In last week’s inaugural Daily Kos Elections Senate forecast post, the good news was that the Democrats, on that day, were positioned to win 51 seats in the Senate and therefore the narrowest majority possible. However, I cautioned:
(I say “today,” because one bad poll in any of half a dozen states could tip the average in that state and move it below the red line tomorrow, meaning Democrats don’t take control of the Senate. That’s the kind of knife-edge we’re on, right now.)
And that’s what happened! Individual bad polls in two different states changed the averages just a little bit, but enough to give the Republican candidate a microscopic lead instead of the Democratic candidate having a microscopic lead. In each case, the movement within the state was small, but if you’re looking at the bottom line, that totally changes the overall result, with the net result being 49 Democratic seats, no change whatsoever from where we are now. (Which might have been a result people would’ve been pleased with at the start of the cycle, when there was a lot of worry about retaining the many Democratic incumbents in red states. Obviously, it’s a less pleasing result now that our confidence has increased across the boards.)
My advice, rather than rending your garments in woe, would be: check back here next week. These results (probably) don’t have anything to do with the ground having been shifted under everyone’s feet. The battle over the Kavanaugh confirmation last week (probably) didn’t cause a huge shift in activating Republicans who were going to stay home otherwise; it (probably) didn’t demoralize Democrats into giving up; it (probably) didn’t move many swing voters out of their undecided status. (For one thing, these polls predate the actual confirmation on Saturday.)
Instead, what changed from week to week was which pollsters decided to release results last week, and what kind of house effects, if any, those pollsters have. (We’ll discuss that in a little more specificity in the following paragraphs.) Pollsters with a likely voter model, or a weighting system, or an approach to making calls (such as random dialing vs. working off voter files), that generates a more favorable result for Democratic candidates, may report in the coming week, and the numbers may bounce back. (Or … you know … maybe they won’t? I wish I could offer more solid reassurance, but polling is full of endless surprises, which often make sense only in retrospect.)
Here’s the “totem pole,” revised for this week:
| STATE |
D CAND. |
D AVG. |
R CAND. |
R. AVG. |
DIFF. |
FLIP? |
| OHIO |
Brown (inc) |
50 |
Renacci |
37 |
+13 |
|
| WISCONSIN |
Baldwin (inc) |
52 |
Vukmir |
40 |
+12 |
|
| NEW JERSEY |
Menendez (inc) |
49 |
Hugin |
41 |
+8 |
|
| MINNESOTA (sp.) |
Smith (inc) |
44 |
Housley |
37 |
+7 |
|
| WEST VIRGINIA |
Manchin (inc) |
46 |
Morrisey |
40 |
+6 |
|
| MONTANA |
Tester (inc.) |
49 |
Rosendale |
44 |
+5 |
|
| NEVADA |
Rosen |
46 |
Heller (inc.) |
43 |
+3 |
D FLIP |
| FLORIDA |
Nelson (inc.) |
47 |
Scott |
45 |
+2 |
|
| INDIANA |
Donnelly (inc.) |
44 |
Braun |
42 |
+2 |
|
| ARIZONA |
Sinema |
46 |
McSally |
46 |
0 |
D FLIP? |
| RED LINE |
RED LINE |
|
RED LINE |
|
|
|
| MISSOURI |
McCaskill (inc.) |
45 |
Hawley |
47 |
-2 |
R FLIP |
| TENNESSEE |
Bredesen |
44 |
Blackburn |
47 |
-3 |
|
| TEXAS |
O’Rourke |
45 |
Cruz (inc.) |
48 |
-3 |
|
| NORTH DAKOTA |
Heitkamp (inc.) |
41 |
Cramer |
52 |
-11 |
R FLIP |
| MISSISSIPPI (SP.) |
Espy |
26 |
Hyde-Smith + McDaniel |
45 |
-19 |
|
One of the states that we’re concerned about, that fell below the red line of death, is Missouri, where last week at this time, Democratic incumbent Claire McCaskill had a 46 to 45 lead over Republican Josh Hawley in the polling average. This week, it’s a 47 to 45 edge for Hawley. That’s entirely because of one poll: a poll from McLaughlin & Associates that gave Hawley a 52-44 lead. If that name sounds familiar, McLaughlin is a Republican pollster that does internal polls for GOP candidates and interest groups. McLaughlin, in fact, is regarded as one of the worst Republican pollsters, with their greatest hit (or miss?) being their prediction that Eric Cantor would win his 2014 primary against Dave Brat by 34 points. (Cantor lost by 12.) Nonpartisan pollsters have always seen this as a tied race, or one within one or two points either way.
