Suppose you’re settling in for a nice evening of watching election returns, and upon checking the very early results, you find that your preferred candidate is suddenly trailing by double digits, despite the pre-election polls that had him or her leading. If you’re fairly smart, you might realize that in some states, rural counties tend to report before the urban counties.
Instead of gnashing your teeth and rending your garments, you could dig further into the election results to see results reported by county, and you might notice that the bulk of the results are coming in from Bumbelch County — at which point you might grab the atlas, look at the teeny-tiny fine print in the back where the population data is, and realize that there are only 5,000 people in Bumbelch County, so it’s only a very small and probably unrepresentative slice of the entire state, and you might as well sit tight and wait for more and bigger counties to report, instead of packing your suitcase to flee the country.
Or, if you’re really smart, you could just bookmark this article and refer back to the tables for various key states, where we isolate which counties are the important ones where most of the population is. Moreover, you could also look at the tables to see what percentage of the vote your preferred candidate should be getting in each of those key counties, in order to stay on the right pace to win statewide. Different counties come in at different times, but using these tables, you can infer right away whether a county that’s reporting early is coming in at the right level for your preferred candidate.
This year, we’re going to look at the states that have “Tossup” or “Lean Republican” Senate races, where control of that chamber is likely to be determined. We’ll also look at a few states with no Senate race (or a generally uncompetitive one), but where there’s an important Tossup gubernatorial race.
First, though, you might be wondering, “wait, does this method actually work?” The short answer is, yes! In the Alabama Senate and Pennsylvania's 18th district special elections, for instance, it was spot on, not just in terms of what numbers Doug Jones and Conor Lamb needed to hit, but closely predicting the actual county-level numbers they got (though, in both cases, I got somewhat lucky, in that in both races they ended up at around 50 percent of the vote, which is the target that I usually model toward).
The long answer, though, is “not always.” For instance, in 2016, there were certain states where Hillary Clinton hit most of her recommended benchmarks in the large counties … and lost those states anyway. Those were the same states where the pollsters whiffed, where Republicans ran up the score in the very small counties that are too unpopulated to make the list, but where Clinton lost, say, 80-20, instead of 70-30 like Barack Obama.
In addition to the “small counties” problem, there were also some large counties (like Milwaukee County, Wisconsin, as a prime example) where she dramatically exceeded her benchmarks in terms of percentage of the vote share, but turnout was inadequate; in other words, Milwaukee County’s share of the total number of votes cast in the state was lower than usual. (My benchmarks tables include both sets of information, not just the percentages of the vote share candidates need to hit, but also the percentage of the state’s total votes that that county needs to account for.)
So, it’s rather a blunt instrument. By design, it has to be, since we just don’t have the information to make assumptions beyond uniform swing from the previous election (2016) to the next one. A candidate with a particular strength in one region of a state could underperform slightly across much of the state, for instance, if he was poised to overperform on his home turf and balance it all out. That’s the kind of pattern that you can’t predict, even by looking at the crosstabs of polls.
But these tables aren’t intended to help you run a better door-knocking campaign or doing anything else at a professional level, where they have access to much more detailed data than I do; they’re just intended to help you be a more informed viewer on Election Night.
For each table, you’ll see three columns of numbers. The left one is the percentage of the state’s total votes that come from each particular county. (Counties with less than 2 percent of the state’s votes aren’t included, just to keep everything a manageable length here. As I stated above, the small counties can, cumulatively, make a big difference, though.) The middle column is the target the Democratic candidate should be shooting for, usually in order to shoot for 50 percent statewide and the barest possible majority. (In a few states, where third party candidates are polling at more than the usual 1 or 2 percent of the vote in the current polling averages, I’m modeling toward 49 instead.) The right column is the 2016 result, so you understand the baseline for the current 2018 estimates.
