Local Logrolling? Assessing the Impact of Legislative Districting in Los Angeles

Introduction: Neighbors at War

In early 2000, the Patel family purchased a picturesque parcel of land nestled in the Santa Monica Mountains within the city limits of Los Angeles, California. ¹ Approximately one year later, the family submitted an initial application to build a single-family residence on the property, consisting of a 7,000-square-foot dwelling, a two-car parking garage, and a swimming pool. Nearly twice the size of the typical home on their street, the Patels’ proposal attracted the attention and eventual ire of some neighbors, including several who served on the boards of local nonprofits. With the help of these and other organizations, the neighbors hired scientists and experts to launch what would evolve into a lengthy battle to block the project using state environmental laws. After failing to persuade the city’s planning department—which eventually issued construction permits in May 2006—and losing their appeal before the local area Planning Commission, opponents brought their case to the Los Angeles City Council.

For several months, Councilman Tom LaBonge, whose council district included the neighborhood, attempted to broker a deal to end the impasse. His efforts to reach a compromise acceptable to both sides failed. At the council hearing considering the appeal over the Patels’ development permit, LaBonge announced that he would side with the project’s opponents. The councilman argued the best path forward was to require a full environmental impact analysis for the project—the first time such a report would be needed for a single-family home in Los Angeles. The prohibitive cost of a full analysis would, in effect, block the proposed construction.

LaBonge’s position conflicted with the advice offered to the council by the city attorney and Los Angeles planning staff. One staffer noted that demanding a full environmental impact review for a single-family development would have unintended consequences. Setting such a precedent, the staffer warned, could furnish development opponents everywhere with a dilatory tactic to challenge future zoning and planning decisions. Empowering such behavior could, in turn, lead to negative impacts on infill development and the supply of housing in the city. On the final vote, only one other council member sided with LaBonge.

We argue that the Patel case and others like it provide important insights for students of both legislative and local politics. First, the final outcome—the majority of the city council voting against the wishes of the member representing the affected area—is not one that standard models would predict. Most prominent theories argue that district-based representation encourages logrolling, with legislators trading votes to support one another’s preferred policy outcomes on issues affecting their respective constituents. Second, the case highlights an important aspect of policy that many descriptions of legislative voting often overlook. Even intensely local issues—such as a land-use controversy over a single home in one neighborhood of America’s second largest city—can and usually do implicate broader policy conflicts and can mobilize a wide range of electorally active political stakeholders.

In this study, we show that such conflicts often become intense enough to undermine existing norms of legislative deference. By drawing on a new data set of nearly seven years’ worth of roll call votes taken by the Los Angeles City Council, we trace the coalitional dynamics underpinning council decisions. We show that, for most ordinary questions that appear on the council agenda, council members routinely defer to the official representing the affected area. We also find, however, that deference sometimes breaks down on potentially controversial policy issues, leaving the local council representative in the voting minority. Adapting the informational theory of congressional organization (Krehbiel 1992) to the local political context, we develop a model of conditional deference that helps account for the unexpected legislative dynamics on the Los Angeles council. We conclude by considering the implications of our findings for the quality of urban governance and the likely policy consequences of the growing use of district elections in American cities.

Do District Elections Lead to Logrolling?

Local governments have undergone tremendous institutional change over the past three decades. Recent trends, such as the growing adoption of term limits, tax and expenditure limitations, and electoral reforms, have altered the nature of political representation in city government. The extent to which these reforms have also impacted local policy processes and outcomes, however, remains unclear.

In this article, we focus on one type of reform that has become commonplace since the 1970s in American cities: the replacement of at-large city council elections with district-based and mixed systems. Once the hallmark of large industrial cities—together with other “machine” institutions such as partisan ballots—district and ward elections have gained wider adoption in mid-sized cities across the country, including former bastions of “reform” in the Southwest (Bridges 1997). As Table 1 indicates, the prevalence of at-large elections has declined sharply since the late 1960s. The trend has been most stark in mid-sized cities with between 250,000 and 500,000 residents. ² In 1968, 70% of these cities relied exclusively on at-large elections to choose city council members. By 1998, only 20% of cities in this population category used this election method; the remainder adopted district-based elections or a mixed system that uses some combination of at-large and geographically districted seats. ³

OPEN IN VIEWER

Table 1.

