Figure 1: The Whole Senate
Control of the senate is very important. If you don’t believe me, just ask Justice Merrick Garland.
In order to have a decent chance of flipping the senate, Democrats need to vigorously compete in all the winnable races, not just the easy ones. That includes competing in some reddish states. They don’t need to win all not-so-easy races, but they need to win at least one. (Actually they should vigorously compete in all races, but for today let’s just talk about the easiest of the hard-to-win races.)
In particular, I’m talking about giving a high priority to all the races with horizontal tie-lines in figure 2. These are ones where data suggests the incumbents should be vulnerable, because they don’t do a good job of representing their constitutents.1
Figure 1 shows the data for the whole senate. The layout of the points shows the behavior of senators (as measured by their dw-nominate score)2, 3 and the political lean of their state (as measured by the Cook PVI).4
Some Democrats have a quaint notion that senators should represent all their constitutents (not just the ones who voted in the partisan primary). One way (but not a good way) of representing this idea is suggested by the diagonal yellow line in figure 2. You can see that there is “some” loose tendency to follow this line; that is, the bluer districts kinda sorta tend to have relatively liberal senators, and the redder districts kinda sorta tend to have relatively conservative senators.
However, the yellow line is not a good fit to the data, and there is no reason why it should be. A large PVI in the R direction means that the state has a larger percentage of R voters, not that the voters are more extreme. Any sufficiently-large PVI means that a senator can get away with extreme behavior, but making the PVI yet larger doesn’t motivate (or justify) yet more extreme behavior.
It makes more sense to fit the Democrats separately, as shown by the dashed blue line. This line shows “some” sensitivity to the constituents, but only 1/3rd as much as the quaint yellow line would suggest.
Note that Sinema is remarkably far to the right. She is the #3 rightmost Democrat in the whole senate. Unlike Manchin or Jones, she cannot argue that she is just representing her constituents, since she is farther right than can be explained by the PVI of her state, since the state is only slightly reddish.
We can also fit the Republican delegation separately, as shown by the red dashed line in the figure. You can see that the political lean of their state has almost no effect on these guys. They exhibit essentially no tendency to represent all constituents. Instead, they form a pretty tight herd near nom1=0.5, with no outliers to the left of there, except for Collins and maybe Murkowski.
The explanation is simple: The Republican party generally does not tolerate moderates. Murkowski got primaried in 2010, and had to run as a write-in candidate to keep her seat. Collins is tolerated because she comes from the bluest (by far) of the R-held seats.
The dashed blue line and dashed red line have different slopes, different by a factor of 5½.
Figure 2 shows a subset of the data, namely the 35 senate seats that will be contested in 2020. The point is to focus on the races that most need our attention.
I would add Graham to the list of senators who seriously need to be challenged. The boundaries of the “gray area” is not super well defined; you could easily broaden the definition to include him. And he deserves to be challenged, just for being such a raging hypocrite.
The horizontal tie-lines on the graph highlight senators who are far enough out of step with their constituents to be remarkably vulnerable in the 2020 elections.
The dark-shaded band in the figure is the proverbial “gray area”. We know from looking at recent House elections that races in this band could go either way.1 So, any R-held seat in this band (or below!) is a pickup opportunity for Democrats. Any D-held seat in this band (or above!) is a pickup opportunity for Republicans.
Note that Charlie Cook5 rates 11 senate races as “Likely D” or bluer, and 19 as “Likely R” or redder. He rates only 4 of the 34 races as toss-ups, namely Jones, Collins, Gardner, and McSally. The problem is that from a Democratic point of view, that is an unacceptable answer. Dems would have to win all four of those toss-up races – plus the vice presidency – to take control of the senate with nothing to spare.
So the action item is clear: Dems need to get their act together and go after Tillis, Ernst, Cornyn, Perdue, and the Isakson seat. Preferably Graham also. That’s in addition to the four toss-up races. It won’t be easy, but it has to be done.
Polls do not tell you what “will” happen; instead, they are at best a lagging indicator of what has already happened. And with all due respect to MLK, the arc of the moral universe does not bend of its own accord. If we want it to bend, we have to get out there and make it bend. Poll numbers are not etched in stone; the whole point of campaigning is to make the poll numbers change. I have seen far too many candidates foolishly wait until the last minute and then obsess over turning out their base.
Turning out the base is not enough. Instead, it would be much better to engage early. Engage on the issues. Explain why you deserve to win. You have to win the argument before you can win the election.
When I say “engage on the issues” I’m not talking about a blizzard of wonky policy papers. In 2016, Hillary thought if she put out just one more white paper she would win the vote by 100 million to nothing. She never figured out why that wasn’t correct. To this day she – and a lot of other people – still haven’t figured it out. The #pussygrabber isn’t selling policy; he’s selling attitude, and his base is still eagerly buying it. Seriously: Campaign loudly and clearly on attitude and character:
A properly-run Democratic campaign should make every district less red. That is, it should move the PVI numbers downward everywhere. This is another reason why it is important to vigorously compete in all races, even the hard-to-win races, even in the reddest of districts. There’s no law that says they have to stay red forever. Even if you don’t win in the short term, you might win in the long term. That is: Even if you don’t win the election, you might win the argument
If you toss a fair coin 4 times, there is only one chance of 16 that it will come up heads every time. So you might think the chance of winning all four “toss-up” races is only 1 in 16. But that’s not how it works, because of correlations.
Suppose you take 4 coins and glue them to an index card, all facing the same way. Then if you flip the card, the chance of getting 4 heads is 1 in 2, not 1 in 16. Big difference. The same idea applies to political races. A national headwind could cause you to lose all 4 races, and a national tailwind could cause you to win all 4 races.
Leading up to the infamous 2016 election, the polling data was correct. However, a lot of people misinterpreted the data. That includes a lot of people who really should have known better. They should have taken correlations into account. I said so at the time. Nate Silver said so at the time.
Suppose I am offered the opportunity to gamble on the roll of a pair of fair dice. It costs me $1.00 per roll to play the game. If I correctly predict the outcome, the payoff is $8.00.
I will happily accept that offer. I predict the outcome will be 7 each time. On average, I am wrong 5 times out of 6, but that means I am right 1 time out of 6, and the payoff is 8-to-1, so I turn a handsome profit on average. Assuming there are no hidden costs, I will happily play this game a million times.
A lot of people don’t understand how this works. They think that if I make a prediction that is wrong 5 times out of 6, it’s a bad prediction. In contrast, I insist that winning is not just how often you win, but how much you win.
This really matters in politics. If you compete in a hard-to-win race, you probably won’t win. But if you do that enough times, you will win one of them, and it could make an eeeeenormous difference.
Please, let’s get with the program. Compete in all the winnable races, not just the easy ones. Recruit some strong candidates. Support them during the campaign. Vote for them.