Kentucky’s 6th Congressional District
COUNTY |
% OF ‘16 DISTRICTWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PREs. |
DISTRICTWIDE |
--- |
50/50 |
39/55 |
fAYETTE |
40.9 |
62/37 |
51/42 |
MADISON |
11.2 |
43/57 |
32/63 |
SCOTT |
7.3 |
43/57 |
32/62 |
FRANKLIN |
7.2 |
56/44 |
45/50 |
JESSAMINE (PT.) |
4.9 |
39/61 |
28/66 |
CLARK |
4.9 |
40/60 |
29/66 |
WOODFORD |
4.1 |
48/52 |
37/57 |
MONTGOMERY |
3.4 |
38/62 |
28/69 |
ANDERSON |
3.4 |
34/66 |
23/72 |
BOURBON |
2.6 |
43/57 |
32/63 |
We’re going to start off with a little bonus content. I’ve never set benchmarks for a House race for a general election before, just because there are so many other races to talk about (though I certainly have benchmarked them for special elections). But I’ll begin with Kentucky’s 6th district, a race that has caught a lot of people’s attention, thanks to a compelling campaign from Amy McGrath.
Largely, though, that’s because Indiana and Kentucky are the first two states to report, and other than Indiana’s Senate race, there won’t be a whole lot else that’s competitive to look at closely in that first hour. Also, it’s a good indicator of what someone has to do to win in a district as red as this one (at least at the presidential level; it was held by a Democratic Rep. until 2012). Here, the formula is actually pretty simple: McGrath is going to need to dominate in Fayette County (where Lexington, and the University of Kentucky, are located), which is the blue part of the district as well as nearly half its population. She’ll need to try and make it into, ideally, the low 40s in the rural remainder of the CD.
ARIZONA
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/49 |
45/48 |
MARICOPA |
60.2 |
50/49 |
45/48 |
PIMA |
16.2 |
58/41 |
53/40 |
PINAL |
5.0 |
42/57 |
37/56 |
YAVAPAI |
4.4 |
36/63 |
31/62 |
MOHAVE |
3.1 |
28/72 |
22/73 |
COCONINO |
2.3 |
60/35 |
54/35 |
YUMA |
2.0 |
52/47 |
46/47 |
Let’s go into a little more detail about the methodology here. If you’re a regular consumer of our county benchmarks, you might know that it’s usually a very simple math process. For instance, in Arizona, the 45/48 result in 2016 would be turned into a 50/43 target in 2018, accomplished by adding 5 to the 2016 Dem vote share in each county and subtracting 5 from the GOP vote share in each county. However, 50/43 isn’t a very realistic target, because it’s a more conventional election this year, with less-disliked candidates, and it’s highly unlikely that the third party candidates this year will take 7 percent of the total vote.
So, what I’ve done this year is create an intermediate step (which I’m not showing in the table, to minimize onscreen clutter, but I’ll describe the process to make it more transparent), where, in Arizona, I would first add 3 to both sides of the 2016 result, so 45/48 becomes 48/51, and the two-party share gets bumped up to 99 percent, which is more typical in most elections. (It’s possible that the split in 2016 wasn’t exactly half-and-half, in other words, the number of Democrats defecting to the Greens or Libertarians in 2016 might not have been equal to the number of Republicans defecting to Libertarians or Evan McMullin’s campaign. There’s no good way to figure that out with any certainty, though, so I’m simply splitting the difference equally.) Only then do I switch over to the old method, where I would add 2 to the Democratic side and subtract 2 to the GOP side, to bring the target vote share to 50/49 that would get Kyrsten Sinema across the finish line by the barest margin possible.
For purposes of an example, there’s no point in applying that to the largest county in Arizona, Maricopa County, since it takes up such a large percentage of the state’s population (it contains Phoenix and most of its suburbs) that, in terms of its numbers, it’s essentially a microcosm (or megacosm?) of the whole state. So if you apply it to Pima County (where Tucson is), instead, you’d see that in 2016, that county went 53/40 for Clinton. After the intermediate step, that becomes 56/43 for Clinton. Adding 2 to the Democratic side and subtracting 2 from the Republican side, that gets turned into a 58/41 target for 2018.
FLORIDA
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/50 |
47/49 |
MIAMI-DADE |
10.4 |
66/34 |
63/34 |
BROWARD |
8.8 |
69/31 |
66/31 |
PALM BEACH |
7.0 |
59/41 |
56/41 |
HILLSBOROUGH |
6.3 |
55/45 |
52/45 |
ORANGE |
5.8 |
63/36 |
60/35 |
PINELLAS |
5.2 |
50/49 |
47/48 |
DUVAL |
4.6 |
50/49 |
47/48 |
LEE |
3.5 |
41/59 |
38/58 |
BREVARD |
3.4 |
41/58 |
38/57 |
POLK |
3.0 |
44/56 |
41/55 |
VOLUSIA |
2.8 |
44/55 |
41/54 |
PASCO |
2.6 |
40/59 |
37/58 |
SARASOTA |
2.4 |
45/55 |
42/54 |
SEMINOLE |
2.4 |
50/49 |
47/48 |
In the perpetually close swing state of Florida, Andrew Gillum and Bill Nelson both need to clean up in the counties of the Miami metropolitan area (Miami-Dade, Broward, and Palm Beach), while eking out a victory in the Jacksonville and St. Petersburg areas, and holding onto at least 40 percent in the rest of the state’s most populous counties.
