By Warren D. Smith, April 2008. Please send me improvements or corrections.
We look at all LA governor races from 1975 to now. (The cutoff date 1975 was selected because that seems to have been the first time they employed the runoff system. As far as I know, LA is the only US state to elect governors with such a runoff.) We shall examine each race with the benefit of hindsight to see what pathologies happened. In some cases we will not have enough information to be certain a pathology occurred and then we shall have some varying degree of certainty.
|Won by landslide majority hence no runoff needed||2007, 1999, 1983, 1975|
|No runoff since candidate withdrew||1987|
|(Hence were 4 races in which a runoff happened. In every one, turnout was higher in the runoff than in the first round, contradicting the claim commonly made by FairVote that runoffs have low turnout.)|
|Probably favorite-betrayal pathologies||2003, 1995, 1991, 1979|
|Non-monotonic||1991 (and 1995? and 2003?)|
|Beats-all-winner loses||1995, 1991 (and 1979?)|
|Exhibiting "no-show paradox"||2003, 1991 (and 1995? and 1979?)|
To summarize this even more concisely (counting uncertain cases fractionally via a somewhat-subjective probability estimate – the figure is the percentage of races in which that phenomenon occurred):
|What happened||Experimental data||Theor. model|
|Landslide majority win||44%||0|
|No runoff due to withdraw||11%||(model not applicable)|
|"No show" paradox||40%||16%|
|Beats-all winner loses pathology||28%||4%|
where I use as the "theor.model" the random election model for 3-candidate Instant Runoff races. As you can see, the theor. model in spite of its crudity didn't do too badly. Its main errors are the first and last predictions. The last one is due to the fact 1D and 2D politics models are better than random election model for this purpose (at least in my experience). The first one is due to fact the Dirichlet model is clearly better than the random elections model for that purpose (since it predicts a nonzero value!).
Nice math puzzle: The Dirichlet model predicts (in a 3-candidate race) that a majority-top winner exists, hence no need for runoff, with probability=56.25%.
2007: Republican Bobby Jindal wins with 54%; closest opponent has 17%.
2003: First Round:
|Kathleen B. Blanco(D)||18.36%|
This is an example of success of Louisiana's runoff system in the sense that Blanco won the runoff despite a big loss in the plurality election. (But she became very unpopular during her term due to perceived poor handling of the Hurricane Katrina disaster and declined to run for re-election.) Most of the voters for the 3rd,4th, etc candidates in round 1, evidently went for Blanco, not too surprisingly since the other high finishers all were in her party. (Also note the turnout was higher in the runoff round than the first round, contrary to IRV-advocates' common assertion runoff turnout is lower. Same thing happened in 1991 and 1995 and 1979, i.e. it happened every time it possibly could.)
Runoff however may have exhibited some pathologies. For example, suppose in round 1 that 2% of the electorate (6% of Jindal's voters) had decided to strategically vote for Leyoub instead of Jindal. That would have made the runoff be Leyoub vs. Jindal, which probably would have resulted in Jindal's victory unlike in the Blanco vs. Jindal runoff which actually happened. [Can anybody find poll data on that question?] That's somewhat speculative, but if so, it would have been an example of dishonesty paying.
Indeed, quite possibly that would have worked even if those Jindal voters had defected to Leyoub in both the 1st and 2nd rounds. That's more speculative, but if so, this race was an example of non-monotonicity – voting against Jindal causes Jindal victory.
There's another pathology which definitely happened – favorite betrayal. If in an election, some subset of voters would have gotten a better result (in their view) by not voting for their favorite top, then we here call that a "favorite betrayal scenario." (Also, in every such scenario on this page, voting their favorite below top was the only way that voter subset could get a better election result.) Undoubtably at least 6% of Jindal's voters (i.e. 2% of the total electorate) preferred Leyoub over Blanco. (In the USA, it is extremely rare for an election result between two near-equal candidates to be biased 94-6 or more.) They would have been better off betraying their favorite Jindal by voting Leyoub in round 1 (or, for that matter, in both rounds). This would have caused either Jindal or Leyoub to win, either way (in their view) an improvement. Also, there seems little or no question at least 14% of Jindal's voters must have preferred Leach over Blanco, hence this race also exhibited a second "favorite betrayal" pathology.
