Los Angeles Mayor races 1960-2009, especially 2001

By Warren D. Smith, April 2009. Please send me improvements or corrections. Historical table.

We look at all LA mayor races from 1960 to now. LA conducts mayor-elections via a 2-round runoff system essentially the same as that used to elect the governor of Louisiana.

As of 2009, Los Angeles is the largest US city, and Louisiana is the only US state, that use this 2-round system. In contrast, New York City (which as of 2007 was by far the largest US city) employs party primaries (which can involve a runoff, for example in 2001 Mark Green defeated Fernando Ferrer in a 2-man runoff to win the Democratic party nomination; the original Democratic primary had 5 candidates with Green & Ferrer finishing top-two) followed by a general election, and in that election at most one candidate from each party runs. In LA, there are no party primaries, any number of candidates from any party can run, and the top 2 finishers (who could be from either the same or different parties) compete in a 2-person runoff if necessary. Los Angeles was the 2nd largest US city as of 2007, with Chicago and Houston next. Chicago and Houston both appear to use an election system similar to LA's.

From the standpoint of a voting theorist, most of these races were not interesting, i.e. practically every voting method would have elected the same winner. But an exceptionally interesting election occurred in 2001. If our educated guesses shown in the table below are correct, Villaraigosa would have won with instant runoff voting (IRV, line 2), and been nearly-tied with Hahn with range (line 3) despite the fact that Hahn defeated Villaraigosa in the official head-to-head two-man later runoff election! (This is not a misprint.)

If that range-voting near-tie had broken Villaraigosa's way this would be the only real-world election known to us as of April 2009 in which there is a reasonable amount of evidence range voting would have "overruled a voter majority" because the minority expressed stronger preferences than the majority (range voting allows strength of preference information in votes). However, see also line 4. Plain plurality and IRV would also have elected Villaraigosa (but for different reasons than range) thus also overruling that voter majority. The reader may want to consider what she thinks about that and compare with our view.

The LA mayor races 1961-2009 – Quick Summary

1961-1969: Sam W. Yorty won the mayorship in 1961, 1965, and 1969. The 1961 win was a surprise defeat of incumbent Republican mayor (1953-1961) C. Norris Poulson. By 1969 Yorty had become unpopular with a scandal-ridden administration; 3 of his commissioners had been convicted on bribery charges, a 4th was awaiting trial; and Yorty, although initially popular with blacks, had failed as promised to repair widespread misconduct by the LA police department or to challenge police chief Parker. The black lawyer and former police lieutenant Thomas J. Bradley (endorsed by the LA Times) had been expected to win in 1969 based both on pre-election polls and the fact that he was ahead of Yorty by over 100,000 votes in the first election round (Bradley took 42% of the votes versus Yorty's 26%). But Yorty beat him in the second and final round 53-to-47% by using racial and militant/radical/communist slurs. Although these were quite unjustified, in the racially-charged atmosphere after the Watts riot of 1965, they were fearful enough to work. Yorty was a Democrat but often supported Republicans and in the 1970s become a Republican.

1973-1992: Thomas J. Bradley won the 1973 rematch with Yorty and then won re-election in 1977, 1981, 1985 and 1989. In all Bradley served 5 terms over a 20-year period. He was only the second African-American mayor of a major U.S. city. Long-duration reigns like Bradley's became impossible later because, e.g, Riordan was forbidden to run again after serving 2 terms, due to a term-limit law. Winning margins:

1977: 59% (over twice Robbins' vote),
1981: 64% (about twice Yorty's vote),
1985: 68%,
1989: 51.8%

Bradley was also famed in political-polling circles for the so-called Bradley effect. That concerned his failed 1982 run for California governor. Bradley was ahead in polls but lost the actual election to George Deukmejian(R) by 52,295 votes. Some theorize this discrepancy was due to racist voters who lied to the pollsters.

