Senate Races: What September Polling Trends Tell Us

Senate Races: What September Polling Trends Tell Us

By Sean Trende - September 25, 2014

About a month ago, I wrote a pair of articles about the 2014 elections, one regarding the Senate, and another focused on the House. The goal of the articles was to try to reconcile two disparate sets of data tied to this election. From the point of view of “fundamentals” -- that is, factors like the president’s job approval and incumbency -- the GOP should have been well on its way toward a clear wave.  On the other hand, from the point of view of polling, this was shaping up to be a very close, even disappointing election for the GOP.

We still don’t have definitive answers to any of these questions.  As I like to say when I give speeches, if I went into a coma and woke up in mid-November, and you told me that the GOP had only picked up two or three seats, I’d be quite surprised, but not completely shocked.  I would have a pretty good idea what had happened.  Likewise, if you told me the GOP had picked up 11 seats, I would be quite surprised, but not completely shocked.  There remains a wide range of possibilities, and many of these races are very close.

Still, the picture has clarified somewhat since then.  In particular, the House polling has done what I suggested it might do: It has converged on the fundamentals, with Republicans opening up a four-point lead.

Some of this is driven by certain pollsters switching on their likely-voter screens (something I alluded to in my piece), but not all of it (we only had three registered-voter polls in August to begin with). But note in particular how the Republicans have opened up their lead.  They jumped from 40 percent to 46 percent in the polling averages.  Democrats did not move at all.  Instead, they remained at around 43 percent, just a couple of points above the president’s national job approval.

What about the Senate?  In fact, we have seen a similar convergence of the polls on the fundamentals.  It has just been quieter.  Consider the following chart, which shows the movement in the competitive Senate races from Sept. 1 through Sept. 24 (Kansas is not included since it was just appearing on the radar screen in early September).

Of the 11 races that RCP tracks, seven have moved toward Republicans.  The average movement is 1.2 points, with a median of 0.8 points.  This isn’t as pronounced a shift as we’ve seen in the congressional races, but it is real.

More importantly is how these races have shifted, especially those races that have moved toward Democrats.  This next chart is a bit dense; I’ll explain after presenting it:

The first column shows the highest vote share that the Democratic candidate has ever recorded in the RCP Averages.  There isn’t a single race where that candidate is currently at his or her peak.  Moreover, only in Minnesota and New Hampshire has that peak been higher than 47 percent.

The next columns show the share of the vote the Democratic candidate was winning as of Sept. 1, and where he or she is today. In general, these candidates have been losing vote share or staying in place since the beginning of the month. There are two exceptions: Arkansas -- where Mark Pryor’s improvement is almost entirely an artifact of a single poll showing him at 47 percent -- and North Carolina.

The final four columns are the key ones. They show the vote share that my “fundamentals” analysis (developed this past winter) suggested Democrats should get given recent polling of the president’s job approval, how far they were from that share in early September, how far they are today and, most importantly, whether they have moved toward the projected vote share or away from it (I do not include Georgia because we haven’t had good polling of presidential approval there).

This helps explain what we’ve seen in North Carolina. As I noted in my last piece, Kay Hagan was the one candidate who was running behind the vote share that the fundamentals suggested she should receive; she is moving upward, but still regressing to the mean.  Note that the main reason she has opened up a lead, however, is that Thom Tillis has seen his numbers plummet in the wake of an advertising blitz, from receiving 45 percent of the vote in mid-August to receiving 40.6 percent today.

So the big question is: Will the Democrats remain range-bound? More importantly, can Republicans really expect the undecided voters to break heavily their way?

This is a question that we can’t answer with certainty. But there is at least some good evidence to support a claim that things will continue to converge on the fundamentals.

A recent piece from Dan McLaughlin at RedState was largely overlooked, but it is hugely important.  McLaughlin went through the polls from RCP and constructed averages for mid- to late-September going back to 2002.  What he found was a substantial amount of uncertainty between what polls showed in mid-September and the actual result. Yet there was order within that uncertainty. Seven races flipped leads in 2012, all toward Democrats; two flipped in 2010, one for each party; three flipped in 2008, all toward Democrats; three flipped in 2006, all toward Democrats; four flipped in 2004, all toward Republicans; and three flipped in 2002, all toward Republicans.

McLaughlin notes, correctly, that in these years, the races broke almost uniformly toward the “wave” party.  But his point is merely that there is a bunch of uncertainty built into mid-September polling.  Our question is a bit different.  After all, we don’t yet know who the wave party is for 2014, if in fact there will be one.  Let’s dig a bit deeper.

