SUCCESS OF CANDIDATES PREFERRED BY HISPANIC AND NON-HISPANIC VOTERS IN RECENT ELECTIONS IN THE HOLLYWOOD PORTION OF LOS ANGELES

Stephen P. Klein, Ph.D.

July 7, 2002

Overview

This report begins by discussing the distinction between at-large and district voting systems. It then summarizes the three Gingles preconditions for a Section 2 challenge to an at-large system and reviews the criteria, data, and procedures that were used to examine which candidates Hispanic and non-Hispanic voters preferred and the success of these candidates in three local and three statewide contests. This analysis includes a description of the number of votes these candidates received citywide and in the majority-minority district that was created for Hollywood by Mr. Clayton.

The results of these analyses suggest that the candidates preferred by Hispanic voters would win as many if not more seats under an at-large plan as they would under a district system. This conclusion is based on the following findings: The candidates preferred by Hispanic voters won three of the six contests studied (Lt. Governor, Sheriff, and Mayor). The latter two races are especially probative because they are local non-partisan contests. No candidate received a majority of the Hispanics’ votes in the state Attorney General’s race or the Insurance Commissioner’s race; i.e., Hispanic voters were not cohesive in these contests. Their preferred candidate for Attorney General even lost in the majority-minority district. However, in both of these races, the Hispanics’ preferred candidates came in second citywide, well ahead of several non-Hispanic candidates. This could only occur if a significant percentage of the non-Hispanics voted for the Hispanics’ preferred candidates because less than a fifth of Hollywood’s voters are Hispanic.

The City Attorney’s race was the only contest that had an outcome that was consistent with the theory that an at-large system would dilute Hispanic voting strength. Dilution did not occur in the other five races—because in three of them the Hispanics’ preferred candidate won citywide, and in the other two, Hispanic voters were not cohesive.

Voting Systems

In a district system, a city is divided into several geographically defined districts, candidates must live in the district they are seeking to represent, and a voter may vote for only one candidate from among those running in that voter’s district. If no candidate receives a majority of the votes cast for the district’s seat, then the two candidates with the most votes have a run-off election. This is the way Los Angeles City Council members are elected.

In the at-large system planned for Hollywood, candidates would run citywide and a voter could cast as many votes as there were seats at stake (but no more than one vote for any one candidate). For example, if there are two seats at stake, each voter could cast one vote for one candidate and a second vote for another candidate. Voters may decide to cast just one of their votes (this is called “bullet voting”). The two candidates with the most votes win. There is no run-off unless there is a tie.

The Three Gingles Preconditions

One potential problem with an at-large system is that it could lead to the majority group filling all of the available seats with their preferred candidates. For example, if Hispanics comprise less than 50 percent of a community’s voters, then at least in theory, non-Hispanic voters could control who won every seat and thereby dilute Hispanic voting strength.

The Supreme Court in Thornburg v. Gingles, 478 U.S. 30 (1986) held that three “preconditions” must be met before an at-large system can be challenged as diluting minority voting strength under Section 2 of the Voting Rights Act. First, it must be possible to create a majority-minority district. This criterion is satisfied if a minority group has a majority of the district’s voting age citizens. Second, the minority group must be cohesive in its support of its preferred candidate; e.g., that candidate should receive more than half of the Hispanic vote. Third, the majority group must vote sufficiently as a bloc to generally defeat the minority group’s preferred candidate citywide. There would be no dilution if both groups preferred the same candidates or if the minority group’s preferred candidate won.
Criteria for Examining the Effects of Election Systems

The available data did not permit a direct determination of how many seats Hispanic candidates would win under a district versus an at-large system. However, indirect information regarding this matter can be gleaned from examining several recent contests to see how the Hispanics’ preferred candidates did citywide and in the majority-minority district (i.e., District #5 in Mr. Clayton’s plan).

Specifically, if the candidates preferred by Hispanic voters consistently lost citywide but won convincingly in the majority-minority district, then this trend would suggest that the candidate preferred by Hispanic voters is likely to have a more difficult time winning a seat under an at-large plan than under a district plan. However, one or more of the three Gingles’ preconditions would not be satisfied if the candidate preferred by Hispanic voters often won citywide or did not win convincingly in District #5 or if Hispanic voters were not cohesive in their support of this candidate.

Data Sources

The County of Los Angeles provided an electronic file that contained the number of Hispanic and non-Hispanic voters in each census block in the November 2000 general election, the district number in which each block fell in the plan proposed by Mr. Clayton, and the number of votes cast for each candidate in each block in the June 2001 City Attorney and Mayoral races. There were 907 blocks with voters in this database.[1]

The County of Los Angeles also provided an electronic file that contained the number of votes cast for each candidate in three 1998 statewide primary contests (Lt. Governor, Insurance Commissioner, and Attorney General) and one local election (Los Angeles County Sheriff).

