Reëlection Watch: May 19, 2012
Today marks the second semi-monthly edition of Reëlection Watch. This time, I’m going to show another model on handicapping the race, using the Cook Partisan Voting Index (PVI).
So, how are things going for the President lately?
Obama’s approval/disapproval rating in the Real Clear Politics average continues to sit mighty close to even. While it was a +0.6 point spread two weeks ago, and +0.7 now, these changes are sufficiently small as to be attributable to random noise. If you look at the graph, you’ll see declining oscillations over time, suggesting that we’re pretty much at stasis now. As I said last time, we should be looking for long-term trend signals in the noise, but those need at least two months of data to be considered signal, rather than noise. The Right Track/Wrong Track polls also remain pretty much the same. Last time it grew a little, and this time it shrank a little. The mid– to upper-twenties of negative spread seems to be the convergence zone here.
Even Congress’s spread, which tightened for a few months, seems to have stabilized around −64 points. The generic Congressional ballot has had more volatility than the other poll compilations, but I’ve come to believe that it’s more a function of the low frequency of those polls than any particular trends. That is, there’s enough noise that it’s essentially drowning out any signal.
As of yesterday, Intrade had Obama at a 57.4 percent chance of reëlection, down 2.5 points from last time. It appears that this is a slow trend of deflating odds, but it’s not clear what’s behind this trend.
Things are still mostly stable in a somewhat pro-Obama state, but with a trend moving toward a closer race.
Unemployment data are essentially unchanged over the past two weeks. We’ll get more data in a couple more weeks.
In terms of oil, West Texas Intermediate crude is trading at about $91, down 13 percent in the last two weeks, while Brent crude (representing the European market) has fallen almost 11 percent to about $107. The drop in oil prices is unusual for this time of year, and it’s hard to say what’s behind it. It’s good news, however, in terms of supporting more economic activity as we head into summer…and that is good news for Obama heading into the fall. I had expected gasoline prices to rise in response to last month’s oil price jump, but it appears that the recent sharp drop in crude prices has held gasoline to a gentle decrease instead. The national average is $3.71 per gallon, more than 20 cents below the same week last year.
Natural gas prices have risen from their bottom at two dollars per million British Thermal Units to about $2.50. This is still half the price of natural gas this time last year.
In general, things have been moving in a positive direction over the past two weeks. But we still need to wait until our next installment to see how unemployment looks.
The Electoral College
Last time, I showed you the margins from the 2008 election, and compared them to the recent polls in the same states, with the intent of illustrating the differences between the national trends and the state trends. This time, I’m taking the same notion from a slightly different perspective.
The Cook Partisan Voting Index (PVI) for states is calculated based on averaging the results of the last two Presidential elections. In particular, it averages the margins of the Democratic and Republican candidates in each state, relative to the national popular vote margin. So, if the Democratic candidate received three percent more votes in 2004 in a particular state than the national average, and five percent more than the national average in 2008, the state’s PVI would be D+4.
Since President George W. Bush received two percent more votes nationally than did Senator John Kerry (D-MA), this would have meant that the hypothetical state described above would have had Kerry taking the state with a four-point margin. This is because Kerry would have had three more percentage points of votes, which would have meant that Bush would have had three fewer percentage points (this presumes an insignificant impact of the other parties; naturally, the calculations are a bit different if other candidates receive significant numbers of votes). That totals a six-point shift from the national average, totaling a four-point margin for Kerry.
To reverse the process for a prediction of the individual state results, take the national poll average, and add the PVI to it. Since current national polling runs even, this essentially means looking strictly at PVI.
I separated the two elections, and created PVIs for 2004 and 2008. Then I calculated the difference between the two, and ran a linear extrapolation based on that difference.
|District of Columbia||R+2.0||D+38|
I readily admit that this is a crude analysis. The factors that led a particular state to move on the PVI scale are numerous, and may not have continued. For example, Hawaii almost certainly increased so much because Obama was born there. The above table is merely meant to be a rough guide to add some perspective.
That said, there are some interesting possibilities that come from this projection. Minnesota, which has long supported Democratic Presidential candidates, has been shifting rightward. Could it be closer to a tossup than we’ve been assuming? Montana has had a similar leftward shift.
If we take this PVI projection as a prediction, the outcome looks like this:
|District of Columbia||D+38||3|
Obama’s margin of 1.7 points would mean that (assuming the projected PVI is accurate), he’d take all states from North Carolina down, giving him 299 electoral votes.
For Romney to win in that breakdown, he’d have to take all of the states up to, and including, Pennsylvania, which would give him 276 electoral votes. This would require a national margin of 1.2 points (my spreadsheet has the additional significant digits). It’s within reach, but certainly not a gimme.
With the PVI trend in hand, I’ll soon be comparing it to the states with appreciable polling. This should give a view to the credibility of PVI trending, which we can use to apply to some of the less heavily polled states as well.