GEORGIA
COUNTY |
% OF ‘16 Statewide VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/49 |
45/50 |
FULTON |
10.6 |
73/26 |
68/27 |
COBB |
8.1 |
53/45 |
48/46 |
GWINNETT |
8.0 |
55/43 |
50/44 |
DEKALB |
7.7 |
84/15 |
79/16 |
CHATHAM |
2.7 |
60/39 |
55/40 |
CHEROKEE |
2.7 |
27/71 |
22/72 |
HENRY |
2.4 |
55/45 |
50/46 |
FORSYTH |
2.4 |
29/70 |
24/71 |
CLAYTON |
2.3 |
88/12 |
84/14 |
Georgia doesn’t have a Senate race this year, but it does have a very important gubernatorial race, and modeling to 50 percent is especially important in this one because Georgia law states that if neither candidate hits 50 percent or more, the top two finishers have to advance to a runoff. Thanks to the presence of a Libertarian candidate, this is a distinct possibility, though he’s only polling around 2 percent, so he has very little impact on the overall totals.
Hillary Clinton winning in suburban Cobb and Gwinnett Counties, long-time Republican strongholds until recently, was a big step forward, but Stacey Abrams will still need to exceed Clinton’s numbers in those counties in order to win statewide. Luckily, polls of Georgia’s 6th and 7th congressional district races (found in those same counties) that have asked the gubernatorial question have found that Abrams, indeed, is leading by a significant amount there.
INDIANA
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 49% |
2016 PReS. |
STATEWIDE |
--- |
49/48 |
37/56 |
MARION |
13.3 |
70/28 |
58/36 |
LAKE |
7.4 |
70/29 |
58/37 |
HAMILTON |
5.7 |
49/48 |
37/56 |
ALLEN |
5.4 |
49/48 |
37/56 |
ST. JOSEPH |
4.1 |
59/39 |
47/47 |
PORTER |
2.8 |
55/42 |
43/50 |
HENDRICKS |
2.8 |
42/55 |
30/63 |
VANDERBURGH |
2.7 |
51/47 |
39/55 |
JOHNSON |
2.4 |
38/60 |
26/68 |
ELKHART |
2.4 |
43/55 |
31/63 |
TIPPECANOE |
2.3 |
55/41 |
43/49 |
MONROE |
2.1 |
71/27 |
59/35 |
Indiana is one of the more important states on this list, not only because Democratic hopes of winning control of the Senate are nearly non-existent if incumbent Joe Donnelly happens to lose (the polling average currently puts him up by 2 points), but also because Indiana, along with Kentucky, is one of the first two states to start reporting. So we’ll have most of an hour at the start of the evening to contemplate how Donnelly is doing, in some detail, before the onslaught of results from a variety of other states start pouring in.
Indiana is one of the two states where I’m modeling to 49 percent rather than 50. That’s because the Libertarian candidate has been polling at 4 percent in our polling average (perhaps thanks to a little ratf*cking by Democrats), and we’ll need to assume that that candidate pulls in something like 3 percent in the end. (Indiana isn’t a state known for its enthusiasm for third-party candidates, compared with, say, Montana. Surprisingly, the Libertarian in the Montana Senate race is only polling at 1 percent in our polling average.)
MISSOURI
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/50 |
38/56 |
ST. LOUIS |
18.4 |
67/33 |
55/39 |
JACKSON |
10.8 |
67/32 |
55/38 |
ST. CHARLES |
7.2 |
46/54 |
34/60 |
ST. LOUIS CITY |
4.7 |
90/10 |
79/16 |
GREENE |
4.6 |
45/54 |
33/60 |
CLAY |
3.9 |
53/46 |
41/52 |
JEFFERSON |
3.8 |
42/58 |
30/65 |
BOONE |
3.0 |
61/36 |
49/43 |
If you’d like me to Show You how Claire McCaskill ekes out a victory in Missouri, here’s how. She has to absolutely dominate in the St. Louis, Kansas City, and Columbia areas (keep in mind that St. Louis proper is an independent city, separate from suburban St. Louis County), while not getting overly blown out in the rest of the state.