Percentage of U.S. Cities Using At-Large City Council Elections.

Population	1968 (%)	1998 (%)
More than 500,000	46	33
250,000–500,000	70	20
100,000–250,000	66	54
50,000–100,000	61	49
25,000–50,000	69	54
10,000–25,000	60	60
5,000–10,000	66	62

Source. International City/County Management Association. Municipal Yearbook, various years.

While the growing use of district constituencies has led to representational gains for minorities, some observers argue that these improvements have come at the cost of undesirable public policy. In their classic book on city politics, Banfield and Wilson (1963) warned that, although district representation may improve representation for neighborhoods and minority interests, this election method may also incentivize public officials to ignore broader citywide concerns. They conclude that district elections encourage parochialism and lead to a legislative process marked by vote-trading among city council members (Banfield and Wilson 1963). Officials elected at-large, by contrast, seek to “legislate for the city ‘as a whole’” (Banfield and Wilson 1963, p. 92).

The logic behind this argument is straightforward: If council members represent a subset of the city and are electorally accountable only to voters who live within the boundaries of their districts, elected officials will invest their scarce political resources to furthering the interests of only their own constituents. As issues impacting other parts of the city do not affect their personal electoral fortunes directly, council members are largely free to ignore such policy concerns and defer to the colleagues who represent the relevant areas, in exchange for their promise—tacit or explicit—to do the same. More recent research, focusing primarily on Congress, has formalized this argument, predicting that district-based elections promote a norm of universalism that produces oversized rather than minimum-winning coalitions (Weingast 1979), lead to excessive pork-barreling by legislators (Shepsle and Weingast 1981), and encourage legislative deference (Weingast and Marshall 1988).

The thesis that district-based representation results in logrolling has undergone little direct empirical testing, although it continues to represent the conventional wisdom. The 2003 edition of the Municipal Yearbook, for example, notes, “At-large election proponents favor having council members elected by the entire city because at-large representation is thought to encourage impartial, community-wide attitudes, rather than parochial views, which in turn promote logrolling” (International City/County Management Association 2003, p. 15).

Existing empirical research has focused primarily on the relationship between municipal political institutions and fiscal policy, generally finding few meaningful differences between cities using at-large and district elections. Council members elected from districts report spending more time on constituent casework than those elected at-large (Welch and Bledsoe 1988) and adopt different electoral strategies (Haselswerdt 1984), but these do not appear to influence policy outcomes (e.g., Lubell, Feiock, and Ramirez 2005). For example, Bridges (1997) shows that reformed cities, which widely adopted at-large elections early in the twentieth century, have spent no less on policies rewarding key electoral constituencies than has been the case in districted machine cities. ⁴ Similarly, Morgan and Pelissero (1980) find no correlation between the reform structure of government and local taxing and spending decisions. ⁵

Extant studies focus almost exclusively on the levels of spending, taxation, and employment, perhaps because these factors are the easiest to quantify, measure, and compare across cities. There are two important weaknesses of this approach, however. As recent critics note (see Langbein, Crewson, and Brasher 1996; Primo and Snyder 2008), theory yields ambiguous predictions about the relationship between local electoral institutions and fiscal policy outcomes. For example, most models predict that district-based elections should not only lead to greater levels of pork-barreling but also depress investment in broadly shared public goods and citywide infrastructure that generate diffuse rather than district-focused benefits. These effects should offset one another in the aggregate measures of fiscal policy, providing no clear expectation about relationship between district elections and total spending levels.

Erie, Kogan, and MacKenzie (2011) also argue that district elections can lead to more liberal-minded representatives being elected to the city council. They suggest that legislative districting increases the electoral influence of poorer residents, whose lower propensity to vote puts them at a significant disadvantage in citywide elections. ⁶ Because city council districts must adhere to the constitutional one-man-one-vote principle (i.e., equal population in every district), irrespective of differences in turnout rates, district elections tend to increase representation for disadvantaged and, in general, more liberal constituents. If cities with district elections do indeed spend more and tax at higher rates, it remains unclear whether this outcome should be attributed to pork-barreling or to the change in the ideological composition of the city council produced by greater representation for disadvantaged groups at the district level.

