Monday, November 5, 2012

Prediction of the Presidential Election

Here is my best guess as to what will happen tomorrow in the presidential race.

Popular vote:  Obama squeaks by with 50 percent to Romney's 49 percent, and 1 percent for others. 

Electoral vote:  Obama wins 281-257:

The popular vote should come as no surprise. A very narrow Obama victory is about what most political science models predicted months ago. Outcomes usually hinge on the economy, and though growth remains sluggish, it is just barely great enough to push the president over the finish line. In recent days, moreover, most national polls have shown a shift in the president's direction. Pew gives a slightly larger margin to Obama, but as Nate Silver explained a few months ago, Pew has the largest "house effect" of any of the major polls (in this case, 3.2 percent in favor of Democrats). 

The electoral vote prediction relies on polls and other public sources of information. The campaigns may have sources that are telling them something different, but we should discount what they say in public about what they're seeing in private. (And place no stock in the crowds the candidates are drawing.  In 1972, George McGovern got huge crowds at the end -- and then he lost 49 states.) My map generally follows the RealClearPolitics "No Tossup" map, with two exceptions.  In Virginia, the president has a one percent lead in the latest poll.  But that poll is from Marist, which has a 2 percent Democratic house effect, so I'm reckoning that Romney is actually ahead by a point.  In Colorado, the polls put Obama slightly in front, but Republicans seem to have an edge in the early vote, so I give it to Romney.

Here are some big caveats:

  • Sampling error: even a flawless poll can calculate a percentage only within a certain range, or confidence internal.  (News reports often call it the "margin of error.")  That is, the actual figure may vary by a couple of points either way:  see here for a thorough explanation.  
  • Other survey errors:  "Unfortunately, there are several other possible sources of error in all polls or surveys that are probably more serious than theoretical calculations of sampling error," the Harris survey has said. "They include refusals to be interviewed (non-response), question wording and question order, interviewer bias, weighting by demographic control data, and screening (e.g., for likely voters). It is difficult or impossible to quantify the errors that may result from these factors."  The increasing use of cellular phones has  made it tougher to get a representative sample. It is possible, though unlikely, that such errors are consistently biasing polls in favor of one candidate or the other.
  • Undecideds and switchers:  A few voters are still undecided and some "decided" voters could switch. "We still have 11 percent of the sample saying 'we could possibly change our mind,'" says Pew president Andrew Kohut.  "This is our projection, and our projections have been pretty good, but there's always the possibility things could change."
  • Sandy:  A previous posts have indicated, the aftermath of Hurricane Sandy could depress turnout in the heavily-Democratic New York metropolitan area, cutting into President Obama's national popular vote total.
The most likely outcome is that Obama wins the popular and electoral vote.  But because of the uncertainties above, other outcomes are within the range of possibility:

  • Romney wins:  a late surge in Pennsylvania could give Romney a 277-261 victory.  The latest Ohio poll is within the margin, so it is just possible that Romney could win there, too.
  • Obama wins by a bigger margin:  victories in Virginia, Colorado, and Florida could make it 337-201 in Obama's favor. 
  • A popular-electoral vote split: as noted, Sandy could tip the popular vote to Romney without switching any states.  Obama would have the presidency, but Republicans would have a talking point.
  • A tie: given trends in key states, the possibility of a tie in the electoral college seems pretty remote -- but it is not quite at zero.