What factors cause areas to vote in a certain manner? Inherently we know some of them. But we cannot completely project elections based on one factor alone. After all, the state with the largest Black population is also one of the most stubbornly Republican, while the wealthiest state votes Democratic in most elections. While a comprehensive projection system is impossible (and unnecessary, since we already have actual results from many past elections), I thought it would be interesting to try to reverse-engineer the results in order to find out what areas are not voting the way they “should.” By “should,” I’m not trying to engage in “What’s the Matter with Kansas?” style analysis. That type of analysis is simplistic. No matter how much one wants or does not want American politics to be a class struggle, the reality is that it’s not. Looking at American politics through the lens of class and class alone misses quite a bit, arguably the majority, or what goes on here. Instead, the reason I am undertaking this exercise is to attempt to figure out where it would be easiest for the respective political parties to grow. In other words, where are there many voters who for the most part should be voting Republican, based on key characteristics (including but not limited to income), but actually vote Democratic? After all, that’s where the Republican Party should be investing (and vice versa).
In order to get data for this diary, I put demographic traits from every county with a population of 150,000 or greater into a table, and then roughly weighted them so that Los Angeles County does not count the same as Ventura County even though both meet the population threshold. For rural areas, I randomly selected a county from each Census PUMA (Primary Use Microdata Area, but that’s not really important) so that the number of low population rural counties did not skew my analysis. All in all, including duplicates (for example, Salt Lake County is included five times due to its large population), I have 1735 units of analysis. Alaska is excluded, as they don’t really do the whole “county” thing.
I began with a basic statistic that impacts politics everywhere: income (specifically median household income, adjusted for cost of living). I then ran a simple regression, looking for the largest outliers.
Here is the chart; as you can see there is a general but very weak trend, with a correlation value of only 0.08. Which, if you are at all knowledgeable about US politics, should not be surprising.
When one looks at the list of places with the largest error from the projection, it’s immediately clear there is more work to be done. The Top 10 “Too Democratic” Outliers:
1. Prince George’s County, Maryland
2. Washington, DC
3. Montgomery County, Maryland
4. Howard County, Maryland
5. San Francisco, CA
6. Maui County, HI
7. Arlington County, VA
8. Marin County, CA
9. Alameda County, CA
10. Santa Clara County, CA
And the Top 10 “Too Republican”:
1. Cimarron County, OK
2. Winston County, AL
3. Casey County, KY
4. Rockcastle County, KY
5. Knox County, KY
6. Real County, TX
7. San Saba County, TX
8. Johnson County, KY
9. Gilmer County, GA
10. Erath County, TX
At this point, the list is really just a list of places with lots of wealthy liberals and places with lots of poor conservatives, and there is more to politics than income. So let’s add another variable, one that we know affects politics in a huge way: race and ethnicity.
The correlation has grown much stronger, up to 0.30. That's a bit better, but we’re ignoring religion, which is another huge factor driving voting patterns. With the list once again adjusted (based on the New York Times’ map of the “religious right” which unfortunately excludes religious Catholics, with projections adjusted up or down based on the number of religious people relative to the national average), here’s what we’ve got.
There are a couple clear problems (The Bronx can't give Obama 120%, nor can Utah County give Romney 114%), but overall the trend is strong. A 0.60 correlation is nothing to sneeze at. That is not to say that my findings here are anything special. What's interesting is what comes next.
First, however, here are the largest outliers.
Top 10 “Too Democratic”:
1. Chittenden County, VT
2. Boulder County, CO
3. Dane County, WI
4. Howard County, MD
5. Multnomah County, OR
6. Marin County, CA
7. Nicollet County, MN
8. Haywood County, NC
9. Windham County, VT
10. Ballard County, KY
Top 10 “Too Republican”:
1. St Martin Parish, LA
2. Karnes County, TX
3. St Landry Parish, LA
4. Claiborne Parish, LA
5. Brazos County, TX
6. Washington County, CO
7. Morton County, KS
8. Elk County, KS
9. Lee County, SC
10. Yuma County, AZ
This list looks a bit more realistic on both sides. The “Too Republican” counties are now more than just the South, and the “Too Democratic” counties are not quite the same as the list of “counties with the most white liberals.” In case you were wondering, the leaders on each end are a whopping 28 points (Chittenden) and an even larger 34 points (St. Martin) away from their expected vote percentages. However, most of the country does vote somewhat similarly to how we’d expect given only income, cost of living, race, and religion.