Our study departs from the existing literature by analyzing legislative behavior directly, rather than focusing on ambiguous and difficult-to-measure downstream policy effects of council actions. In particular, we examine the nature of local legislative coalitions to test the prediction that geographic districts result in logrolling and undue deference among city council members. Although our research focuses on a single city, preventing us from examining the relationship between alternative electoral institutions and legislative behavior directly, the data do shed light on the frequency of logrolling on the council. We leverage variation in voting patterns across different policy types to examine whether council members defer to one another on bills affecting their respective constituents compared with policies affecting the city as a whole, as standard accounts predict.

A Model of Conditional Deference

Banfield and Wilson’s (1963) work on urban politics and more general legislative models predict that district elections provide a sufficient condition for universalism and logrolling. An individual city council member elected from a small geographic constituency should rationally promise to defer to her colleagues on bills affecting other districts in exchange for their pledges to do the same on bills targeting her constituents.

Although observing a high degree of deference in actual voting behavior would appear to support these predictions, other mechanisms may explain the norms of collegiality and unanimity among legislators seen in many political contexts. As Krehbiel (1992) notes, a legislator may follow the lead of her colleagues simply because she believes they possess a higher level of expertise about the policy question at hand than she does (see also Lupia and McCubbins 1998). Indeed, rational lawmakers may adopt a norm of deference to encourage specialization by individual legislators, yielding information that can help the entire legislative body make better decisions (see Gilligan and Krehbiel 1987, 1989, 1990; Krehbiel 1992).

In the context of local politics, deference may achieve informational efficiencies by encouraging council members to dedicate precious time and resources to learn about their own districts. Most city charters require council members to live in the districts they represent, increasing their exposure to the key issues confronting their constituents such as traffic patterns, local infrastructure needs, and sources of social tension. Such relevant policy information is likely to be less accessible to legislators who reside in other parts of the city, particularly in less professionalized municipalities where elected officials may lack staff and other resources necessary to conduct in-depth policy research. Aside from expertise about policy, which may also be available from professional staff in some cities, district representatives are also likely to invest greater effort in learning about the political ramifications of the policies affecting their district such as the positions that various stakeholders and interest groups take on these issues—the actors most likely to be active in local elections. If council members believe that other legislators will regularly overrule them, they have little incentive to use scarce staff time and resources to invest in gaining legislative expertise about their district and should instead direct such resources toward constituency services, advertising, and other election-minded strategies that improve their political standing (Mayhew 1974). As a result, encouraging individual members to become experts about their districts by deferring to them on issues affecting their constituents represents a smart and superior strategy for a legislative body in many cases, one that can improve the quality of decision making for the council as a whole.

There is, however, an important difference between the informational rationale for legislative deference and the traditional accounts that stress logrolling: If deference reflects and encourages expertise, it will not take the form of a rubber stamp. Although other members of the city council will defer to a legislator from the affected district on basic housekeeping matters—for example, on votes approving repaving a local street, the replacement of streetlights, or other similar uncontroversial and routine questions—they should not offer their support on issues that have significant redistributive implications or threaten to aggrandize the affected district at the expense of the rest of the city. Instead, the informational model of voting predicts that deference will be conditional. On policies affecting only an individual district, council members will defer to the legislator with the greatest expertise about the area. For issues that have the potential to affect the city as a whole—such as the vote on the level of environmental review required for the Patel house, which would have clear citywide implications—other members of the council should be far less willing to defer to the council member from the district in question.

Both sets of theoretical explanations lead to the following hypothesis:

The empirical prediction from Hypothesis 1 is that votes on district bills will pass with higher margins compared with votes on bills that do not target specific geographic areas. We test this hypothesis below by examining how frequently votes are unanimous and whether this rate varies across different types of bills.