In the meantime, let’s see what the current Electoral College looks like, based on current polling data:
Things have moved a bit left in the past two weeks. Here’s what’s changed since last time, from most likely to go Romney’s way to most likely to go Obama’s way:
- South Carolina and Georgia are considered “Leans Romney” by RCP. I think that’s too generous, particularly for South Carolina, whose PVI moved to the right from 2004 to 2008. Even Georgia, which moved one PVI point to the left from 2004 to 2008, would still be at R+6 if it moved another point this time. I just don’t see either of these states being in play at all.
- Montana is now considered by RCP to be a “Leans Romney” instead of “Likely Romney” state. With the most recent Rasmussen poll giving Romney a seven-point lead, I’m inclined to agree, particularly since Montana’s PVI shifted over four points to the left from 2004 to 2008.
- Arizona is still called a tossup by RCP. The home state effect makes Arizona’s three-point PVI shift to the right from 2004 to 2008 suspect as a trend this time around. But it’s still a pretty conservative state. We’ll have to wait and see what numbers show up later.
- Missouri got another new Rasmussen poll, giving Romney a mere three point lead. The polls would put the Show Me State into tossup territory for sure. On the other hand, the projected PVI is R+5, which suggests that perhaps it’s near the border between “Tossup” and “Leans Romney”.
- North Carolina remains solidly in tossup territory. The Public Policy Polling and Rasmussen polls each show Obama and Romney, respectively, as the leader. Rasmussen shows a bigger lead for Romney than last time, but it’s hard to conclude that any movement over the past two weeks is signal, rather than noise. We could be seeing some impact from the aftermath of Amendment 1, though. North Carolina has also been moving to the left over the past several years, giving a projected PVI of R+2. So it’s a bit on the Romney side, but not strongly so.
- Florida, had two new polls, one from Quinnipiac and one from Suffolk. They straddle the even mark, leaving Florida a pure tossup from a polling perspective. On the PVI side, Florida has been inching to the right, with a projected PVI of R+3. This gives Romney a slight edge.
- Ohio had three new polls since last time. Purple Strategies, Public Policy Polling, and Quinnipiac all give the edge to Obama, by five, seven, and one point, respectively. That’s more Obama than in Florida, but not quite enough to leave “Tossup” land for “Leans Obama”. It’s purple, particularly when factoring in a projected PVI of R+3.
- Nevada hasn’t had any polls since my last installment. The PVI projection is EVEN. That said, Rasmussen’s most recent poll gives Obama an eight-point edge. It’s just not enough to make me push the (Nate?) Silver State into “Leans Obama”.
- Virginia got a new poll from the Washington Post; the seven-point Obama lead lines up with the recent Public Policy Polling result. Virginia’s projected PVI is R+4, which suggests to me that it’s probably going to go purple. I’m putting Virginia barely on the “Tossup” side of “Leans Obama”.
- Pennsylvania had one new Quinnipiac poll, which shows no real change over the past couple of months. With the projected PVI of D+1, Pennsylvania remains (barely) “Leans Obama”.
- Colorado got a new poll from Purple Strategies, showing the two candidates tied, though the projected PVI is D+3. I’m leaving it “Leans Obama”, and waiting for another poll to see if Purple Strategies is right to color the state purple.
- New Hampshire has a recent poll from Public Policy Polling that gives Obama a 12-point lead. The projected PVI is EVEN, but for now I’m putting New Hampshire in the “Leans Obama” column.
- Minnesota has one recent poll from SurveyUSA, showing a 14-point lead for Obama. The projected PVI is EVEN, though. For now, I’m leaving Minnesota at “Leans Obama”, since it’s pretty much the same story as New Hampshire.
- New Mexico is solidly in “Leans Obama” territory, with polls lining up with the projected PVI of D+7.
Based on this month’s polling, Obama needs 20 electoral votes (down from 34 last time) out of the 107 in the tossup group. Nearly any pair of tossup states, or Florida on its own, would do the trick.*
The more data I collect, the closer the election seems. That feels right to me intuitively. If you’ve been reading my Reëlection Watches over the past year, you know I believe we have a relatively weak incumbent running against a relatively weak challenger. This is a situation for which we have few examples in history upon which to draw.
For what it’s worth…on this day four years ago, Obama had a three-point lead over McCain, though that quickly narrowed to a statistical tie.
Obama remains in a good position relative to Romney, but there’s still plenty of opportunity for that to change between now and November.
*Note: in an earlier version of this article, I erroneously left Wisconsin in the “Leans Obama” category in the bar graph, which led to the incorrect conclusion that Obama needs only ten electoral votes. Wisconsin is still a “Tossup”.
- Jeremiah Wright and Kenya birthers could reëlect Obama and bring back Speaker Pelosi (thehill.com)
- Six Months Out, President Obama’s Campaign Still Faces Stiff Headwinds (swampland.time.com)