Midterm Madness: Predictions for Tuesday

Although Chet Edwards (D-TX) has the backing of the NRA and remains highly popular in his district, voter disgust with Washington will likely sweep him from office.By Kirk Jing

It is possible a few people came in hopes of hearing eerily reminiscent assertions (to 2006), that the Democratic House would be safe come November. But Midterm Madness 2010, a discussion sponsored last Friday by the Rockefeller Center and Government Department, was no support group for wounded Democrats. Anyone who came in hopes that their preexisting political narratives would be ‘vindicated’ in an exercise of group verbal self-indulgence probably left disappointed. 

To a small extent, the turnout was symbolic. The room was filled with a smattering of presumed Government majors, people snapping pictures for possible future use in a newspaper, a relatively large contingent of non-students, and perhaps another smattering of people interested in election math.  

The panel wasted no time. Instead of issuing vapid talking points or empty prognostications, the panel jumped straight into the hard numbers of polling and elections.

Charles Franklin, a political science professor from the University of Wisconsin-Madison and cofounder of Pollster.com, spoke first. Franklin’s background is in the statistical analysis of elections, so many in the audience were likely braced for a blast of math coming their way. Instead, Franklin actually vowed to keep his discussion on more qualitative analysis, and asserted that highly negative national factors for candidates, such as economic stagnation and disillusionment with President Obama, could override positive local factors for them, such as being strong or popular candidates. 

He fittingly demonstrated this with the rumble in his own backyard: the Wisconsin Senate Race. He noted that Ron Johnson, the Republican and political neophyte who had jumped in to run a mere week before the Republican state convention, and incumbent senator Russ Feingold were in a statistical tie even when the vast majority of the electorate didn’t know Johnson well enough to express any kind of opinion on him.

Without any witchcraft-related surprises to seriously sway voter opinion, Johnson eventually surged to a solid lead in most polls as he gained more name recognition, despite running a campaign much like Tam Michaels, the Republican whom Feingold roundly trounced in 2004. Franklin then noted that those same polls showed that Feingold was still liked by the majority of the electorate, or held “net favorables,” showing that even some people who liked Feingold were voting against him because of discontent with the national Democratic Party. 

Incumbents in the United States Senate are renowned for their electoral resilience, and Feingold is a reasonably popular incumbent with several policy overtures, mostly social, aimed towards centrists in a lean-Democratic state. (I’m uncomfortably reminded here of Feingold’s farcical attempt to have the federal government guarantee Wisconsite access to Packers games.)  

With a mathematical flair that could only be honed from adventures in Government 10 statistics, Professor Joseph Bafumi, one of Dartmouth’s own, came up next, treating the audience to its first delightfully math-heavy analysis of the night. He presented a model for using generic ballot results (a poll about whether respondents would vote for a generic Republican or generic Democrat) to predict the actual election results.

Bafumi used election data and generic polling from 2006 to help formulate separate equations for both incumbent seats and open seats, to translate aggregated generic polling results into predicted two-party vote totals in contested districts. The hard numbers and mathematical techniques from which the model was derived remained obscure, however, due to how infuriatingly small the charts on the Powerpoint were. 

Possibly for audience appeal, Bafumi noted his opinion on Senate control, which happened to be similar to Franklin’s prediction — a 51/49 Senate — before diving into the crux of his presentation. Running the model 1000 times to help account for uncertainty, Bafumi found that the Republicans took control of the House 79% of the time, picking up an average of 51 seats. Although 51 seats far-surpassed the magic 39 required for the GOP to take control of the House, a glance at the graph Bafumi had displayed on the screen while explaining his points would have revealed a huge clump of Democrat incumbents that would stay safe if the Republican preference on the generic ballot underperformed. As ominous as the model was for Democrats, a Democrat with an alternative interpretation could find solace in what appeared a remarkably deep Democratic firewall that might theoretically just barely stop the GOP wave. 

Tossing out another fun fact about the model, Bafumi mentioned that the Democrats would have to hold the GOP to a 51-49% lead in the generic ballot in order to have an even chance of holding onto to the House, a somewhat tall order when most of the panelists signaled their belief that the true generic ballot lead for the GOP is between 6% and 9%. 