There are a few caveats to this data, before I begin to discuss the results and how they relate to the premise of this post.
1. Mormonism: For some reason, all Mormon-heavy counties were listed as far more Democratic than they should be, probably due to the overwhelmingly observant religious nature of these areas, which dwarfs anything in even the Deep South. Thus, I ignored these counties. Hopefully that isn’t too much of a problem, since I doubt Democrats are going to be competing for the religious Mormon vote anytime soon.
2. Conservative Catholicism: Many Catholics are just as religious and just as conservative as their evangelical counterparts, but unfortunately there’s no way to distinguish between them and “cafeteria Catholics” at the county level. This may be a particular concern in Louisiana.
3. Cheap, Wealthy Areas: A few cheap, wealthy areas (such as Williamson County, TN) have projected Republican percentages approaching 100%, which is unrealistic. I did not ignore them, but I’d just like to note this occurrence. The inverse problem occurs in the Bronx, which has such a high cost of living and is so non-White that the expected Democratic vote percentage is over 100%.
4. Assumes 100% Voting: This is a very big one. Poorer Americans vote less, and many aren’t citizens or are ineligible felons. An American electorate with 100% voter turnout would be very different than today’s electorate. This effect is particularly large in heavily Hispanic and/or Asian areas, which have many non-citizens.
With that said, obviously the most fertile places to look for gains by one party or the other (with some exceptions) are the areas voting out of sync with the projections. In some cases these are due to localized factors which need to be taken into account, and in other cases it seems to be more of a fluke. But let’s begin to take a look, starting with places which are too Democratic (and therefore would be fertile areas for Republicans to adopt persuasion-based techniques).
Category 1: Peoples’ Republics
Target Level: 3 out of 10
Population: 13 million
Examples: King County, WA; San Francisco County, CA; Boulder County, CO; Tompkins County, NY; Washtenaw County, MI; Cumberland County, ME; Johnson County, IA; Dane County, WI
Analysis: There’s a reason the secular Whites here are more liberal than their secular White counterparts in the rest of the country: these places attract those on the left. Everywhere in this category is either a college town (Iowa City, Ann Arbor, Boulder, Ithaca, Lawrence, Madison, Fort Collins) or has some other reason for liberal politics (environmental issues in Maui or Gilpin County, CO; traditionally liberal cities like both Portlands, the Bay Area, or Asheville). The GOP could probably attract some of these voters by moderating on social and environmental issues, but it likely would not be worth the cost. If I were a Republican strategist, this would not be a top priority.
Category 2: Government Counties
Target Level: 1 out of 10
Population: 6 million
Examples: The entire DC Area, Eaton County, MI; Thurston County, WA
Analysis: It’s a feature, not a bug, that the anti-government party does poorly with government workers. Improving in these areas would hurt the GOP with base groups. Not worth the cost.
Category 3: Upper Midwest
Target Level: 6 out of 10
Population: 8 million
Examples: Small and medium-sized counties in Minnesota, Wisconsin, and Iowa
Analysis: This region’s Nordic heritage and collectivist mentality has made it a tough nut to crack for Republicans relative to how one would expect they’d be performing. However, there’s no crystal-clear reason they shouldn’t be doing better here. And for a party desperate to expand the playing field in the Electoral College, Minnesota (and to some extent Wisconsin, which still tilts Democratic but is very competitive) is a good place to start. But it would probably require a bit of movement leftward on economic issues.