If we look closely at the data, there seems to be a “break point” in races where a candidates’ lead is five points or less.  There are candidates who have led by more than that at this point but who went on to lose -- Max Cleland in 2002, for one -- but they are few and far between. On the other hand, candidates who have led by five points (rounded) or less have won only 15 of the 33 races in which they have been involved -- about a 50-50 proposition. Of the 10 races listed above, only Minnesota shows a larger lead than five points.

But take a look at the following chart.  The x-axis (horizontal) shows how far above or below the president’s job approval the candidate of the president’s party was in September, while the y-axis shows how many points the candidate of the president’s party improved between the September average and Election Day.


The data are noisy, but the relationship is statistically significant: The farther behind the president’s job approval a candidate is, the more he or she improves up to Election Day. Likewise, if a candidate is above the president’s job approval, the tendency is that he or she finds his or her vote share is stagnant.

This makes sense.  In 2002 and 2004, Republicans were able to pick up a large share of undecided voters because the president was generally popular; they converged on his job approval number in the close states.

 In 2006, however, Republican vote shares hardly improved from their September showing. Tom Kean was leading Bob Menendez, 45 percent to 41 percent; he received 44.3 percent of the vote. Mike DeWine trailed Sherrod Brown, 47.7 percent to 44 percent; he received 43.8 percent of the vote. George Allen led Jim Webb, 48 percent to 43.3 percent of the vote; he received 49.2 percent of the vote.  The president’s job approval in New Jersey, Ohio and Virginia (respectively): 35 percent, 41 percent, and 45 percent.  The one candidate who saw major movement in his direction, coming from behind to nearly win a race where he trailed badly: Conrad Burns, who hailed from a state (Montana) where the president was still fairly popular.

We see much the same thing in 2008 -- Republicans failing to improve their vote shares while their challengers moved ahead of them -- while in 2012 Democrats generally surged to meet the president’s job approval in their state. A classic example is the Massachusetts Senate race, where Scott Brown led 43.4 percent to 42.6 percent at this point; he lost, 53.7 percent to 46.2 percent, as Elizabeth Warren moved toward the president’s 63 percent approval rating and won over almost all of the undecided voters.

Taken together, this tells a story that sticks together reasonably well.  From 2002 to 2012, candidates of the president’s party have tended to converge on the president’s job approval.  It isn’t an absolute tendency, but it is nevertheless real.

The Democrats’ problem is that they seemingly find themselves in a position similar to that of Republicans in 2006: They are in tight races.  But so far, they seem unable to move past where the fundamentals suggest they should be able to go: Recall again that their maximum showing has generally been bounded at 47 percent. 

This is true even though Democrats have generally dominated the air wars recently.  They’ve succeeded in driving down Republican numbers, or holding them in check.  But they haven’t improved their own. Here, the Peters-Land race in Michigan is instructive.  In early September, Gary Peters’ lead was 3.8 points.  Today it is 5.4 points.   But his numbers are largely unmoved: 45.3 percent at the beginning of the month, 44.3 percent today.  Terri Lynn Land’s numbers, however, have tanked: from 41.5 percent to 38.9 percent. 

But we know that this race will not have a 44.3 percent-38.9 percent outcome -- sooner or later the undecided voters will begin to decide.  And given that the Democrats are winning the votes of almost everyone who approves of the president’s job, they will have an uphill -- though hardly insurmountable -- battle with undecided voters. 

If this theory is right, we should expect to see these races continue on the basic trajectory we’ve seen over the past few weeks: Democrats holding at their current levels.  Eventually, Republicans should begin or continue to improve, as undecided voters engage and make up their minds, and as Republicans narrow the spending battles.  Even if this theory is true, it won’t occur in every race, but it will be the general tendency.

Again, I want to emphasize that this is only a theory.  As I said, I wouldn’t be shocked to wake up after Election Day and find that Republicans made minimal gains.  But I do think it offers a pretty good way of synthesizing the available data, and splitting the horns of the “fundamentals vs. polling” dilemma. Under this theory it isn’t really a dilemma any more than it was in 2006.  We just need time to see it resolve.

Sean Trende is senior elections analyst for RealClearPolitics. He is a co-author of the 2014 Almanac of American Politics and author of The Lost Majority. He can be reached at Follow him on Twitter @SeanTrende.

Senate May Confirm Up To 88 Federal Judges
Alan Fram · December 16, 2014
A Reply to Rush Limbaugh
Mona Charen · December 9, 2014
Against the Wave: Three Winning House Democrats
David Byler · December 14, 2014
Obama, Congressional Dems Show Cracks in Unity
Jim Kuhnhenn · December 7, 2014

Sean Trende

Author Archive

Follow Real Clear Politics

Latest On Twitter