Shifting Tides in the Battleground States
As I sit here only days away from the U.S. presidential election, wondering whether I will be drinking to celebrate or drinking to drown my sorrows (the drinking part is a given), I can’t help but wonder if elections are exercises in true indeterminism or simple predestination. In other words, does anything we do amount to a hill of beans? I suppose it depends on how one defines political momentum.
The Myth of Political Momentum
Political momentum is one of those handy terms overused and poorly defined by the media. Typically, it refers to short-term momentum, actions happening in the present tense, measured through political polls. Candidates are tracked day by day, minute by minute, with the hope of creating predictive models of success based on snapshots in time.
The challenge with measuring short-term momentum is that it is based on opinions not actions, and opinions are notoriously difficult to quantify. Courtney Kennedy, director of survey research at the Pew Research Center, identified many of the weaknesses of political polls: different polling organizations conduct surveys in different ways, many pollsters have no credentials, the margin of error is often double what is reported, surveys often fail to adjust for education level, and estimates of public opinions of policy are much more trustworthy than tracking the “horse race.” Furthermore, evidence shows that telling the public a candidate is likely to win often reduces voter turnout. In my previous life as a market research analyst, I often quipped, “let me know what results you want, and I’ll design the questionnaire for you.”
More importantly, short-term political momentum is tactical rather than strategic. As its name implies, it measures change in the moment, not changes over the long-term. Long-term analysis, on the other hand, is strategic. It assumes voter populations change slowly, and it therefore provides a way to guess the future by studying the past.
Are All Politics Local?
Measuring long-terms trends is all well and good, but to quote former Speaker of the House Tip O’Neill, “all politics is local.” Or is it? Andrew Gelman, writing for FiveThirtyEight nearly 10 years ago, refuted O’Neill’s adage, saying, “elections in the United States have become increasingly nationalized in recent decades…the decline is gradual, but we’re clearly at a lower level of variation now than we were 30 years ago…the red-blue map is much more stable from election to election than it used to be.”
If geographies don’t shift dramatically over time, do they shift dramatically in space? In other words, do counties typically vote like their neighbors?
Take Vigo County, Indiana, for example. Vigo County (county seat Terre Haute) has been considered a bellwether for presidential elections since 1888. Since then, its voters have chosen the winning candidate in all but two elections. Not surprisingly, its voters split pretty evenly across party lines. In the last five presidential elections combined, 49% of all votes were cast for Republicans and 48% for Democrats. Compare Vigo to next door Clay County, where 66% of the combined presidential votes went for Republicans, and 32% for Democrats. Two neighboring counties, two political populations. One might say, “all politics is painted nationally but written locally.”
Political Slope Analysis
I decided to measure this variation over time and space using Political Slope Analysis (my own term) to create a geographic map of party trends county by county. By measuring the mean and slope of the percentage of Democratic and Republican votes, I was able to get a birds-eye view of both general political attitudes and long-term shifts.
In my analysis, I measured the following:
- Percentage Points between Democratic and Republican votes for each election between 2000 and 2016
- Average percentage points for all the elections between 2000 and 2016
- Slope of the linear regression line of both Democratic and Republican votes across the 16-year period, (normalized by dividing by 0.0625, the maximum possible slope value)
I collected raw data from the MIT Election Data and Science Lab. Among their many datasets are national election results for all 3,141 U.S. counties.
As an example, consider the aforementioned Vigo County, below. Although the percentage of Democratic and Republican votes swing wildly election by election, the overall trend shows a slight decrease in Democratic voters (-5.6%) and a slight increase in Republican voters (2.6%). Therefore, a presidential candidate campaigning in Vigo County would be well to consider the long-term political trends as well as any short-term political issues.
POLITICO identified eight states as critical battlegrounds for the 2020 presidential election: Arizona, Florida, Georgia, Michigan, Minnesota, North Carolina, Pennsylvania, and Wisconsin. These states were selected based on polling, demography, past and recent election history, voter registration, interviews with party officials, and polls.
These swing states serve as perfect templates to test Political Slope Analysis. For each state, I have mapped out county population, average percentage points (in favor of Democrats or Republicans), and slope of both party voters. All maps were created using the online tool mapchart.net.
What becomes obvious in the resulting patchwork of blue and red is the complexity of the battlefield. Some party strongholds are stable, some are becoming more partisan, and some are shifting in the opposite direction. Note that these maps are not meant to highlight the full intricacies of the election process. Rather they are meant to uncover true political momentum: unsexy, unexciting political mobility in slow time.