MONTANA
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/49 |
35/56 |
YELLOWSTONE |
14.0 |
46/51 |
31/58 |
MISSOULA |
12.1 |
67/30 |
52/37 |
GALLATIN |
10.7 |
60/37 |
45/44 |
FLATHEAD |
9.5 |
43/57 |
28/64 |
LEWIS & CLARK |
7.0 |
56/41 |
41/48 |
CASCADE |
6.9 |
50/50 |
35/57 |
RAVALLI |
4..5 |
42/58 |
28/66 |
SILVER BOW |
3.3 |
67/32 |
52/39 |
LAKE |
2.6 |
51/49 |
36/57 |
One piece of trivia you might not know is that Jon Tester has never hit 50 percent of the vote in either of his previous statewide victories. That’s in large part thanks to a fairly large vote total taken by the Libertarian candidate in both previous wins. Montana just tends to be more amenable to third-party candidates in general, though, as you can see that Clinton plus Trump added up to only 91 percent of the vote in 2016. However, the Libertarian candidate in the Montana Senate race this year is only pulling in 1 percent in the polling averages, so we’re modeling to 50 percent here.
NEVADA
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 49% |
2016 PRES. |
STATEWIDE |
--- |
49/49 |
48/46 |
CLARK |
68.2 |
53/45 |
52/42 |
WASHOE |
18.7 |
47/48 |
46/45 |
DOUGLAS |
2.5 |
31/65 |
30/62 |
CARSON CITY |
2.2 |
39/55 |
38/52 |
LYON |
2.1 |
27/70 |
26/67 |
Nevada is the other state where we’re modeling to 49 percent, because in its Senate race, “other” is running at 3 percent in our polling average (presumably “other” includes Nevada’s unique, and notorious, option of “none of the above”). As you can see from the table, it really doesn’t matter what happens outside Clark and Washoe Counties (where Las Vegas and Reno are, respectively); the mostly-empty rest of the state barely makes up more than 10 percent of the total vote.
NORTH DAKOTA
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/50 |
27/63 |
CASS |
23.5 |
62/36 |
39/49 |
BURLEIGH |
13.9 |
46/54 |
23/68 |
GRAND FORKS |
8.8 |
59/41 |
36/54 |
WARD |
8.0 |
44/55 |
21/68 |
MORTON |
4.6 |
42/58 |
19/72 |
WILLIAMS |
3.7 |
36/64 |
14/79 |
STARK |
3.6 |
36/64 |
14/79 |
STUTSMAN |
2.9 |
48/52 |
25/66 |
RICHLAND |
2.2 |
50/50 |
27/63 |
Heidi Heitkamp’s last stand, if she’s going to win it, is going to be heavily dependent on winning by wide margins in the two most cosmopolitan places in the state: Cass County (where Fargo is) and Grand Forks County. It’ll also be heavily dependent on winning by wide margins in the rural counties that contain reservations; individually, though, those counties are too small to show up on this list. (Counties that fall in that category that you might check out separately include Rolette and Sioux Counties.)
OHIO
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/50 |
43/51 |
CUYAHOGA |
11.0 |
72/28 |
65/30 |
FRANKLIN |
10.6 |
67/33 |
60/34 |
HAMILTON |
7.4 |
60/40 |
53/42 |
SUMMIT |
4.7 |
59/41 |
52/43 |
MONTGOMERY |
4.7 |
54/46 |
47/48 |
LUCAS |
3.6 |
63/37 |
56/38 |
STARK |
3.2 |
46/54 |
39/56 |
BUTLER |
3.2 |
41/59 |
34/61 |
LORAIN |
2.5 |
54/46 |
48/48 |
WARREN |
2.1 |
36/64 |
29/66 |
LAKE |
2.1 |
47/53 |
40/55 |
MAHONING |
2.1 |
56/44 |
49/46 |
Ohio’s Senate race isn’t looking very competitive any more, but it has one of the most crucial gubernatorial races in the nation. While the bulk of the state’s Democratic votes are in Cuyahoga and Franklin Counties (Cleveland and Columbus, respectively), you might also keep an eye on more blue-collar counties like Lorain and Mahoning (the latter is where Youngstown is), which have been Democratic strongholds in the past, gave Clinton only very narrow wins in 2016, but now seem poised to snap-back significantly.
