Senate races are annoying. Not in the general sense of most politics having an annoying element, but from the perspective of their occurring only every six years. House races are clockwork; same people up every two years. Presidential races are simpler, thanks to term limits. But the staggering of Senate races and the regularity of incumbent victories tends to put them out of step with the analytical tools that help us try and see patterns.
Despite this, the Wall Street Journal this week dove into the data to try and figure out what the past might tell us about the immediate future. That is, what the uneven pattern of Senate races might suggest about the November elections. Among the charts was a variant of the one below, showing that, over time, the winners of Senate races have less and less frequently been from the opposite party as the winner of the most recent presidential race in that state.
That's interesting, but the numbers are pretty small and, again, the Senate is weird. A few entrenched incumbents can throw off the whole mix.
So we decided to look at the patterns that apply to this year's most contentious Senate races, comparing the presidential vote over time to the Senate vote -- for both classes.
In general, there are three patterns that emerge. Three states can serve as an example.
Colorado's history is interesting. The presidential vote (the dashed line) has hovered around the midpoint of the vote. (The vertical axis shows the margin of victory for the winner of the race, with a 100 percent Democratic victory at top to a tie in the middle to a 100% Republican victory at the bottom.) The Senate results (loosely) track with the presidential results, but generally are more extreme.
The slopes of the lines between the Senate races are an interesting way of tracking the state's sentiment. If there are two incumbent senators, it looks very different than if there are a series of close Senate results (as you'll see in a second). In Colorado's case, the lines track with the presidential results; the slopes of each are the same, meaning that as the state voted more or less Democratic or Republican for president, it was doing the same for the Senate. In only one case, 2004, was the party that won the Senate race different than the closest presidential victor.
Iowa is very different. Iowa loves its incumbent senators, Tom Harkin (D) and Chuck Grassley (R). 2014 is unpredictable from this data because Harkin's retirement leaves his seat open.
And then there are states like Arkansas, in which the candidate makes a lot of difference. Incumbents have faced little or nominal opposition several times (those big spikes in 1990 and 2008). In a system without any sort of incumbent advantage (or with tight term limits), we'd expect to see what we see in Colorado -- the vote more closely reflecting party choices that you see in the presidential race. But that's not what we have.
The state has been voting more Republican in its presidential vote. But that doesn't seem to have meant much -- particularly when you consider why it voted for the Democratic president in 1992 and 1996 (hint: A place called Hope).
Alaska had an incumbent imbalance for a long time, too -- a streak broken in 2004. Mark Begich's squeaker of a victory in 2008 was an aberration, but the state has been trending more blue in its presidential votes as Senate races have gotten closer.
With the exception of the uncontested race in 1990, Georgia's vote has tracked reasonably closely with the presidential results. The exception is that 2000 race, a special election in a presidential year, won by ostensible Democrat Zell Miller.
The winner of the Kansas Senate race will likely be incumbent Sen. Pat Roberts(R). But it may be the closest race in the past 30 years in the state.
Kentucky is an interesting case. It's Senate races have been closer than its presidential results in all but one case since 1998. Which is the point at which we come back to the bigger picture: A battle between Alison Lundergan Grimes (D) and Sen. Mitch McConnell (R) is it's own beast, which could evolve in a million ways. But expecting it to be closer than a 20-point victory -- about the margin in 2012 for Romney -- doesn't seem like a bad bet.
Louisiana is a weird state, with some big incumbent wins and an overall rightward tilt. But compare its bipartisan voting patterns with ...
... Michigan, which has voted Democratic pretty consistently for president and Senate. That downward tick in 2012 denotes a nine-point win for Obama, which is a bigger margin than it might appear.
In Montana, the good news for Amanda Curtis (D) is that Democrats have done consistently better in close Senate races over time, but the overall pattern mirrors Iowa's. The bad news is that Curtis is the party's accidental candidate.
North Carolina is a state, like Colorado, that has seen a lot of close Senate and presidential results. As we noted in a separate post, don't be fooled by the downtick in 2012's presidential results; North Carolina has been slowly growing more Democratic, compared to the country on the whole. (That's why Michigan's 2012 downtick might not matter much. It's still on a very slight Democratic trend compared to the rest of the country.)
Oregon Sen. Jeff Merkley (D) has his first opportunity at benefiting from the incumbent advantage that Democratic senators in Oregon have enjoyed quite a bit.
Happily, West Virginia comes last alphabetically, letting us save an interesting one for the end. Let your eye follow the path of the Senate races, jumping across the gaps when there were no elections. It's the same overall wave form, except the presidential results peaked in 1996 and the Senate one peaked in 2000. Lots of gaps! Not many data points! But we work with what we have, and it would not be surprising at all if the 2012 election -- a reconfirmation of Sen. Joe Manchin (D), who'd only been elected two years prior -- were an outlier.
This doesn't mean that Republican Shelley Moore Capito is going to win. If, hoever, you happen to be in Vegas, don't bet against her.