The informational and distributive accounts of policy making lead to diverging predictions about what happens to the subset of district bills that produce policy spillovers across geographic and political boundaries, giving rise to conflicting interests among elected officials. Distributive models, which predict that deference is unconditional, yield the following hypothesis:

By contrast, the informational account stresses the limited nature of deference and predicts that the majority of the council will be hesitant to follow the lead of the legislator from the district primarily affected by the policy if doing so has potentially negative consequences for their own constituencies or the city as a whole.

In short, logrolling predicts that council members will carry the vote on controversial issues most closely affecting their districts, while our alternative account suggests that they will end up in the minority on divisive votes. To examine which of these hypotheses is most consistent with the empirical record, the bulk of our analysis focuses on tracing the legislative history of contentious bills. We identify contentious bills by examining the size of the legislative coalition on each vote. Specifically, we define a contentious bill as one on which there was a nonunanimous vote cast, with at least one city council member voting “yes” and at least one voting “no.” Although we cannot observe the precise mix of costs and benefits entailed by each proposal, the lack of unanimity serves as a useful indicator of conflicting interests between the district directly affected by the policy and representatives in other parts of the city. ⁸

Empirical Analysis: The Los Angeles City Council

To test the hypotheses described above, we turn now to analyze roll call voting on the Los Angeles City Council. We focus on Los Angeles for two reasons. First, from a theoretical standpoint, Los Angeles is similar on many dimensions to cities that have transitioned from at-large to district elections over the past three decades. Similar to most other major cities in the American Southwest, Los Angeles has been governed by a quintessential reform charter emphasizing nonpartisanship and powerful, autonomous bureaucratic departments since the early twentieth century (see Ingram 2009). Unlike the typical reform city, however, Los Angeles has historically combined these institutions with district elections and a strong, independently elected mayor. As other Southwestern cities abandon at-large elections but continue to maintain other vestiges of reform, they are increasingly coming to resemble Los Angeles with regard to their governing institutions.

Similar to other cities with district elections, the Los Angeles City Council has built a reputation for collegiality. This norm attracts the most attention from outsiders on the rare occasions when someone on the council violates it. The most visible recent controversy surrounded the city’s decennial redistricting after the 2010 Census, when a 13-member council majority adopted new district boundaries over the objections of two very vocal opponents, Councilwoman Jan Perry and Councilman Bernard Parks. Media coverage of the process repeatedly noted the unusual dynamics that the issue created on the council. “To the outside world,” one recent article in the Los Angeles Times begins, “the Los Angeles City Council is a body that works harmoniously, with members almost always voting unanimously and working congenially when they don’t” (Zahniser and Linthicum 2012). Put another way, Los Angeles is precisely the kind of city in which one would expect to see a deep-rooted practice of legislative logrolling on the city council.

Second, the city’s legislative records are particularly conducive to examining the hypotheses we develop in this article. Since 2003, the city clerk has recorded and maintained an electronic database that archives the votes cast by the city council. In addition to providing a brief description of each policy proposal and each member’s vote, the data set records which, if any, of the districts were specifically affected by each piece of legislation. The relevant districts are identified by city staff when items are scheduled on the council agenda and allow the elected officials, local residents, and other stakeholders to identify which upcoming votes are likely to affect their geographic area of interest. ⁹ These geographic codes allow us to identify which legislator represents the area affected by each policy proposal and to separate primarily district-focused from broader citywide votes. In this analysis, we focus on approximately 47,000 votes cast between July 2003 and March 2010.

During the period under study, the city council consisted of 15 members elected by district from various parts of the city. Of the total membership, just two councilmen were Republicans (Dennis Zine-District 3, Greig Smith-District 12), making partisanship a largely unimportant cleavage on the council. Because our time period spans several election cycles, our full sample includes votes cast by 21 different individuals who have served on the council during the period between 2003 and 2010. Figure 1 provides a map of the districts, while Table A1 in the appendix reports basic demographic information about each district taken from the 2000 Census.

OPEN IN VIEWER

Figure 1.

Los Angeles City Council districts (2001–2011).