Finally, there was a "no show paradox." Suppose an extra contingent (equivalent to 12.2% of the electorate) of voters had magically appeared, all ranking Hunt Downer top and Jindal dead last. Then it would have been a Jindal vs Downer runoff, which Jindal would have easily won. Thus these extra voters, by ranking Jindal last, cause him to win. They would be better off not voting (as in fact, they didn't).
1999: Mike Foster wins with 62.17%, closest opponent (Bill Jefferson) has 29.53%.
1995: First Round:
|Arthur D. "Jim" Nichols(I)||16616||1%|
Runoff round: Foster wins over Fields 64-to-36 percent.
At naive first glance, the magnitude of Foster's final-round victory makes you think this race could not have exhibited any pathologies. But a closer look leaves me almost-convinced of the opposite!
Only 20,000 votes would have put either (former governor) Roemer or (congresswoman, state treasurer, and soon to be senator) Landrieu into the runoff instead of Fields. Both had led in the polls until nearly election time.
Fields was black (indeed, the first black candidate for LA governor in over 100 years) and felt to have no chance of election against the white old-wealth multi-millionaire Foster in a state which had last election put Neo-Nazi Ku Klux Klan Wizard David Duke into the runoff. (As part of this 1995 campaign, Foster paid more than $150,000 to David Duke for his mailing list of supporters. Duke also endorsed Foster. After failing to report the purchase as a campaign expenditure, Foster became the first Louisiana governor to admit and pay a fine for violating the state's ethics code... a distinction which of course set him up for a landslide re-election win. Fields' support was very narrow; exit polls indicated he got all but 2% of his votes from blacks in the first round.) Hence many Democrats felt Fields running at all was therefore a stupid act, strategically speaking (and they were proven right in the end) but naturally this view was not popular with Fields supporters.
Fields wasn't 100% dead since a Mason-Dixon poll on October 17 found that Louisianans thought (by 75%-to-18% margin) "a black" could theoretically win – and they even thought that Fields probably would win (by 45%-to-21%, with 34% undecided, said a 27 Feb. Mason-Dixon poll) – provided his opponent was David Duke! Against pretty much anybody else, though, Fields was going to lose a head-to-head race (see polls below) by a huge margin.
Fields also was involved in a mini-scandal when he abused his congressional "franking" privileges to bill taxpayers $46,000 for sending out gubernatorial-campaign literature. In 1997 the FBI videotaped him stuffing about $20,000 in cash from Edwin W. Edwards into his pockets, but that happened after this race ended and hence did not affect it.
Former Democrat governor Roemer now ran as a conservative Republican with an anti-crime anti-welfare platform. Foster switched from Democrat to Republican during the race and spent $2 million of his own money to move slightly ahead of Roemer in the Republican vote.
But it seemed likely that either white (Roemer or Landrieu) would have beaten Foster in the runoff since either was more centrist. The poll data (below) support that speculation for Roemer, but not for Landrieu. So if just 9000 Fields voters had voted Roemer instead (or if just 5000 had voted Landrieu instead?) they would have gotten a winner they preferred.
Thus this race probably exhibited one (and perhaps two) favorite betrayal pathologies (and no-show paradoxes too – since those Fields voters would have been better off either "staying home" or voting for Roemer; ditto many Landrieu voters). It also may have been non-monotonic: if about 20,000 Fields and 10,000 Landrieu voters had voted for Foster, that would have put Roemer into the runoff, whereupon he probably would have beaten Foster (perhaps even if these defecting voters still voted Foster even in the runoff, although that is more speculative).
