virginia 60.68 35.71 D
5 15,352,114 Los Angeles County, California Camden County, New Jersey 64.98 30.12 D
6 10,583,764 Westchester County, New York Clayton County, Georgia 62.7 32.47 D
7 13,012,875 Middlesex County, Massachusetts Suffolk County, New York 60.63 34.87 D
8 10,662,615 Broward County, Florida Salt Lake County, Utah 59.11 34.59 D
9 13,253,843 Santa Clara County, California Hillsborough County, Florida 57.92 37.92 D
10 11,608,274 Monmouth County, New Jersey Fayette County, Kentucky 53.41 41.23 D
11 12,136,236 Collin County, Texas Gloucester County, New Jersey 51.02 44.35 D
12 12,211,731 Knox County, Tennessee Vanderburgh County, Indiana 50.41 44.11 D
13 11,904,288 Hampden County, Massachusetts Clay County, Missouri 47.33 47.52 R
14 11,764,983 Onondaga County, New York Pulaski County, Arkansas 47.73 47.16 D
15 8,835,181 Washtenaw County, Michigan Kalamazoo County, Michigan 44.18 50.77 R
16 15,030,062 Maricopa County, Arizona Washington County, Rhode Island 44.94 49.3 R
17 11,554,247 Lebanon County, Pennsylvania Livingston County, Michigan 41.15 53.84 R
18 12,309,300 Riverside County, California Saratoga County, New York 41.73 51.82 R
19 11,792,515 Hinds County, Mississippi Wilson County, Tennessee 40.33 55.0 R
20 11,828,605 Minnehaha County, South Dakota Nash County, North Carolina 35.53 59.93 R
21 12,064,134 Anchorage Municipality, Alaska Moore County, North Carolina 36.64 58.3 R
22 10,757,729 Lapeer County, Michigan Fauquier County, Virginia 35.46 59.42 R
23 12,877,087 San Bernardino County, California Upson County, Georgia 34.25 60.85 R
24 12,042,693 Lane County, Oregon Allen County, Kentucky 33.47 61.5 R
25 11,948,202 Bonneville County, Idaho Meriwether County, Georgia 32.98 62.07 R
26 11,889,133 Orange County, Vermont Cass County, Iowa 32.3 62.86 R
27 11,909,889 Izard County, Arkansas Yukon-Koyukuk Census Area, Alaska 34.16 57.86 R


What I think is most notable about these results is that the structural disadvantage for Democrats in our real Electoral College isn’t solely limited to the fact that rural states have more electors per voter than urban states. Instead, it’s the fact that in any territorial apportionment of voters, Democrats are at a disadvantage due to the fact that they are “inefficiently” clustered together in cities. And this phenomenon is really one that’s exclusively about the biggest of big cities. In these models, the second and third districts (that is, the next-to-most urban districts) are usually about as Democrat-leaning as the highest-numbered districts (that is, the most rural districts) are Republican-leaning. But the most urban district is much more Democrat-leaning than the most rural district. Thus Democrats have a surplus of voters in these handful of large cities, running up their margins in places which translates into no electoral advantage.

Obviously, this sort of an electoral system is never going to be instituted. But it does suggest something about the geography of partisanship that cuts against the “urban–rural” divide that has become so familiar. If a few hundred thousand Democratic voters were to decamp from the densest cities and move, not to the true rural hinterland, but to the kinds of counties which are right at the middle of U.S. density patterns—to cities like Chattanooga, Tallahassee, and Springfield, MA, all of which are in tipping-point districts in the models above—they would go a long ways towards shifting the geographic terrain of political campaigns in their favor.

Crossing the T in Treason

Let me preface this by saying, unequivocally, that Donald Trump should be impeached. Never mind impeached: publicly spanked, or shot into the sun would be even better. We should all wish for Congress to act swiftly to make this happen, … Continue reading

The tip toe

“Toe of mouse, injected, x30.” Hardwicke’s Science-Gossip: Monthly Magazine of Interchange and Gossip for Students and Lovers of Nature 25 (No. 242, February 1885), p. 25.


Bill Traylor, Untitled (Rabbit), (1940). Smithsonian American Art Museum.

In memory of Donald Hall

Yesterday, I was moving books around in the glass cabinets, and I took out Seasons at Eagle Pond, flipped through it briefly, and replaced it on a different shelf. Later that afternoon I found out that Donald Hall had died … Continue reading