In Table 2, we summarize the actions taken by the council during this seven-year period according to their geographic scope. Policies targeting one city council district represent the modal category, accounting for 63.8% of the votes taken by the council during our period of study. By contrast, citywide legislation—policies that did not focus on individual council districts—represented 25.8% of the total. Local legislation affecting more than one council district, such as the resurfacing of a street crossing district boundaries, constituted less than 2% of all votes. The final category includes procedural votes—for example, whether to accept a report from an advisory council—that were not coded for geographic scope by city staff or the clerk and dealt largely with citywide issues or parliamentary process. Such procedural votes accounted for 8.8% of roll call votes taken by the council.

OPEN IN VIEWER

Table 2.

Los Angeles City Council Votes, 2003–2010.

Type of Vote	% of Votes
Citywide legislation	25.8
Legislation primarily affecting one district	63.8
Procedural motions	8.8
Legislation primarily affecting multiple districts	1.6

Legislation affecting specific parts of the city (about two-thirds of all votes cast during this period) appeared to be well distributed across various council districts. As Figure 2 shows, each district was targeted by approximately 5% of the bills voted on by the council. One clear outlier is District 9, which was the subject of roughly twice as many council votes as other areas of the city. ¹⁰ This council district includes the downtown area of Los Angeles, which saw particularly heavy development during our period of study. Between 2000 and 2009, this area accounted for nearly a third of all new multifamily dwelling units, more than a fifth of all new office space, and almost half of all new retail floor space constructed in the city (Deener et al. 2013, Table 14.1), explaining its prominence on the council agenda.

OPEN IN VIEWER

Figure 2.

Distribution of votes by affected district.

Our first hypothesis predicts that district bills will be more likely to garner unanimous approval when compared with citywide legislation. The data, which we summarize in Table 3, show that the voting behavior of the city council supports this hypothesis. While the vast majority of all legislation was adopted by consensus—indeed, more than 97% of all votes were unanimous—district bills were significantly more likely to pass without dissent. Overall, 99.8% of all district bills received unanimous votes, compared with 97.7% for citywide bills, a statistically significant albeit substantively modest difference. ¹¹

OPEN IN VIEWER

Table 3.

Proportion of Unanimous Votes.

	Citywide Legislation	District Legislation	Difference Legislation
Proportion unanimous	97.7%	99.8%	2.1%***

Significant at *p < .05. **p < .01. ***p < .001 (two-tailed test).

While the high rate of unanimity may be surprising to some readers, it is important to place these figures in comparative context. Although media coverage often focuses on the most contentious and divisive proposals, the vast majority of bills in most legislatures pass unanimously. During the highly partisan and polarized 110th Congress (2007–2009), for example, nearly 94% of all measures adopted by the U.S. Senate passed by unanimous consent (Rybicki 2008). In other words, the level of unanimity on the Los Angeles City Council does not appear to be atypical of other legislative bodies.

Which explanation—nefarious logrolling or deference to local area experts—best explains the modestly higher rate of unanimity on district bills? As a first step toward answering this question, we coded a sample of approximately 1,000 votes taken by the council to identify the precise nature of the bills under consideration. This sample included all of the approximately 450 contentious votes that occurred during the seven-year period under study as well as a random sample of approximately 500 unanimous votes taken by the council.

Table 4 summarizes the content breakdown for both district and citywide bills, identifying key differences between the two. The chi-square statistic is significant at the p < .01 level, suggesting that the composition of district bills is substantially different from that of citywide legislation. Bills concerning land-use issues account for nearly half (46%) of the district votes in our sample compared with just 9% for citywide votes. These are precisely the kinds of issues on which individual council members likely possess substantial expertise—at the very least, more expertise than their colleagues from other parts of the city. By contrast, contract awards, a type of vote council members are probably equally qualified to cast, made up more than a quarter of citywide votes compared with just 5% of district legislation. The stark differences between the composition of district and broader legislation suggests that legislators have legitimate reasons to defer to the specialized knowledge of individual council members, even though this type of behavior may look suspiciously like opportunistic logrolling to outside observers.

OPEN IN VIEWER

Table 4.

Comparison of Bill Categories for Citywide and District Votes.

