RESPONSIVENESS IN CITY GOVERNMENT

The term responsiveness means that government “responds” to changes in citizens’ views by moving policy in the direction of those views. Cross-sectionally, this implies that places where the public holds more conservative views should have more conservative policies (Erikson, Wright, and McIver Reference Erikson, Wright and McIver1993). This definition of responsiveness is based on liberal notions of popular sovereignty. At a minimum, in a representative democracy the views of citizens should influence government policy decisions (Achen Reference Achen1978).Footnote ¹

Many scholars argue that municipal governments are unresponsive to the views of their citizens (Craw Reference Craw2006; Morgan and Watson Reference Morgan and Watson1995; Peterson Reference Peterson1981; Ruhil Reference Ruhil2003; Wolman, Strate, and Melchior Reference Wolman, Strate and Melchior1996). This view suggests that elected city leaders have limited control over policy outcomes due to a multitude of institutional constraints (Gerber and Hopkins Reference Gerber and Hopkins2011). First, cities are subordinate to states and the national government. There are a variety of statutory or constitutional constraints on specific local policies (Ladd and Yinger Reference Ladd and Yinger1989). For instance, many states restrict local governments’ ability to levy sales or property taxes. Moreover, there are a number of areas where responsibility over policy is shared between levels of government (Berman Reference Berman2003; Craw Reference Craw2006; Nivola Reference Nivola2002; Peterson Reference Peterson1995). Federal and state governments also exert indirect control over local policy by providing restricted grants and funding streams for specific programs. In addition, cities face constraints due to economic competition from other jurisdictions (Bailey and Rom Reference Bailey and Rom2004; Ladd and Yinger Reference Ladd and Yinger1989; Peterson Reference Peterson1981; Rae Reference Rae2003). Indeed, cities have little control over the movement of people, industry, and capital across their borders.

Overall, the consensus in the literature on municipal politics is that the policy decisions of city governments are unresponsive to the views of their citizens. However, there are reasons to believe that city governments should be responsive to the policy preferences of their citizens. The central assumption of American politics scholarship over the past 30 years is that elected officials are primarily motivated by electoral incentives (Kousser, Lewis, and Masket Reference Kousser, Lewis and Masket2007; Mayhew Reference Mayhew1974). Re-election minded officials have incentives to adhere to the will of their constituents in order to gain their votes. This means that they should be responsive to the median voter in their constituency (Downs Reference Downs1957; Erikson, Wright, and McIver Reference Erikson, Wright and McIver1993).

A variety of scholars have found that citizens hold local officials accountable for their decisions in office. Arceneaux (Reference Arceneaux2005) finds that survey respondents connect their evaluation of mayors’ performance on traffic congestion and other salient issues to their vote choice. Similarly, Howell and Perry (Reference Howell and Perry2004) show that respondents’ evaluations of city services in four large cities (Charlotte, Chicago, Detroit, and New Orleans) were significantly related to mayoral approval ratings. Finally, Stein, Ulbig, and Post (Reference Stein, Ulbig and Post2005) find that mayoral approval significantly predicts vote choice in several recent Houston mayoral elections.

Responsiveness in cities does not necessarily depend on a traditional view of the legislator-constituent relationship, in which constituents observe the actions of legislators and reward whoever best represents their policy preferences. Tiebout (Reference Tiebout1956) offers a model of city politics in which citizens locate themselves in cities which best match their preferences for public goods provision. It may be the case the citizens vote on the basis of vague notions of approval for the policies of their city, and that their choice to move or stay creates a market-based mechanism for city representation. In the Tiebout model, citizens need not have a deep knowledge of the actions of their public officials, they need only know what level of public goods they are receiving. This logic applies equally well to other policies besides public goods provision, for instance the kind of public goods provided. One implication of a Tiebout type model is that representative institutions may not matter very much. Elected politicians are incentivized to pursue policies that retain and attract like-minded citizens, regardless of whether they are city councilors or mayors, partisan or nonpartisan, or whether voters can change policy directly at the ballot box.