Arizona (Full-Sized Map)
- Average percentage points R+7.6%
- Democratic Slope 0.0%
- Republican Slope -2.8%
On a statewide level, Arizona leans Republican. However, Democratic votes have remained consistent over the long term, while Republican votes have trended very slightly downward.
High population counties are not necessarily Democratic strongholds. Maricopa County (Phoenix) leans slightly Republican and Pima County (Tucson) leans slightly Democratic. Republican Maricopa County is trending slightly more Democratic, and deep-blue Apache County is trending slightly more Republican based on the downturn in Democratic votes.
Florida (Full-Sized Map)
- Average percentage points 0.5% (R)
- Democratic Slope 0.4%
- Republican Slope -1.0%
Florida is the swingiest of all the swing states, with only a 0.5% percentage point advantage to Republicans. Despite slight shifts over time, party affiliation remains pretty steady.
Of the highest populated areas, Palm Beach, Broward, Miami Beach, Volusia, Pinellas, and Hillsborough Counties lean Democratic, whereas Lee, Orange, and Duvall Counties lean Republican. Osceola, with its slightly Democratic majority, appears to be trending even more Democratic.
Georgia (Full-Sized Map)
- Average percentage points 9.3% (R)
- Democratic slope 3.8%
- Republican slope -5.0%
Georgia has traditionally voted Republican in presidential elections, but the numbers show a slow shift towards more Democratic and fewer Republican votes. The suburban counties surrounding Atlanta have tipped slightly Republican in the past, but the entire greater Atlanta area, including the surrounding suburbs, appears to be trending Democratic. Deep red Columbia County (Augusta) in the northeastern part of the state appears to be trending slightly blue as well.
Michigan (Full-Sized Map)
- Average percentage points 6.9% (D)
- Democratic slope -2.0%
- Republican slope -0.2%
Michigan has leaned Democratic in the past, but Democratic votes have dropped very slightly. Four of the five elections have gone to Democratic candidates, but 2016 went to Donald Trump by 0.2 percentage points. Areas to watch include red Ottawa and Kent Counties in western Michigan, which are trending slightly blue, and blue Marquette County in the Upper Peninsula, which is trending slightly red.
Minnesota (Full-Sized Map)
- Average percentage points 5.1% (D)
- Democratic slope -0.5%
- Republican slope -1.5%
On average, Minnesota leans more Democratic than Republican, and Democratic counties tend to be pretty stable. Republican advantage appears to be trending downward slightly.
Areas of possible change are Lake, Carlton, and Olmstead Counties. Lake and Carlton in the northwest typically lean Republican but appear to be shifting more Democratic. Olmsted County (Rochester) typically leans more Democratic, but appears to be tipping more Republican.
North Carolina (Full-Sized Map)
- Average percentage points 6.1% (R)
- Democratic slope 4.3%
- Republican slope -7.2%
Although North Carolina typically goes Republican, an upward Democratic shift and downward Republican shift might trend it Democratic over time. Most of the trends appear to fall along party lines (blues getting bluer, reds getting redder). In particular, Guilford (Greensboro) and Wake (Raleigh) Counties, both populous Democratic-leaning counties, are trending even more blue.
Pennsylvania (Full-Sized Map)
- Average percentage points 4.3% (D)
- Democratic slope -1.8%
- Republican slope 1%
Pennsylvania leans Democratic, but the past five elections have shown a small downward shift for Democrats and a small upward shift for Republicans. Most of the map trends along party lines, but a couple of areas are worth watching. Right-leaning Lancaster and Cumberland Counties appear to be trending more Democratic, and left-leaning Luzerne and Erie Counties appear to be trending more Republican.
Wisconsin (Full-Sized Map)
- Average percentage points 4.1% (D)
- Democratic slope 0.1%
- Republican slope -1.7%
Overall, left-leaning Wisconsin has remained steady over time. However, many of the left-leaning counties in the western part of the state are trending more Republican. Whether this shift will be offset by deep blue Dane and Milwaukee Counties is yet to be seen.
Political Slope Analysis provides a methodology for mapping out long-term trends in the political landscape. By comparing population maps with voter statistics, campaigns can direct resources to high-profile areas more likely to flip political persuasion. It’s a long game to be sure, but the best politics always are.