As I alluded to above, this is a frequently-polled race; if you don’t like the current weather in that race, just wait a few days and other pollsters will come along, and McLaughlin’s result will get shoved further down the trendline. (You might be asking: so why even include pollsters like McLaughlin in the trendline? Well, it wouldn’t be very intellectually honest to cherry pick pollsters, even the ones we don’t like; in fact, it would rather defeat the whole purpose of averaging the polls. Theoretically, partisan pollsters on each side should cancel each other out. You also need to allow for the possibility that the partisan pollsters might actually be right, which happens from time to time.)
The other state that slipped a few points and went below the line is Tennessee, an open seat in a dark-red state but one where Democratic ex-governor Phil Bredesen has been running neck-and-neck with Republican Rep. Marsha Blackburn. Last week, Bredesen had a 47 to 46 lead in the polling average; today, Bredesen trails 47 to 44. As with Missouri, the problem boils down to one poll: a poll from Anderson Robbins/Shaw giving Blackburn a 48 to 43 lead. Without any other new data this week, that swung the trendline quite a bit.
If you’re wondering who “Anderson Robbins/Shaw” is, they’re better known as “Fox News,” or more precisely, the pollsters that Fox News uses. Before you start screaming that we shouldn’t be including Fox News, of all sources, in our polling averages, keep in mind that “Fox News” polling is widely well-regarded within the polling world, putting up numbers that are generally in line with polling averages. (For instance, the same tranche of Fox polls here found Democratic leads in Arizona and Indiana.) “Anderson Robbins” and “Shaw” are two different partisan pollsters, one Democratic and one Republican, who presumably keep a check on each other or at least fight each other to a draw. (NBC News’s in-house polling operation works the same way, with a consortium of one D pollster and one R pollster.)
It should be noted that Fox’s previous poll of Tennessee, from early September, also gave Blackburn a similar-sized lead of 3 points. So one way to think of it is that Fox may have an R-leaning house effect in Tennessee (either in their weighting techniques or their likely voter model), even if they don’t have one in, say, Arizona. And that’s why it’s always a good idea to sanity-check polls by looking at the previous result in the same race by the same pollster; what may look like a big drop, relative to the polling average, isn’t actually a big drop when you do apples-to-apples comparisons of the same pollster. It doesn’t, of course, change the polling average, but it gives you more grains of salt to take with the polling average (which you should always do, regardless).
There’s one other race where Democratic fortunes took a turn for the worse, and that’s in North Dakota, where incumbent Heidi Heitkamp is facing Republican Rep. Kevin Cramer. Last week, Heitkamp trailed Cramer 48 to 45 in the polling average (so down 3), but this week, she’s trailing 52 to 41, an 11-point gap. Again, these polls were in the field a week ago, so this has nothing to do with Heitkamp’s “no” vote on the Kavanaugh confirmation; that happened after these polls were completed.
On the one hand, this didn’t affect the bottom line, because Heitkamp was already losing. If we were running a full-on model with Monte Carlo simulations, her larger deficit here would have a bigger impact on the Democrats’ overall chances. (Bigger, in fact, than the very small drops in Missouri and Tennessee, because a 3-point gap is well within the average’s error bars … while a 3-point differential doesn’t map out to 50-50 odds, it’s not that far from it when there are only a few polls in the sample. An 11-point gap, however, gives you about zero chance of winning, closing off one possible route to a majority if the Democrats don’t win say, Missouri or Tennessee.)
But on the other hand, one reason that we’re confident that Heitkamp is losing is because this is based on two new polls this week, not one. Of those polls, one is from Anderson Robbins/Shaw (who, again, also put up the bad result in Tennessee, though good results in Arizona and Indiana). They had her down 4 in early September, so this is a big drop even when doing apples-to-apples. The other is from a mysterious new pollster called “Strategic Research Associates,” who have literally no track record prior to this week. There’s no reason to think SRA is up to no good, though; they’re working on behalf of local TV affiliates, not candidates, and they simultaneously released results in Florida and West Virginia that were right in line with the polling averages, if not slightly better than average for Joe Manchin in West Virginia. Accordingly, we at Daily Kos Elections downgraded Heitkamp to “Lean Republican” in our qualitative ratings, our only ratings change on the Senate side last week.
Is there any good Senate news to report? Well, Jon Tester’s position in Montana improved incrementally; he’s now leading by 5 in the polling averages (49-44), boosted by two new polls from PPP and Gravis. That solidifies his position as being in the best shape of all the Tossup races. Beyond that, though, the best advice, as always, is to keep watching the averages every day, but without assigning too much significance to any one move, because when you have races that are literally on the knife’s edge, they’re very much going to oscillate daily back and forth between which side of the knife they’re on. (Also, keep checking our House and gubernatorial forecasts later this week, where the news is very good and mostly keeps getting better.)