The election of 1946 was hailed as the “greatest Republican victory since Appomattox” after the GOP gained 55 seats. But Harry Enten ‘11, known for both his savvy with polling and unique 90’s-inspired (and thus inherently laudable) approach to weathertainment, gave the most audacious House prediction of the night, predicting the Republicans would pick up 57 seats in the House, far exceeding the 39 required to take control of the chamber. 

He felt that such a gain was within the range corroborated by statistical evidence, but that most political handicappers were hesitant to predict numbers these high, due to lack of much historical precedence. After all, not many pollsters would want to put their professional reputations on the line in order to seize the honor of predicting Appomattox 3.0. Enten, however, also volunteered a more conservative estimate on the Senate than the others, predicting the Democrats would retain 52 or 53 seats in the upper chamber.

Interestingly enough, though, the event of a party taking control of the House of Representatives but not the Senate, as both all the panelists who ventured predictions and most political handicappers admit is the most likely, also has little historical precedent, happening last in 1930 when Democrats took the House by one seat. 

But of course, like most of the panelists, Enten’s presentation focused on a topic less often debated upon: the predictive value of polls dating back to June. In his presentation, he demonstrated that there had been remarkably little movement in the generic ballot since June. Aggregating generic ballot polls, he found that Republicans had only gained 1.29% among registered voters, and an even more miniscule .44% among likely voters. Looking at Senate races, although some races had fallen off the map for certain reasons (such as earlier mentioned, people actually learning who Ron Johnson is), for the most party predicted net party gains had also barely moved. 

After Enten, Ted Schroeder ‘11 took a surprising detour. Instead of focusing on midterm elections themselves, Schroeder presented about a study he had worked on with the moderator, Dartmouth professor Dean Lacy, over the summer surrounding the special elections that precipitate these midterm elections. And although probably everyone in the room came with knowledge about Scott Brown’s famous special election victory, the study was on a huge span of special elections spread across various decades. 

Robert Erikson, a political science professor from Columbia University, spoke last. His presentation focused on quantitatively tracing election trends. Interestingly enough, in a partial parallel with Franklin’s first presentation, Erikson’s data showed that Democrats perform considerably better on the local level than they do on the national level, underscoring the overarching dominance the national atmosphere will have on this election cycle and showing that even locally strong Democratic candidates could be given a “thumping” by locally weak Republican candidates simply because of the national atmosphere. 

Erikson compared the election trends across the 2006 and 2008 cycles, and his charts showed that while Democrats started 2004 with a slim lead over Republicans, it slowly turned into a huge lead. It also showed that what happened to the Republicans was very different. Republicans started off 2008 with a similarly huge deficit in voter support, but very quickly surged to a narrow lead by the middle of the session in 2009, where they have essentially remained since now. It was also very convenient for audience members to see a visual representation of how consistent polling has been for the last few months, as mentioned earlier in Enten’s presentation. 

Enten also noted that, interestingly enough, 2010 is unique in that although the 1994 Republican Revolution was a huge shock, and most pollsters were hesitant to even suggest a Democrat takeover in 2006, there has been so far a unique consensus among pollsters of a Republican takeover of the House in 2010.

A variety of questions came up as soon as their presentations were done. The possibility that had been generating hope in the liberal blogosphere – that the polls are off because many exclude cell phones – was raised. That hope was unanimously shredded by the panel in a rather lengthy discussion. They noted that cell-phone bias has been an issue for a decade and that pollsters have been working to compensate for it ever since because of the expense of calling cell phones. It was also noted that many robocall pollsters, despite being prohibited by law to call cell phones, were stunningly accurate: the best example being the 2009 New Jersey gubernatorial elections, when most robocalls predicted a Christie victory and most person-to-person polls incorrectly predicted Corzine would hold on. 

Several more topics came up, such as the effect of weather on elections (the panel pointed to the surprisingly close 2006 Senate election in Tennessee as an example), the downballot effect of gubernatorial elections on federal elections (mostly notably the strange situation in Colorado), and the seeming inability of Democrats to forge either a message or competent campaign. Perhaps many Democrats heard their sentiment argued when a panelist mentioned in passing, “A party too stupid to campaign…doesn’t deserve to be in power.”