1993: Richard Riordan(R), a 63-year-old rich venture capitalist, won a 54-to-46% victory over Mike Woo(D, age 41) in the runoff on 8 June. The first round on April 20 had given Riordan and Woo 33% and 24% respectively followed by Joel Wachs (12%), Richard Katz (10%) and Linda Griego (8%). There were 24 candidates in all. For portraits of the most prominent, see Jane Fritsch's article in the New York Times Magazine 18 April 1993; these included 8 democrats including Woo but only 1 republican, Riordan. Bradley, age 75, did not run for the first time in 24 years. Woo was endorsed by the LA Times, Bill Clinton, and County Supervisor Gloria Molina and was widely expected to win the runoff. Riordan spent $6 million of estimated $100 million personal fortune to fund his campaign. Whites voted Riordan by over a 2:1 ratio, while the other racial groups supported Woo. The net result, according to polls, is Woo would have won with 100% turnout, but since the demographic groups supporting Riordan turned out in higher percentages – "Moderates" and "Conservatives" and voters earning over $20K/year all supported Riordan – he won. Riordan became the first Republican mayor in over thirty years.

LA Times: 9 June 1993 page 1 article "Exit Poll Finds Voters Sought Lesser of 2 Evils" by Richard Simon: Those who supported Woo said they were influenced by his reaching out to minorities and his early opposition to former Police Chief Daryl F. Gates. They also said they were turned off by charges that millionaire Riordan was trying to buy the election and to a far lesser degree by Riordan's alcohol-related arrests.
Riordan apparently struck a chord with his mailers and TV ads portraying Woo as an ineffective politician. Nearly half of Riordan's supporters cited Woo's council record as their reason for supporting Riordan.

The turnout in the first round was 32% of registered voters (NY Times 22 april p.16) and of those no candidate received more than 33%, which is why it went to a runoff. In the runoff round, Riordan won by (according to initial figures reported by NY Times 10 June 1993 page 1) 302786 to Woo's 259820. The turnout was 44% of registered voters. Note the turnout increased from 32% round 1 to 44% in round 2.

1997: Richard Riordan(R) won re-election with 61.5%, defeating Tom Hayden(D, 34.5%) and 3 other minor rivals.

2001: Complicated and interesting 15-candidate 2-round election ultimately won by James K. Hahn(D). [Riordan was not running due to term-limit law.] Will be discussed below.

2005: Councilman Antonio Villaraigosa (former speaker of CA assembly) defeated sitting mayor James K. Hahn, becoming the city's first Hispanic mayor since the 19th century. Two rounds needed.

General election (i.e. first round, held 8 March 2005):

Runoff (i.e. second and final round, held 17 May; note the turnout increased from 377,699 in round 1 to 493,084 in round 2):

Based on this data, the only reasonably-plausible way to contend Villaraigosa was a "wrong winner" in this election would be to postulate that Hertzberg was the "right winner." (Which would have been possible if, e.g, enough Hahn voters preferred RMH>AV.) Incidentally, the LA Times endorsed both AV and RMH in round #1. RMH endorsed AV in round #2.

On 28 Feb. 2005 the LA Times asked 710 "likely voters" by telephone:
"Among rivals of James K. Hahn, who has the most honesty and integrity?" The results were

A.Villaraigosa    23%
R.M.Hertzberg     17%
B.C.Parks         16%
R.Alarcon          7%
others/don't know 37%

which suggests that if Hahn were removed from the picture then Villaraigosa would still have been preferred versus Hertzberg. [But the wording's restriction to "honesty and integrity" makes that less clear than one would want, plus the (16+7+37=60)% other-way pollees confuse the issue.] This is supporting evidence that Villaraigosa was the "right winner." But it remains entirely possible (indeed, I consider it more likely than not) that Hertzberg "should have" placed 2nd instead of Hahn since, e.g, enough Villaraigosa voters preferred him over Hahn but were not allowed to say so in their votes. This problem (if it existed) would still have occurred with "instant runoff voting" since IRV would have ignored RMH>JKH preferences of Villaraigosa voters; but it would not have occurred with range voting. Indeed, a range-style telephone poll of 710 likely voters was conducted 22-27 February 2005 by the LA Times which confirmed this thinking:

"Averages" based on equispaced rating scale VF=3, SF=2, SU=1, VU=0 but the same finish order would have arisen with any 4 reasonable values.