Policy Category	Citywide Votes	District Votes	Difference
Appointment	4.0% (28)	0.5% (2)	−3.5%
Budget	18.5% (128)	18.7% (75)	0.2%
Contract	27.0% (187)	5.0% (20)	−22.0%
General ordinance	12.4% (86)	1.2% (5)	−11.2%
Land use	9.4% (65)	46.4% (186)	3.70%
Other	28.6% (198)	28.2% (113)	−0.4%
n	692	401

Note. χ² = 266. p < .01. Number of observations in parentheses.

That is not to say that all local land-use bills are dispatched without opposition. Table 5, which looks specifically at district bills and compares the composition of unanimous and contested votes in our sample, shows that land-use questions account for nearly two-thirds of all contested district votes. Note, however, that this should not be read to mean that most district land-use votes are contentious. Almost all votes on district bills are unanimous, and this is true of the vast majority of all district land-use votes. Indeed, projecting from the distribution in our sample, more than 99% of all district land-use bills receive no opposition, a higher rate of unanimity than all citywide legislation. Rather, Table 5 suggests that the general norm of unanimity breaks down slightly more frequently on land-use questions than on other types of district bills, leading this category of bills to account for the largest share of the contentious district votes on the council.

OPEN IN VIEWER

Table 5.

Comparison of Unanimous and Contested District Bills.

Policy Category	Unanimous Votes	Contested Votes	Difference
Appointment	0% (0)	1.9% (2)	1.9%
Budget	19.3% (57)	17.0% (18)	−2.3%
Contract	4.1% (12)	7.5% (8)	3.4%
General ordinance	1.0% (3)	1.9% (2)	0.9%
Land use	39.3% (116)	66.0% (70)	26.7%
Other	36.3% (107)	5.7% (6)	−30.6%
n	295	106

Note. χ² = 46.1. p < .01. Number of observations in parentheses.

Although most district bills receive unanimous votes from the council, we argue that deference among council members should not be absolute. When the interests of the city as a whole conflict with the preferences of the member from the affected districts, we expect deference to give way and the majority of the council to defeat such measures.

To explore our second hypothesis, we turn now to examine only the contentious—that is, all nonunanimous—votes taken by the council. Our key question is whether legislators from the district affected by each piece of legislation are more likely to prevail on these votes, as the logrolling explanation predicts, or whether they are more likely to end up in the minority, as our alternative model of conditional deference hypothesizes.

Using the votes on contested bills, we created a new data set that used each bill–legislator combination as the unit of analysis. We coded the dependent variable equal to “1” if legislator i was on the council majority on vote j for all i and j in our data and “0” if the legislator was in the minority. ¹² We treated council members who were absent or chose to abstain from the vote as a missing observation and thus excluded them from the analysis for that specific vote. The independent variable of interest is whether bill j affected the city council district represented by legislator i. We coded each bill–legislator combination as “1” if the legislator represented the affected district and “0” otherwise. Because our dependent variable is dichotomous, we estimate logistic regressions predicting each legislator’s probability of being on the winning side of each roll call vote. As each bill appears multiple times in the data set—the unit of analysis is the legislator–bill pairing—we used robust clustered standard errors to account for factors that may be correlated across legislators on any given vote. We present the results from the estimation in Table 6.

OPEN IN VIEWER

Table 6.

Logit Estimates for Probability of Being on Winning Side of Roll Call Vote.

	Model 1	Model 2	Model 3	Model 4
Own district	−0.87** (0.29)	−1.06** (0.31)	−0.81* (0.32)	−0.99** (0.36)
District policy			−0.11 (0.15)	−0.09 (0.17)
Budget vote			0.29 (0.23)	0.33 (0.23)
Contract vote			0.60** (0.20)	0.72*** (0.25)
General ordinance			0.83** (0.25)	0.94*** (0.26)
Land-use vote			0.45* (0.21)	0.53* (0.22)
Other vote			0.47** (0.21)	0.56** (0.19)
(Intercept)	1.58*** (0.04)	1.04*** (0.15)	1.18*** (0.17)	1.60*** (0.33)
Legislator fixed effects	No	Yes	No	Yes
Pseudo-R2	.02	.05	.01	.08
No. of observations	10,696	10,696	7,561	7,561
No. of bills	487	487	342	342

Note. Robust standard errors clustered on vote in parentheses. Excluded category of votes is appointments.