Category 4: Oregon, Washington, and Idaho
Target Level: 5 out of 10
Population: 6 million
Examples: Western Washington and Oregon and the larger Idaho counties
Analysis: Eastern Washington and Oregon are performing at expected levels for the GOP, which means it’s time for them to try harder West of the Cascades. I’m as surprised as you that Idaho is on this list at all, but Ada (Boise), Bonneville (Idaho Falls) and Kootenai (Coeur d’Alene) counties are all worse for Republicans than they should be, which apparently means the GOP should be racking up the score like crazy. This won’t matter one iota in the Electoral College or in any race whatsoever, really, so it’s not a priority but is interesting nonetheless. Oregon and Washington are more socially liberal than the Upper Midwest, and it’ll take some balancing to appeal to both. But these last two categories really show that the late 90s/early 00s gains in these rural White states have been lost but really shouldn’t have been.
Category 5: Philadelphia Metropolitan Area
Target Level: 6 out of 10
Population: 2 million
Examples: Chester and Montgomery Counties (PA), Gloucester and Burlington Counties (NJ)
Analysis: The Philly suburbs are electorally important, at least on the Pennsylvania side of the river, and the upper-income ones are weak for the GOP. This should be a pretty easy place to improve.
Category 6: Border South
Target Level: 9 out of 10
Population: 4 million
Examples: Grundy County, MO; Roane County, TN; Trigg County, KY; Anderson County, TN; Christian County, MO; Anderson County, KY
Analysis: This should be the juiciest fruit of all, and it would pay dividends at the state level in defeating the remaining Blue Dogs. These areas should continue their realignment for the forseeable future, but it’s noteworthy that they haven’t made it there yet. Perhaps the evangelicals here are less conservative than other areas. But if an evangelical isn’t voting GOP by now, it’s hard to imagine what else could win them over.
Category 7: Rural New England
Target Level: 4 out of 10
Population: 2 million
Examples: Hampshire County, MA; Kent County, RI; Tolland County, CT; Berkshire County, MA
Analysis: These are more rural versions of Peoples’ Republics in many cases, but they slipped away from the GOP more recently as the party took on a Southern-fried flavor. Winning the areas back would require some major shifts, and it isn’t really that much territory in the first place.
Thus, if I were a GOP strategist, I’d advise making a play for: the Border South (where the party hasn’t maxed out), the Upper Midwest, Western Oregon and Washington (where the party has really lost ground in the past decade), and the Philadelphia suburbs.
Finally, here is a list of populous counties which fit none of these eight categories but are also fertile Republican ground (or should be).
McLean, Tazewell, and Lake Counties, IL
Monroe, Bay, and Oakland Counties, MI
Wayne and Wood Counties, OH
Porter County, IN
Jefferson County, CO
Warren County, NJ
Most of these are in the Midwest and many are in electorally important states; they could use work too. I’m not sure what’s up with Jefferson County.
Democrats, despite a moderately advantageous Electoral College position, have plenty of areas to grow as well. Here they are.
Category 1: Hispanic Semi-Rural
Target Level: 10 out of 10
Population: 12 million
Examples: Yuma County, AZ; Kings County, CA; Madera County, CA; Cameron County, TX; Yakima County, WA; Fresno County, CA; Nueces County, TX; El Paso County, TX; Dona Ana County, NM
Analysis: These areas suffer from one of two problems: Hispanics are disproportionately non-citizens or Hispanic citizens vote at lower rates than they do nationally. The first problem just takes time, while the second could use some immediate work, but either way Inland California, South Texas, Yakima, Yuma, and New Mexico are good areas for Democrats to focus on. But of course, most of you knew about all the unregistered Hispanics. The next group, however…
Category 2: Deep South and >33% Black
Target Level: 7 out of 10
Population: 8 million
Examples: Williamsburg County, SC; Lowndes County, AL; Sumter County, GA; Bolivar County, MS; Marion County, SC; Randolph County, AL; Ouachita County, LA; Dougherty County, GA
Analysis: I’m a bit confused how, even with the highest Black turnout ever in 2012 (evident even in non-swing states in the county level results), these areas show underperformances. It isn’t racism, because the discrepancy essentially only appears in counties with large minorities (or majorities) of Black voters. So I guess that, surprisingly, there’s still room to grow with rural Black turnout in the Deep South.