The 1998 data were provided by Redistricting Unit (RDU) rather than by census block. An RDU is essentially a census tract and it contains an average of about 30 census blocks. There were 47 RDUs in Hollywood, but some of them extended beyond Hollywood’s borders. Data on the number of Hispanic and non-Hispanic voters within a census block or RDU were not available for the 1998 primary, but they were available for the November 2000 election. These considerations led to adopting the procedures described below for obtaining a count of the number of Hispanic and non-Hispanic voters and the number of votes cast for each candidate in the Hollywood portion of each of its 47 RDUs.

1. There were 34 RDUs that fell entirely within Hollywood. For these RDUs, the analysis used the number of votes that were cast for each candidate in the RDU in 1998 along with the number of Hispanic and non-Hispanic voters in November 2000 in the RDU’s census blocks.

2. For each of the other 13 RDUs, a count was made of the number of its November 2000 voters that lived in the census blocks that fell in Hollywood. A count also was made of the total number of voters in this RDU; i.e., regardless of whether they lived in Hollywood or not. The proportion that lived in Hollywood was then used to estimate the number of votes cast for each candidate in the Hollywood portion of this RDU.[2]

To illustrate this last step, 50 of the 68 census blocks in RDU #191900 were in Hollywood. There were 1,493 voters in the blocks in Hollywood and 640 voters in the blocks that were not in Hollywood; i.e., a grand total of 2,133 voters. Hence, the Hollywood portion of this RDU accounted for 70 percent of its voters (because 1493/2133 = .70). There were 206 votes cast for Baca in this RDU. The estimated number of votes that were cast for him in the Hollywood portion of this RDU was therefore .70 x 206 = 144.

Voter Preferences

The ballot box is secret. The percentage of Hispanic and non-Hispanic voters who preferred each candidate has to be estimated. Plaintiffs’ experts in Section 2 cases typically use “ecological regression” to do this.[3] This statistical procedure estimates Hispanic and non-Hispanic support for a candidate on the basis of the relationship between the percentage of voters in a geographic area (such as a census block or RDU) who are Hispanic and the percentage of voters in that area who cast a ballot for that candidate.

Table 1 shows the ecological regression estimates of the percentage of all the votes that were cast by a group for each candidate. The unit of analysis for these regressions was the RDU for the 1998 races and the census block for the 2001 contests. Table 1 also shows the number of votes each candidate received citywide and in District #5.

Local Contests. The most probative elections for a Section 2 challenge are the 1998 Sheriff’s race and the 2001 Mayor and City Attorney races because these are local non-partisan contests. The Hispanics’ preferred candidate won two of these three races. Specifically, Baca edged out Block for Sheriff and Villaraigosa trounced Hahn for Mayor by a 3 to 2 margin (61% to 39%).

In the City Attorney’s race, ecological regression estimated that Hispanic voters preferred Delgadillo over Feuer while the reverse was true for non-Hispanics. Although Feuer won citywide, Delgadillo came in first in three of Mr. Clayton’s five districts and second citywide. Hence, there must have been substantial cross-over voting by non-Hispanics. These results suggest that the Hispanics’ first choice and the non-Hispanics’ first choice would both be elected in the type of at-large system planned for Hollywood.[4] In addition, the same non-Hispanics who voted in this contest also preferred Villaraigosa over Hahn (see Table 1). This shows that candidate characteristics other than ethnicity determined who won in Hollywood.

Table 1

Ecological Regression Estimates of the Percentage of Voters in Each Group Casting Ballots for Each Candidate as a Percentage of the Total Votes Cast by that Group and the Number of Votes Cast for Each Candidate Citywide and in the Majority-Minority District in Mr. Clayton’s Plan

		Percentage of votes cast by:		Number of votes cast
Contest	Candidates	Non-Hispanics	Hispanics	Citywide	District 5
1998	Baca	33	39	4,211	531
Sheriff	Block	34	34	4,174	522
	All others	33	27	3,870	433
2001	Villaraigosa	55	73	10,851	1,889
Mayor	Hahn	45	27	6,854	872
2001	Delgadillo	35	72	7,911	1,624
City Attorney	Feuer	65	28	8,725	963
1998	Bustamonte	40	74	6,213	884
Lt. Governor	Miller	17	- 7*	1,704	104
	Hentshel	10	11	1,340	166
	All others	33	22	4,301	472
1998	Calderon	15	45	2,681	423
Attorney	Lockyer	28	25	3,565	433
General	All others	57	30	6,792	606
1998	Martinez	23	41	3,853	606
Insurance	Brown	37	18	4,380	426
Commissioner	Quackenbush	29	22	3,571	390
	All others	11	19	1,481	195

* See footnote 5.

Statewide Contests. The less probative but still relevant statewide elections showed a similar pattern of success for candidates favored by Hispanic voters. In the Lt. Governor’s race, Bustamante was the Hispanics’ and non-Hispanics’ first choice and he won. Voting was not ethnically polarized. The –7 value for Miller in this race was due to a violation of an assumption underlying the ecological regression methodology.[5] This violation does not affect the counts of votes cast for each candidate but it does raise a question about the trustworthiness of the ecological regression estimates.

In the Attorney General’s race, Hispanic voters preferred Calderon while non-Hispanics preferred Lockyer who won citywide. However, less than half of the Hispanics voted for Calderon and Lockyer finished first in every district, including #5. These results suggest that Hispanic voters were not cohesive and that a district system would not assure that their preferred candidate would win in the majority-minority district.