Significant at *p < .05. **p < .01. ***p < .001 (two-tailed test).

In the first model, we report the results of a simple bivariate regression. In the second model, we include a fixed effect for each city council member to account for the existence of council factions that may cause a legislator to systematically end up in the majority or minority. The use of fixed effects creates a within-legislator design, allowing us to estimate whether any given council member is more or less likely to prevail on a vote targeting her district compared with the same legislator’s probability of being in the majority on all other votes, including citywide legislation or bills affecting other members’ districts.

Overall, the first two models provide evidence that is clearly inconsistent with logrolling. The coefficient on the Own District variable is negative and highly significant, indicating that city council members are less likely to prevail on a contentious vote when the issue concerns their own district compared with issues targeting other areas or the city as a whole. Put another way, the majority of the city council has few hesitations about voting against the wishes of the legislators from districts affected by specific pieces of legislation on contentious policies. Not only are the results robust to the inclusion of legislator fixed effects (model 2) but the magnitude of the effect is also slightly larger in this specification.

Because the model coefficients from a logistic regression do not yield themselves to intuitive interpretation, we calculate simulated probabilities to estimate the substantive size of the effect using the more conservative results from the first model. For contentious citywide bills and legislation targeting other districts, a city council member will, on average, be on the winning side of the vote 83% of the time. Yet, when legislation affects her own district, the council member’s probability of voting with the majority drops to 67%, a substantively large decline. The predicted effects along with their confidence intervals for the first two models are summarized in Table 7. In each case, representing the affected district substantially reduces each council member’s probability of voting with the council majority on a contentious proposal.

OPEN IN VIEWER

Table 7.

Predicted Probability of Voting with Council Majority.

	Citywide/Other District	Own District	Difference
Model 1	0.83 [0.82, 0.84]	0.67 [0.53, 0.78]	−0.16 [−0.04, −0.29]
Model 2	0.89 [0.85, 0.92]	0.73 [0.58, 0.85]	−0.16 [−0.05, −0.30]

Note. Predicted values generated via simulation. Confidence intervals reported in brackets.

In models 3 and 4, we also include controls for bill type, using the same typology as discussed earlier. The analysis in these models excludes purely procedural votes, which explains the reduced number of cases included in the regression. In addition, these models include a dummy variable for district votes, with citywide votes and appointments serving as the omitted categories. The inclusion of these additional controls does not change the primary findings: When proposals attract at least one dissenting vote, council members from the affected district are more likely to be in the minority compared with other members of the council and their own winning rate on other policy proposals. Although the predicted effects are slightly smaller in the full specification, representing the affected district still reduces the probability of being in the majority by approximately 13 percentage points. Finally, we estimated an additional model that included an interaction effect between Own District and land-use bills. ¹³ The results, available from the authors upon request, show that the coefficient for representing the affected district is only significant and negative for land-use bills, although the very small number of nonland-use contentious district votes in our sample cautions us against in putting much confidence in or emphasis on this finding.

Our analysis points to two important conclusions. First, as evidenced by the higher rate of unanimity on district bills, individual council members are indeed willing to defer to their colleagues on legislation that affects districts other than their own. Second, our results show that deference is conditional. When policies are contentious—perhaps because they favor a particular district at the expense of the rest of the city or because they mobilize stakeholders who actively lobby against their passage—the majority of the council is more likely to send these bills to defeat, often over the objection of the legislator who represents the district directly targeted by the vote. Together, these results suggest that the norm of deference on the council reflects a reasoned decision on the part of elected officials to recognize the expertise that their colleagues may possess over the areas they represent, rather than a practice of blind deference and logrolling.

Discussion

Banfield and Wilson (1963) argue in their classic treatise that district elections have both benefits and drawbacks. In particular, they assert that parceling the city into districts fosters better representation for specific neighborhoods and increases the representation and influence of minorities, whose voices can be drowned out in a citywide contest. In examples all over the United States, the adoption of district elections has fulfilled the promise of amplifying both the voice of specific neighborhoods and racial and ethnic groups (Marschall, Ruhill, and Shah 2010).

