Politics and Population Density

There has been considerable attention recently to the increasing alignment between population density in the United States and partisan election voting. Cities, such as Madison and Milwaukee have increasingly voted for Democratic candidates, while small towns and rural areas in Wisconsin have become increasingly supportive of Republicans.

The chart below compares the relationship between the percentage of the vote that went to Tony Evers in Milwaukee County municipalities and the approximate distance between Milwaukee and each suburb. R2, the coefficient of determination, is often described as showing how much of the variability of one factor can be caused by its relationship to another related factor. In other words, in Milwaukee County, about 45% of the variation of the partisan vote reflects the distance from Milwaukee.

Note that in all these cases, I only include votes for the two major-party candidates: Tony Evers and Tim Michels. Thus, one can calculate the Michels vote by subtracting the Evers percentage from 100%.

As can be seen from the plot, Milwaukee ranked second for the percentage of its voters supporting Evers. Shorewood had the highest proportion. This may be consistent with the model. Because Milwaukee covers so much territory, many Milwaukee voters will be further from central Milwaukee than the Shorewood voters.

Evers lost the election in only three Milwaukee County suburbs: Franklin, Hales Corners, and Oak Creek. All three are on the south side of the county. However, along with South Milwaukee, they are the most distant from Milwaukee.

The next graph shows the results of the same analysis when applied to Ozaukee election results. Here, the relationship is even stronger than for Milwaukee County, with an R2 above 70%.

Ozaukee County was not Evers Country. His only comfortable win was in the small portion of Bayside located in Ozaukee County, which is separated from the rest of Ozaukee County by a ravine, meaning one must go through Milwaukee County to access that part of Bayside.

The next closest municipalities, Thiensville and Mequon, were ones that Evers barely won or barely lost. Two others where the vote was very close, Cedarburg and Port Washington, are the kinds of places likely to attract Democrats.

Finally, the next graph applies the same analysis to Waukesha County. While weaker than in Ozaukee County, the same pattern appears, with an R2 of around 40%. Evers won no Waukesha County municipality, but–think Elm Grove–several were close.

Evers did considerably better in the city of Waukesha than would be predicted by using only its distance from Milwaukee. But the city of Waukesha presents a special case, as it is itself something of an urban environment.

Within the Milwaukee metropolitan area, there is a pattern that repeats itself again and again: the further away a community is from Milwaukee it is, the more Republican it votes. By the same token, the closer it is the more Democratic it votes. This pattern repeats throughout the data.

But what accounts for this pattern? Are Democrats attracted to more urban neighborhoods and Republicans to more rural and suburban ones?

Is there some signal that counties give off that attracts people sympathetic to one party while repelling those in the other party? Something that says, “you will be comfortable (or not) here”?

I don’t want to overstate the issue. A healthy minority of Democratic voters live in Ozaukee County, for instance—about 23,000 or 44.5% compared to almost 29,000 Republican voters. So, whatever it is that made the county attractive to more Republican voters stopped well short of making the county uniformly Republican.

While the imbalance is somewhat greater among Waukesha County voters—86,000 Democratic voters to 128,000 Republicans, or 40% to 60%, that is still a large number of Democratic voters. In fact, if counties in the state are ranked by their number of Democratic voters, Waukesha would come in third highest.

That the phenomenon is real is supported by the evidence. The cause, however, is a mystery.

Note on the distance calculations: In calculating the distances between Milwaukee and other towns, villages and cities, I started with a function in Excel which generates data for geographic entities, including Wisconsin’s municipalities. Among the data generated are the latitude (how far north or south the entity is) and longitude (how far east or west of Greenwich England it is). I then took the differences of the latitudes and longitudes, squared the differences, added up the squares, and took the square roots or the differences, which I used as the distances between municipalities.

Sticklers for accuracy may note that this calculation ignores the earth’s curvature. The further away from the equator the closer together the north-south lines are. As a result, a given difference in longitude will shrink as we go north. For simplicity I chose to ignore this effect.