Category 3: East and Central Texas
Target Level: 5 out of 10
Population: 4 million
Examples: Brazos County, TX; Grimes County, TX; Nacogdoches County, TX; Rusk County, TX; Erath County, TX; McLennan County, TX
Analysis: This problem is mostly unique to Texas and exists in both rural and more populous counties. It’s clearly due to White voters and not Hispanic ones. I have no idea what could be causing it. But Democrats involved in Battleground Texas shouldn’t forget about the Eastern or Central parts either.
Category 4: Retirees
Target Level: 2 out of 10
Population: 3 million
Examples: Indian River County, FL; Onslow County, NC; Cochise County, AZ; Yavapai County, AZ; Ocean County, NJ; Jones County, NC; Sumter County, FL
Analysis: This is one of those geographical quirks I mentioned before. Age matters.
Category 5: Cajun Country
Target Level: 3 out of 10
Population: 2 million
Examples: Essentially all of Southern Louisiana
Analysis: Probably lots of conservative Catholics here. If that’s the cause, this region is voting how one would expect. If it isn’t, then I’m not sure what would be causing it, but Democrats should be trying harder. John Kerry performed quite well just ten years ago, despite being a New England liberal.
Category 6: Super-Rural Plains
Target Level: 4 out of 10
Population: 2 million
Examples: western Kansas, Nebraska, Oklahoma, and Texas, plus eastern Colorado
Analysis: This area is super White. It’s not extremely religious, yet votes as if it were. The states listed don’t really matter, other than Colorado, so it’s not really too much worth focusing on as a party, but many of these voters could very well be persuadable (or perhaps they aren’t, due to other issues such as guns which aren’t accounted for here).
Category 7: Coal Country
Target Level: 1 out of 10
Population: 2 million
Examples: West Virginia and Kentucky, plus Fulton County, PA
Analysis: Definitely not worth the cost. We know why these not-so-religious, poor voters are Republican supporters. The Democratic Party has made a calculated decision to embrace environmentalism and abandon coal. That has cost them tons of support in these two states (and some in the more important state of Pennsylvania), but also led to gains in other places. While I would personally prefer to win West Virginia and lose Connecticut, that’s not the route the party has chosen to go and it would be dumb to risk so much to gain so little.
Category 8: Rural West
Target Level: 4 out of 10
Population: 1 million
Examples: Lake County, OR; Fremont County, CO; Tehama County, CA; Lake County, MT
Analysis: See Category 6. Guns may be a factor here, too. Again we see parts of Colorado show up; it’s just a demographically strange state when it comes to voting. If not, I’d say the party should go for it: Colorado and Montana could both pay dividends.
Category 9: Pennsyltucky
Target Level: 7 out of 10
Population: 2 million
Examples: Central Pennsylvania
Analysis: Pennsylvania is a key state, and Republicans are doing better than they should in the central part of the state. I’m not sure what else would be causing it except historical loyalty, which means it should be fertile for Team Blue.
Other Populous Counties Democrats Should Work For:
Randolph, Wayne, Davidson, and Alamance Counties, NC
Whitfield County, GA
Colleton and Berkeley Counties, SC
Lee County, AL
Allen County, OH
Richmond County, NY
As you can see, a majority of these counties are in the Carolinas and nearly all are in the South. They aren’t heavily evangelical, though. North Carolina is of course key and should be priority number one for the party after Hispanics. Allen County, Ohio is an interesting one to appear, as is Staten Island.
So those are the areas each party should go for. I hope you learned something from this diary. Please express your thoughts in the comments.