2009: Antonio Villaraigosa re-elected in the first round (3 March) election by 56.6% majority (hence no runoff 2nd round occurred). Villaraigosa outspent Walter Moore by a ratio of 15 to 1. Moore got 22.1% of the vote; everybody else got less than 8.5% each. 10 candidates ran in all.

The interesting 2001 election

The outgoing mayor R.Riordan, the LA Daily News, and the Daily Breeze all endorsed Steve Soboroff(R); while the LA Times issued a dual endorsement of James K. Hahn and Antonio Villaraigosa (both Democrats). The results of the first round (10 April) were:

Runoff (i.e. second and final round, held 5 June; note the turnout increased by over 50,000 voters in the 2nd round versus that in the 1st round):

For the second round, Riordan switched his endorsement to Villaraigosa; and Wachs, long-time county supervisor Gloria Molina, the AFL-CIO, the California Democratic party, and the LA Times also endorsed him (while Soboroff and Becerra remained neutral). Despite this and his win in the primary, Villaraigosa did not get enough support from Soboroff & co's voters so Hahn defeated him.

LA Times pre-election polls indicated that Soboroff's voters were overwhelmingly white, but unfortunately for him in that respect he suffered a vote-split with Connell and Wachs. With IRV, range, or approval voting that vote-split would not have existed, which with IRV probably would have moved Soboroff ahead of Hahn into 2nd place.

Details: The racial, religious, and ideological compositions of voters in this election were elicited by an LA Times exit poll held 10 April and reported in table 7.1 page 172 of Karen M. Kaufmann: The urban voter: group conflict and mayoral voting behavior in American Cities, University of Michigan Press 2004:

  10 April 2001 LA mayor exit poll results in percentages

%voters     Villargsa Hahn Soboroff Wachs Becerra Connell
TOTAL          30      25     21     11      6       5
RACE-BASED:
52 White       23      19     30     17      3       6
14 Black       12      71      5      3      2       4
20 Latino      62       7      8      3     17       2
 4 Asian       23      32     25      5      5      10
SELF-PROFESSED IDEOLOGY:
47 Liberal     41      24      8     12      7       6
27 Moderate    23      29     23     13      4       4
26 Consrvtv    17      22     43      8      5       4
RELIGION:
33 Protestant  19      37     23      9      4       5
30 Catholic    40      17     21      6     12       3
16 Jewish      26      16     27     22      1       7

J.Rainey & G.Krikorian: THE MAYOR'S RACE; Hahn Won on His Appeal to Moderates, Conservatives, LA Times 7 June 2001 page A1. Exit-poll interviews of 3427 voters as left 62 polling places. "Hahn's advantage came mainly from voters who previously had backed Republican businessman Steve Soboroff and Councilman Joel Wachs in the first round of the election. Soboroff voters went 83% to Hahn and Wachs' voters 62% for Hahn in Tuesday's vote, The Times exit poll found."

Michael Finnegan: Valley's White Conservative Voter Bloc Backs Soboroff Series, LA Times 6 April 201 page B1. "Early in the campaign, mayoral candidate Joel Wachs, a city councilman, was widely seen as a potential favorite of those conservatives. But [Steve Soboroff], according to a Times poll published this week, has pulled well ahead of Wachs among that group. Among all likely voters, Soboroff is in a tight race with two other leading candidates, ... Hahn and... Villaraigosa."

If all 60K of the "moderate" and "conservative" white voters for Wachs and Connell had voted for Soboroff while their 30K "liberal" whites split between V & H, that would have been equivalent to introducing about 45K extra Soboroff votes. Those not only would have moved Soboroff ahead of Hahn, but indeed would have given him a near-tie for the lead with Villaraigosa. However, Villaraigosa also suffered a Latino vote-split with Becerra and hence, if that too were removed, would have returned to the lead. Actually probably this 60K-30K monolithic-split model was too pro-Soboroff biased, but even if it were, say, 50K-40K, or even 45K-45K, then still that would have been enough to move Soboroff ahead of Hahn. So it seems clear that Soboroff would have been second, not Hahn, with single transferable voting. End of details.