Hispanics probably comprise a majority of the voting age citizens in District #5 (i.e., the “majority-minority” district in Mr. Clayton’s plan). If so, this would satisfy the first Gingles criterion. However, Hispanics constitute less than half of this district’s voters. This imbalance may help to explain why Calderon did not finish first in this district (and why Block came very close to finishing ahead of Baca there too—see Table 1).

In the Insurance Commissioner’s race, the non-Hispanics’ first choice (Brown) came in first citywide while the Hispanics’ first choice (Martinez) came in first in District #5. However, according to ecological regression, only 41 percent of the Hispanics casting ballots in this contest voted for Martinez. Hispanics gave the remaining 59 percent of their votes to others. Thus, Hispanics were not cohesive in their support of Martinez. The non-Hispanics also split their votes among several candidates. They did not have a clear favorite. Again, this pattern suggests that the Hispanics’ preferred candidate would win at least one of the seats at stake in an at-large election.

Summary and Conclusions

The local and statewide contests had similar results. Taken together, these results show that there is no empirical basis for a claim of a pattern of vote dilution in Hollywood. Specifically:

· The candidates preferred by Hispanic voters won three of the six races studied, namely: Sheriff, Mayor, and Lt. Governor. The Sheriff and Mayor races are particularly probative because they involved non-partisan local contests.

· In the Attorney General’s race, the Hispanics’ preferred candidate (Calderon) lost citywide and in District #5. Part of the reason for this is that less than half of the Hispanics voted for him. This contest also is important because it shows that Mr. Clayton’s proposed district system would not assure the success of the Hispanics’ preferred candidate (note also that Block received almost as many votes as Baca in this district).

· In the Insurance Commissioner’s race, only about 41 percent of the Hispanics who voted cast a ballot for Martinez. Although Hispanics were not cohesive in their support of this candidate, she was nevertheless able to finish second citywide. This occurred, in part, because of substantial cross-over voting by non-Hispanics. This situation also suggests that candidates preferred by Hispanic voters are likely to win seats in a multi-seat at-large contest in Hollywood.

· The City Attorney’s race was the only one of the six contests studied in which Hispanic voters gave cohesive support to a candidate who lost citywide but who came in first in District #5. However, the non-Hispanics who cast ballots in this contest were the same voters who preferred Villaraigosa over Hahn for Mayor. Thus, candidate characteristics other than ethnicity determined who won in Hollywood.

One out of six contests does not constitute a pattern of white bloc voting generally defeating the candidate preferred by minority voters. In fact, just the opposite is true. Thus, on balance, the empirical evidence indicates that an at-large system would not result in vote dilution in Hollywood.

---

[1] Voters were assigned to a census block primarily by geocoding, with checking and completion by mapping of precinct geography (voter registration lists contain each voter’s name and address). The combination of geocoding and mapping was then used to create a crosswalk between precincts and blocks (i.e., an estimate of the number of voters from each precinct that resided in each block). This conversion was used to assign a share of the votes cast in each precinct to each of its blocks. The determination of whether a voter was Hispanic was based on the US Census Bureau’s dictionary of Hispanic surnames. Only 18 percent of Hollywood’s voters had a Spanish surname.

[2] Four of the 13 RDUs required very little allocation because they had multipliers over .95 or under .05. If a census block did not fall entirely in Hollywood, then the portion of it that did fall in Hollywood was identified by inspection of the relevant maps. This proportion was multiplied by the number of Hispanic and non-Hispanic voters in the block to estimate how many of these voters lived in Hollywood. For example, if 50 percent of a block was in Hollywood, then half of this block’s Hispanic voters and half of its non-Hispanic voters were assigned to the Hollywood portion of this block. Less than one percent of the blocks were partially in Hollywood. The same types of procedures were used to allocate an RDU’s voters and votes to the Clayton district(s) in which they fell; e.g., if the RDU’s blocks fell in more than one district.

[3] Other methods include exit polls, homogeneous precinct analysis, and other statistical procedures, such as Ecological Inference and the Neighborhood model (Freedman, D., Klein, S., Ostland, M., & Roberts, M. On “solutions” to the ecological inference problem. Journal of the American Statistical Association, 1998, 93, 1518-1522).

[4] Of the various types of at-large systems that could be adopted, the plan proposed for Hollywood is generally considered (even by the plaintiffs’ experts in Section 2 cases) as being the most friendly to minority voters. It enjoys this status because it often results in their preferred candidates winning seats in communities where there are significantly more non-minority than minority voters casting ballots.

[5] Ecological regression assumes that Hispanic voters are just as likely to vote for a given candidate regardless of where they live in the community. For example, the percentage of Hispanics voting for Baca is assumed to be the same regardless of whether they live in a more or less affluent tract, one with a relatively high or low percentage of high school graduates or Republicans, etc. There is a corresponding assumption about non-Hispanics. This is called the “constancy” assumption. The accuracy of the ecological regression estimates depends on this assumption being correct.