Fields was definitely a lose-to-all loser (among the top four) – and probably Roemer was a beats-all Condorcet winner. Louisianans were asked in several Mason-Dixon telephone polls how they would vote in hypothetical 2-person runoffs. The results were:
|Hypothetical Runoff||Poll result %s||Poll date||Sample|
|Roemer v. Fields||Roemer 54, Fields 27||17 Oct 1995||826|
|Roemer v. Fields||Roemer 55, Fields 26||25 Sep 1995||827|
|Roemer v. Fields||Roemer 61, Fields 22||28 Aug 1995||828|
|Roemer v. Fields||Roemer 59, Fields 24||8 July 1995||826|
|Roemer v. Fields||Roemer 58, Fields 26||8 May 1995||818|
|Landrieu v. Fields||Landrieu 51, Fields 21||17 Oct 1995||826|
|Roemer v. Landrieu||Roemer 43, Landrieu 42||17 Oct 1995||826|
|Roemer v. Landrieu||Roemer 43, Landrieu 42||25 Sep 1995||827|
|Roemer v. Landrieu||Roemer 45, Landrieu 40||28 Aug 1995||828|
|Roemer v. Landrieu||Roemer 43, Landrieu 38||8 July 1995||826|
|Roemer v. Landrieu||Roemer 45, Landrieu 41||8 May 1995||818|
|Foster v. Landrieu||Foster 42, Landrieu 39||17 Oct 1995||826|
|Foster v. Fields||Foster 53, Fields 24||17 Oct 1995||826|
|Roemer v. Foster||Roemer 40, Foster 44||17 Oct 1995||826|
|Roemer v. Foster||Roemer 47, Foster 32||25 Sep 1995||827|
|Roemer v. Foster||Roemer 43.5, Foster 38||(avg of both above)||1653|
1991: The famous "Lizard versus Wizard" race replete with pathologies. It was so utterly insane that it has a web page of its own!
1987: First round:
|Billy Tauzin(D [became R in 1995])||10%|
|J.H. "Jim" Brown(D)||9%|
Runoff round: did not happen since Edwards withdrew. (This race was the only time Edwards finished other than first in an election.) This withdrawal was generally believed to have been an intentional strategic masterstroke designed to allow Edwards to come back and win the next governor race! (Wikipedia: "[Edwards] was cleverly setting a trap for Roemer. By withdrawing, Edwards denied Roemer the opportunity to build a governing coalition in the general election race, and denied him the decisive majority victory that he surely would have attained. In one stroke, Edwards made Buddy Roemer a minority governor.") Both Roemer & Edwards were from the same party (Democrat) at the time, although Roemer and Tauzin later switched to the Republican Party. It was felt by many (including Edwards) that to win this race, he had to go against Livingston in the runoff. Edwards talked up Livingston during the campaign but it did not work.
Based on all this craziness, one might think pathologies must have been present in this race! But I do not see any! (I suppose one could argue that maybe some Livingston supporters intentionally-dishonestly voted Roemer because they thought Livingston could not win the runoff. The Republican Livingston later resigned from the US House [even though speaker-elect at the time] due to a sex scandal [in 1999].)
1983: Edwards wins with 62% of the vote in round 1, closest opponent (Treen) has 36%.
1979: First Round:
|David C. Treen(R)||297674|
|Ed.G."Sonny" Mouton Jr.(D)||124333|
|Luther D. Knox(D)||6327|
|Greg Nelson (no party)||4783|
Runoff round: Treen wins by 690691 to Lambert's 681134.