TENNESSEE
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/50 |
35/61 |
SHELBY |
13.5 |
77/23 |
62/34 |
DAVIDSON |
9.9 |
75/23 |
60/34 |
KNOX |
7.2 |
50/48 |
35/59 |
HAMILTON |
5.7 |
54/41 |
39/55 |
RUTHERFORD |
4.3 |
49/49 |
34/60 |
WILLIAMSON |
4.2 |
44/53 |
29/64 |
SUMNER |
2.9 |
40/59 |
25/70 |
SULLIVAN |
2.5 |
35/64 |
20/75 |
MONTGOMERY |
2.3 |
53/45 |
38/56 |
WILSON |
2.3 |
40/58 |
25/69 |
BLOUNT |
2.1 |
38/61 |
23/72 |
True story: when Phil Bredesen was being re-elected governor of Tennessee in 2006, he actually won every single county in the state. Tennessee has changed quite a bit in the last decade in terms of its willingness to vote for moderate Democrats downballot, and many of those counties that voted for him in 2006 wouldn’t in a million years vote for him now (especially for federal office). He’ll need to absolutely dominate in Shelby and Davidson Counties (Memphis and Nashville, respectively), while eking out wins in the smaller cities and trying to hold down the damage in the rural areas.
TEXAS
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/49 |
43/52 |
HARRIS |
14.6 |
61/39 |
54/42 |
DALLAS |
8.5 |
68/32 |
61/35 |
TARRANT |
7.5 |
50/49 |
43/52 |
BEXAR |
6.6 |
61/38 |
54/41 |
TRAVIS |
5.2 |
73/24 |
66/27 |
COLLIN |
4.0 |
46/53 |
39/56 |
DENTON |
3.3 |
44/54 |
37/57 |
FT. BEND |
2.9 |
58/42 |
51/45 |
EL PASO |
2.4 |
76/23 |
69/26 |
MONTGOMERY |
2.3 |
29/70 |
22/73 |
WILLIAMSON |
2.3 |
49/48 |
42/51 |
For Beto O’Rourke to scratch out a win against Ted Cruz, he’s going to put together a long string of feats that, even if they happened individually, would be considered a surprising accomplishment. He needs to push way past the 70 percent mark in his native El Paso and in Travis County (where Austin is), and come close to that mark in Dallas County. And he’ll need to outright win in GOP strongholds like Tarrant County (Fort Worth) and Williamson County (Austin’s suburbs). Collectively, it seems a little like running an under-two-hour marathon, but O’Rourke may well be the man to do it.
WISCONSIN
COUNTY |
% OF ‘16 STATEWIDE VOTE |
WHAT WE NEED TO BREAK 50% |
2016 PRES. |
STATEWIDE |
--- |
50/49 |
46/47 |
MILWAUKEE |
14.8 |
69/31 |
65/29 |
DANE |
10.4 |
74/25 |
70/23 |
WAUKESHA |
8.0 |
37/62 |
33/60 |
BROWN |
4.3 |
45/54 |
41/52 |
RACINE |
3.2 |
49/51 |
45/50 |
OUTAGAMIE |
3.2 |
45/55 |
41/53 |
WINNEBAGO |
2.9 |
47/52 |
43/50 |
WASHINGTON |
2.6 |
31/69 |
27/67 |
KENOSHA |
2.6 |
51/49 |
47/47 |
ROCK |
2.6 |
56/43 |
52/41 |
MARATHON |
2.3 |
42/58 |
38/56 |
LA CROSSE |
2.1 |
55/43 |
51/41 |
SHEBOYGAN |
2.0 |
42/56 |
38/54 |
Contrary to what you may have learned from political Twitter, Waukesha County actually isn’t all that crucial. Tony Evers doesn't need to win it, or even come that close there, in order to win statewide against Scott Walker. What he does need to do is dominate in Milwaukee and Dane Counties (the latter being where Madison is), stay about even in the smaller industrial cities of Wisconsin, and keep from getting blown out in Milwaukee’s conservative suburbs like Waukesha.