Los Angeles is perhaps one of the best examples of how using district elections to elect representatives to the city council can bring diversity to the composition of the legislative body and enhance minority influence. The council has been instrumental in giving a political voice to previously underrepresented groups: at the time of our writing, 6 of the current council’s 15 members were themselves ethnic minorities—3 Latinos and 3 African-Americans. In addition, Tom Bradley, the iconic African-American who served as the city’s mayor for two decades, began his political career on the council. Without the presence of district elections, it is likely that Tom Bradley’s decorated public service career would have been significantly delayed or scuttled altogether (Sonenshein 1993).

Every electoral system requires trade-offs between desirable democratic goals. A number of scholars have argued that district elections come at the expense of citywide interests by leading to rampant logrolling and, as a result, civic waste. Fortunately in the case of Los Angeles voters, there is little evidence that district elections have produced a city council that meets these expectations. Despite Banfield and Wilson (1963) warning that district elections result in parochialism and subordinate the city’s best interest to narrow district concerns, our analysis reveals that the city’s elected officials have been able to find a balance between these competing considerations. While our results indicate that the city council often defers to individual members when issues affect their districts, we also show that elected officials have at times been conscientious trustees for the entire city by rejecting a significant portion of district-specific legislation that may have negative policy implications for other areas. While we agree electoral incentives encourage individual council members to privilege the interests of their district and pursue policy outcomes that favor their particular locale, we also argue that the need to build council (and thus citywide) majorities to pass legislation gives other officials a veto over policies that may benefit one district at the expense another or the city as a whole. Together, our two main findings depict the Los Angeles City Council as a legislative body that has utilized district elections to encourage members to specialize—thereby enhancing the expertise of the entire council—without sacrificing the best interests of the city.

More broadly, this article makes two important contributions to the literatures on urban and legislative politics. First, we develop an informational model of legislative deference that yields clear empirical implications distinct from standard logrolling accounts. Moving beyond previous research that focuses largely on cross-sectional analysis of spending and tax levels, we apply our theoretical framework to develop predictions for legislative behavior. Second, we test these implications using city council roll call voting data, a rich data source that scholars of urban politics have largely overlooked.

We acknowledge that our analysis covers only the city of Los Angeles, which differs from smaller jurisdictions in several respects that may limit the generalizability of our findings. Compared with most municipalities, the Los Angeles council has achieved a high degree of professionalization. For example, specialized policy committees review most policies before they reach the full council for review, and individual council offices employ a number of full-time staff who advise elected officials. Together with other city employees and the Office of the Chief Legislative Analyst, these resources provide council members with ample opportunity to identify potentially controversial proposals and rework or simply block them before they reach a full council vote, perhaps explaining the high degree of consensus observed on the council. We emphasize, however, that these factors cannot explain why district bills pass with a higher degree of unanimity than other policies, because both types of bills are reviewed by council committees and professional staff. The difference in the disposition of district and citywide proposals, albeit modest, suggests that local council members do bring an additional level of expertise to the table that their colleagues acknowledge.

The second key difference is that Los Angeles uses particularly large electoral districts, with each council member representing between 230,000 and 260,000 constituents (see Table A1). This may help explain why the city has not experienced the pathologies critics associate with district elections. Banfield and Wilson (1963), for example, argue that parochialism and logrolling both decrease with district size. Other models suggest, however, that incentives for logrolling should depend on the number rather than size of districts, increasing as the number of constituencies grows (see the literature on the “Law of 1/n,” for example, Weingast, Shepsle, and Johnsen 1981). On this dimension, the 15-member Los Angeles City Council is substantially larger than is the case in most cities. Given this atypical institutional context in Los Angeles, our results do not provide a definitive account of the legislative impact of district elections. Additional research that compares the level of council unanimity between cities using at-large and district elections, as well as interactions between the method of election and other institutional factors such as partisan elections, promises to produce even richer insights about the nature of legislative politics and local policy processes. We believe these questions provide a fruitful direction for future researchers. Our findings do suggest, however, that district elections need not give rise to pernicious logrolling and other forms of undesirable legislative behavior, an important consideration for local governments and political reformers weighing the pros and cons of alternative institutional arrangements.