Disproving the Spoiler Effect

Conventional wisdom says that a third party candidate will hurt the most similar major party candidate by splitting their votes, leading to the possibility that the least favored candidate will win the election (the so-called “spoiler effect”).  Even election experts like Professor Rick Hasen and the staff at Fair Vote assert that third parties may “spoil” elections in precisely this way.  Last week, we questioned that premise, presenting several reasons why the “spoiler effect” deserves greater scrutiny.  Here, we go further to demonstrate that conventional wisdom is backwards – significant third party candidates help the major party candidate closest to them in a three candidate race by splitting the vote with the more dissimilar candidate.

The best evidence of this comes from recent elections.  There have only been three Senate elections going back to 2000 in which there were a Republican, a Democrat, and one prominent challenger clearly associated more closely with one of the major parties than the other (due to his or her membership in it).  In all three, the interloping challenger appears to split the votes with the more distant, unassociated major party – the opposite of what Hasen and Fair Vote say you should expect.

ALASKA 2010

Republican Senator Lisa Murkowski lost the Republican primary to Joe Miller, who was endorsed by the Tea Party.  Murkowski decided to run a write-in campaign, and became the first write-in candidate to win a Senate election in over fifty years.  Murkowski received 36% of the vote, slightly better than Miller’s 35%.  The Democratic candidate Scott McAdams received only 23% of the vote.

CANDIDATE	PARTY	VOTE SHARE
Lisa Murkowski	Write-in	36.08
Joe Miller	Republican	35.27
Scott McAdams	Democrat	23.32

Comparing those numbers to the next closest Senate elections in Alaska illustrates how badly the extra conservative candidate hurt the liberal candidate.  In the 2008 Senate election, the Democrat beat the Republican 48% to 47%.  In 2004, when Murkowski was the official Republican candidate, she won, but the Democrat still received 46% of the vote.  So, when two Republicans ran for Senate against one Democrat, the Democrat did the worst and only about half as well as when competing against one Republican.

FLORIDA 2010

There was no incumbent in this race.  Marco Rubio won the Republican primary; Kendrick Meek won the Democratic primary.  Republican Governor Charlie Crist announced his intention to run for the seat, but dropped out of the Republican primary when it was clear he stood no chance of defeating Rubio.  Here are the final results from the election, which Rubio won easily.

CANDIDATE	PARTY	VOTE SHARE
Marco Rubio	Republican	48.89
Charlie Crist	Independent	29.71
Kendrick Meek	Democrat	20.20

Like the Alaska election, two candidates associated with the Republican Party ran against one Democrat.  And just like the Alaska election, the Democrat got trounced in defiance of the “spoiler effect.”  In the 2006 Senate election, the Democrat won with 60.3% of the vote.  In 2004, the Democrat received 48.3% in a losing effort.

CONNECTICUT 2006

When Joe Lieberman lost his primary in 2006, he chose to run anyway – and won.  His primary opponent Ned Lamont finished second in the race.  The Republican candidate, Alan Schlesinger, finished third with less than 10% of the votes.

CANDIDATE	PARTY	VOTE SHARE
Joe Lieberman	CT for Lieberman	49.71
Ned Lamont	Democrat	39.73
Alan Schlesinger	Republican	9.62

In the Senate elections before and after 2006, the Republican candidate received 34.17% (2000), 32.13% (2004), and 43.22% (2010).  There was no major third party or independent challenger competing in any of them.

That’s only three elections.  But since third party challengers are so rare in the United States, they are by far the best examples of how third party challengers affect election results.  Each contains a fairly large electorate (statewide), clear distinctions between the similar candidates and the dissimilar candidate (as opposed to the 2008 Senate race in Minnesota), and a sufficiently large deviation from the norm to confidently attribute it to the third candidate’s participation (as opposed to the 2000 presidential election).  That each emphatically defies the “spoiler effect” shows that (a) we probably should discard it, and (b) we probably use it politically rather than descriptively.  The latter is essentially the point of our last post – a point that becomes much stronger when you realize that the “experts” exclude such prominent counter-examples.