Story behind this race: After 102 straight years of Democrats as Louisiana governors (the last Republican was the Vermont "carpetbagger" W.P.Kellog in 1873-1877), the Republican Treen was elected. Treen, a 51-year-old lawyer, had begun his political career as a segregationist and counted on white votes to propel him to victory, whereas his Democratic opponent Lambert hoped for Black votes and was endorsed by organized Labor. Treen had pledged to keep misleadingly-titled "right to work" laws outlawing unionized shops, thus drawing the ire of labor unions. Treen had a reputation for personal honesty and (amazingly for a Louisiana politician at that time) disclosed his personal finances. That contrasted well with the fact governor E.W.Edwards and Lt.Gov. J.Fitzmorris were both involved in an FBI bribery investigation – but this was not widely known until after the election was over and hence probably did not affect it. [Edwards explained to the press that he had "played games" with an FBI undercover informant trying to bribe him after "it became obvious the conversation was being recorded." Later he explained how $10,000 from Korean businessman Tongsun Park was not a gift to him but rather to his wife. After the election, E.G.Mouton Jr. (who was named "executive counsel" by Treen and had received an unrecorded $15,000 contribution) also was subpoenaed by the grand jury, as was Treen's Lt.Gov. Robert E. Freeman.] There were dozens of major errors found when rechecking voting machines in the first round. Fitzmorris initially led Lambert but a retabulation put Lambert ahead of Fitzmorris by about 2506 votes.
Fitzmorris sued, alleging Lambert had benefited from fraudulent votes and improper counting procedures (thousands of alleged fraudulent acts and irregularities were listed in the suit; in a few days Fitzmorris magically lost about 2000 votes while Lambert gained about 2000). The court "sympathized" with Fitzmorris but refused to let him in the runoff because he'd only clearly proven a few hundred votes bogus, which wasn't enough. That all stimulated Treen to play it safe by hiring guards and 1000s of volunteers to help guard and observe the machines and ballots in the runoff round. That worked when Treen won by a tiny 9557-vote margin out of 1.4 million votes cast (50.3% to 49.7%) over Lambert. About $20M was spent campaigning for this $50K/year job. Treen became unpopular during his term which is why he was resoundingly defeated by Edwards next election.
Treen was the only Republican in the race. All other major contenders were Democrats. Fitzmorris, Hardy, Henry, and Mouton all crossed party lines to endorse Treen, and Hardy later switched to the Republican party (and became Lt.Gov. 1988-1992; also Treen once elected appointed Fitzmorris as his "special assistant for industrial development"). Lambert blamed his defeat mainly on Fitzmorris' suit: "If it hadn't been for that lawsuit, I would have beat him 55 percent to 45 percent."
Probable pathologies: If just a few thousand of Lambert's supporters had switched their votes to Fitzmorris (or if there really was a fraud and that fraud were undone), then, considering how close Treen's victory was even over the scandalized Lambert, Fitzmorris would probably have defeated Treen. (Wikipedia: "Treen realized that he was fortunate in that Fitzmorris may well have been a stronger opponent than Lambert.") Indeed, if enough Lambert voters had switched their votes to either Fitzmorris or Hardy to get them into the runoff, then they too probably would have beaten Treen, which those voters would have preferred. (There is little doubt that at least 60,000 Lambert voters existed who preferred Hardy over Treen, and certainly at least 2,000 existed who preferred Fitzmorris over Treen.)
In other words this election (probably) exhibited a favorite betrayal pathology (in fact probably several).
It also exhibited no-show paradoxes, e.g. if sufficiently many Lambert-top Treen-bottom voters had "stayed home" then a candidate they preferred probably would have won – their act of voting honestly Treen bottom caused Treen to win. (If Fitzmorris would have beaten Treen, then only a few thousand would have needed to stay home, and almost certainly they could also have stayed home during the runoff round also.)
It also is plausible – although this is more speculative – that Fitzmorris or perhaps Hardy was actually the "beats-all" Condorcet winner (e.g. presumably Hardy, as a future Republican, would have been preferred by Treen supporters over Lambert, and also the usual thinking was that the Democrats would normally beat a Republican and Treen only beat Lambert due to the lawsuit and associated higher-than-usual scandal).
Luther Devine Knox, tired of getting few votes, had his name legally changed to "None Of The Above." However,the Secretary of State refused to let him on even as "L.D. 'None of the Above' Knox" and the legal name change came through too late (ruled a judge) to get him on the ballot under that name.1975: Edwin Edwards wins with 62.35% (closest rival has 24.29%).
Would be nice to add details, but... Wikipedia, State of Louisiana, Newspaper articles, and Polls.
Return to main page