If so (assuming the election rules still finished with a 2-man runoff) Hahn would not have been in that runoff and hence would have lost.

Was Hahn a "wrong winner"? What would have happened with other voting systems?

As we've seen, it is probable that with Instant Runoff Voting Hahn would not have made it into the top two (which would have been Villaraigosa vs. Soboroff) whereupon probably Villaraigosa would have won. However, the legitimacy of that Villaraigosa victory could be questioned because of the fact that Hahn defeated him 53.5-to-46.5% in the head-to-head runoff!

Incidentally, this IRV election (with V defeating S in the final round) would have exhibited a "favorite betrayal" pathology: the Soboroff>Hahn>Villaigarosa voters would have been better off dishonestly-strategically voting for Hahn. Indeed, quite possibly, they in fact did so.

But with range or approval voting, the fact that the Villaraigosa voters (who were the liberalest) tended to prefer Hahn>Soboroff would have been able to come into play, which probably would have prevented Soboroff from making it into the top 2. (IRV would have ignored these preferences, but Approval and Range would have taken them into account.) Hence our prediction (or "postdiction") would be that the top two would still have been Villaraigosa & Hahn.

What would have happened then? Approval- and range-style polls (as well as "second choice" and plurality-voting-style) all were conducted in this election by the LA Times, and their results were interesting:

"Averages" in the range-voting tables above were computed using equally-spaced scores VF=3, SF=2, SU=1, VU=0. However, equal spacing was not necessary in June; Hahn's victory was clear since he would have won using any VF>SF>SU>VU. In April, most reasonable weights except for those highly-resembling "approval voting" (meaning VF=SF=1, SU=VU=0) would also have yielded a near-tie in which the Villaraigosa-Hahn margin was smaller than the exit poll's statistical error. But Hahn won with approval voting both using the above data and also with this poll data from late-February:

Graphical studies indicate that IRV is "biased" to favor extremists while approval voting is biased to favor centrists. Range voting, however, has little or no such bias. These vague views are fully supported by the data in the present election, viewing Villaraigosa as leftist, Hahn as centrist, and Soboroff as rightist.

What about some less-discussed voting systems: Borda, Condorcet, and Median-based "range" voting?

With Borda voting, because there were 15 candidates the temptation and power of strategic voting would have been overwhelming (with a strategic voter having ≈14 times greater discriminating power among the top contenders, than an honest one), and the situation is a textbook setup for a DH3 pathology guaranteeing that none of the apparent "top 3" (Hahn, Villaraigosa, Soboroff) could win and thus assuring victory for a "dark horse" such as Connell. (Read a story about just such a DH3 pathology.) In contrast, with range and approval voting DH3 cannot occur since the usual forms of strategic voting do not misorder candidates (there is no strategic purpose served by doing so that could not be served as well or better without misordering).

With honest Condorcet voting, it seems clear Hahn would have won (since the V voters preferred H>S, while the S voters preferred H>V). Indeed, we can tell that the finish-order among the "Big Three" would have been Hahn>Vilaraigosa>Soboroff. However, the DH3 pathology also can occur with Condorcet voting and hence with strategic voters put in the mix I do not know what would have happened. E.g. quite plausibly Connell would have won.

M.Balinski & R.Laraki have suggested a voting system like range voting but in which the greatest median (not the greatest average) score wins. With median-based range it would have been at least a 6-way tie since all 6 of the top 6 candidates in the April 2001 race have the same median score ("Somewhat Favorable", SF). Exact ties are common with medians but extremely rare with averages, which is why Balinski & Laraki recommend a certain tiebreaking procedure. If we apply that procedure then Villaraigosa would have won since after remove κ of the SF scores (where κ is any constant with 20%<κ<28%) from each candidate, then only Villaraigosa still has median score SF; all the others then have median=SU or worse. [Jameson Quinn points out his alternative "GMJ" and "MAV" tie-breaking methods also enthrone Villaraigosa.] The Balinski-Laraki system has a number of disadvantages versus ordinary average-based range voting (e.g. it is more complicated) two of which are visible in this election:

1. Nonmedian ratings are largely ignored. For example, Connell's score-distribution was (VF,SF,SU,VU)=(16,46,27,11)%. If this had been changed to (16,46,0,38) or (16,46,38,0) that would have represented a fairly large change in the perceived quality of Connell. Indeed, the latter change would have moved her up to third place with mean score 1.78 and the former down to 6th (or much worse) in the view of ordinary range voting. Meanwhile analagous changes could have moved Villaigarosa and Hahn's mean scores down to 1.76. But all these changes would have been invisible to Balinski & Laraki's voting method. And indeed, the average rating of Hahn exceeded that of Villaraigosa but Balinski & Laraki ignored the outer ratings that caused that and thus elected Villaraigosa.
2. Adding "all-zero ballots" can alter the election result with Balinski & Laraki's system (but cannot affect it with ordinary range voting, approval, etc). In particular, in the LA 2001 election their system elected Villaraigosa, but if an extra 33% worth of voters had magically appeared, all voting "all zero ballots" (i.e. scoring every contender≤"very unfavorable") then Hahn would have been victorious with the uniquely maximal median score "SF" (and thus no need for tiebreaking).

What is our view about range voting occasionally (and perhaps in LA 2001) "overruling voter majorities"?

As we have seen, Villaraigosa would have won with IRV and might have won with range voting (too close to call) on 10 April. Assuming for the purposes of argument that Villaraigosa would indeed have won with Range, that would be a fascinating situation in which Range Voting, like IRV, actually went against a voter-majority. What do we think of that?

From the theoretical point of view (both abstractly, and looked at in the particular case of this LA 2001 election), it does not bother me:

1. If 47% of voters prefer V>H strongly while 53% of voters prefer H>V weakly, then if the voters were being honest then in my view it is better for society that V win; range voting worked.
2. If the voters were instead being strategic (i.e. sometimes dishonest) about their preference-strengths then it is less-clear that V should win; but if this sort of strategicness acted in a balanced manner in both directions, then I still contend V should win. Hence in general I believe (and this is supported by Bayesian Regret simulation-measurements) that range will deliver a better winner on average than unweighted voting.
3. In the particular LA 2001 mayor election here, Villaigarosa voters tended to prefer V>H>S since V was "liberal," H "moderate," and S "conservative." Under these circumstances, "strategy" presumably have caused greater exaggerating of H>V preference strengths than V>H preference strengths because the V-voters, by exaggerating, would have been risking a greater backfire (i.e. electing S). Despite any such advantages Hahn may have enjoyed in the range-exit-poll due to such strategic lying to pollsters, V still defeated H in that poll (assuming favorable tie-break).
   
   I would conclude (under this tie-break assumption) that V was the "right winner"; and the LA voters themselves seemed to agree V's defeat was (in hindsight) a mistake, since they elected V over the incumbent H in the next (2005) election.
   
   In other words, any attack on range voting's overrule of majority vote based on hypotheses about strategy, fails because strategy would, if anything, have worked in the opposite logical direction. [Also, if one believes range voters would have been more strategic in a real election than in a poll, then that would either have had no effect or would have caused H to win the range-voting-election, in which case there would have been no "overrule" and the whole issue would be moot. This in fact is exactly what happened in June's 2-contender range poll and also in the approval-voting polls, experimentally confirming the thinking in previous sentence.]

From the propaganda point of view, though, this property of range voting could hurt it since propagandists will use it to attack range voting. Indeed, this already has happened with IRV-propagandists at "FairVote" attacking range voting for precisely this reason. In the 2001 LA election we are speaking of here, however, the IRV-propagandists have some difficulty employing that attack since

1. Villaraigosa did not clearly-win with range voting; instead it was a too-close-to-call near-tie between V & H.
2. IRV also would have "overruled the voter majority" and elected Villaigarosa (and on this, the evidence is clear).

Details of poll data

Return to main page