Some recent work supports the notion that local policymakers are responsive to the views of their constituents. For instance, Gerber and Hopkins (Reference Gerber and Hopkins2011) show that Democratic mayors spend less on police and fire services than their Republican counterparts.Footnote ² Palus (Reference Palus2010) examines policy responsiveness in 26 large cities, and finds strong evidence that the ideological preferences of citizens are reflected in the spending decisions of governments. At the county level, Choi et al. (Reference Choi, Bae, Kwon and Feiock2010) find that Democratic votes for president are correlated with greater expenditure levels and a larger share spent on redistribution. Overall, these previous theoretical and empirical studies on responsiveness lead to our first hypothesis:

THE IMPACT OF INSTITUTIONS ON RESPONSIVENESS IN CITIES

There are a number of institutions that reformers have established to improve the quality of municipal government (Lubell et al. Reference Lubell, Feiock, La Cruz and Ramirez2009; Trounstine Reference Trounstine2008). In this section, we focus on five such institutions: the presence of a city manager rather than an elected mayor, the presence of direct democracy provisions, the presence of nonpartisan elections for mayor and city council, the presence of term limits, and the presence of at-large versus single-member elections.

These institutions are a good test case for the importance of municipal institutions because they were designed with representation in mind. In particular, reformers around the end of the 19th century sought to cultivate “better” government by decreasing the power of party machines, increasing professionalization and promoting political involvement by the “right” kind of people. The Progressive Era reformers wanted to diminish the power of narrow interests and wrest power from local bosses. Although the power of party bosses greatly diminished over the course of the 20th century, the broader impacts of these reforms is unclear. In particular, we do not know whether they enhance or reduce representation in cities as a whole, especially now that the power of local political parties are greatly diminished even in cities that retain prereform institutions (Trounstine Reference Trounstine2008).

According to one recent study, “the most frequently analyzed and politically debated feature of municipal government is the balance of electoral versus managerial power in the executive branch of city government” (Lubell et al. Reference Lubell, Feiock, La Cruz and Ramirez2009). In the early 20th century, most cities had an elected mayor that led the executive branch and a city council that handled legislative functions (Schiesl Reference Schiesl1977). The Progressive reform movement came to link mayor-council systems with the inefficiency and corruption of party machines. Reformers argued that city governments should be run by experts rather than politicians (Hofstadter Reference Hofstadter1956; National Municipal League 1916). The “reform” council-manager system eliminated the political position of an elected mayor as chief executive (Lubell et al. Reference Lubell, Feiock, La Cruz and Ramirez2009). Instead, cities hired a professional city manager to run the government and make daily administrative decisions. The mayor was reduced to a figurehead with little real power. Most cities have adopted a council-manager form of government (Ruhil Reference Ruhil2003; Svara Reference Svara1990). While city managers may be better than elected mayors at promoting efficiency and economic development (Stein Reference Stein1990), the dominant view among scholars is that cities with an elected mayor are more responsive to the views of their citizens than cities with a “reform” council-manager system (Sharp Reference Sharp1997; but see Lubell et al. Reference Lubell, Feiock, La Cruz and Ramirez2009). This leads to the hypothesis that:

Progressive reformers also believed that partisan elections helped to increase the power of party bosses. As a result, they promoted the creation of nonpartisan elections for municipal office (Trounstine Reference Trounstine2010). In nonpartisan elections, parties do not officially nominate candidates for office, and candidates’ party affiliations generally do not appear on the ballot. Most cities in the United States have adopted nonpartisan elections. Scholars have reached conflicting results on the effect of nonpartisan elections on representation. Hansen (Reference Hansen1975) finds some evidence that cities with nonpartisan elections have weaker representation. More recent work on nonpartisan judicial elections has found that judges elected through nonpartisan elections are more responsive to public opinion since they cannot rely on partisan cues to signal their policy positions (Caldarone, Canes-Wrone, and Clark Reference Caldarone, Canes-Wrone and Clark2009). But others argue that nonpartisan elections typically have lower turnout than partisan elections, which may increase the power of special interests (Schaffner, Streb, and Wright Reference Schaffner, Streb and Wright2001). We examine the following hypothesis:

A more recent reform designed to increase democratic responsiveness is the development of direct democracy provisions. The potential for citizen initiatives may create stronger incentives for elected officials to be attentive to constituent interests. As a result, scholars argue that policy choices are more likely to be responsive to voters’ preferences when direct democracy exists (McCabe and Feiock Reference McCabe and Feiock2005). While few studies have studied the effect of direct democracy on representation at the local level (Lubell et al. Reference Lubell, Feiock, La Cruz and Ramirez2009), the evidence at the state level is mixed. A number of studies have found that majoritarian interests are more likely to prevail in states with direct democracy institutions (Gerber Reference Gerber1999; Matsusaka Reference Matsusaka2010; Tolbert Reference Tolbert, Bowler, Donovan and Tolbert1998). But other studies find no significant relationship between the presence of the citizen initiative and democratic responsiveness (Lax and Phillips Reference Lax and Phillips2012). We examine the following hypothesis:

Reformers in the late 20th century have also argued that the presence of term limits affects the link between elected officials and their constituents (Carey, Niemi, and Powell Reference Carey, Niemi and Powell2000). Some scholars find evidence that the turnover caused by term limits leads to less experienced elected officials (Kousser Reference Kousser2005). This reduces the capacity of lawmakers to assess and respond to public opinion. Moreover, it may reduce legislators’ incentives to respond to public opinion by limiting the value of elected office (Carey et al. Reference Carey, Niemi, Powell and Moncrief2006). However, other scholars argue that the turnover caused by term limits reduces incumbency advantages and leads elected officials to better reflect current constituents’ preferences (Moncrief, Powell, and Story Reference Moncrief, Powell, Story, Niemi, Cain and Kurtz2007). This leads to our fifth hypothesis:

Reformers in the early 20th century also promoted city-wide (“at-large”) elections to prevent narrow interests from exerting too large an influence on local government (Trounstine Reference Trounstine2010). In unreformed cities, the municipality is divided into geographic areas of roughly equal population size, and each district elects a single city councilor. In reformed cities, councilors are elected by the municipality as a whole. Most previous studies on the impact of at-large elections focus on descriptive representation. These studies generally find better descriptive representation for African-Americans and other racial minorities under a districted system than an at-large system (Bullock and MacManus Reference Bullock and MacManus1993; Davidson and Grofman Reference Davidson and Grofman1994; Trounstine and Valdini Reference Trounstine and Valdini2008; Welch Reference Welch1990). However, while at-large elections may harm descriptive representation, there are good reasons to believe they should enhance responsiveness by “shift[ing] electoral power toward a single median voter and away from geographically concentrated interests” (Trounstine Reference Trounstine2010). This leads to our final hypothesis:

RESEARCH DESIGN

Measuring City Policy Preferences

As the starting place for our model of city policy conservatism, we estimate a large sample of citizens’ ideal points using an approach similar to the one taken by Tausanovitch and Warshaw (Reference Tausanovitch and Warshaw2013). First, we pool together data from seven recent large-scale surveys of the American public (the 2006, 2007, 2008, 2010, and 2011 Cooperative Congressional Election Surveys (CCES) and the 2000 and 2004 Annenberg National Election Surveys (NAES)). Each of these surveys asked between 14 and 32 policy questions to 30,000–80,000 Americans.

We assume that all survey respondents have a quadratic utility function with normal errors. Each item j presents individual i with a choice between a “Yes” position and a “No” position. We assume that respondents’ policy preferences lie in a one-dimensional policy space. A preliminary test of this assumption is provided in Appendix B.Footnote ³ We estimate respondents’ ideal points using a Bayesian Item-Response (IRT) model (Clinton, Jackman, and Rivers Reference Clinton, Jackman and Rivers2004). In all, we estimate the ideal points of over 275,000 Americans.

Next, we estimate city-level policy conservatism by combining our individual-level data with a multilevel regression and poststratification (MRP) model (Park, Gelman, and Bafumi Reference Park, Gelman and Bafumi2004). This approach employs Bayesian statistics and multilevel modeling to incorporate information about respondents’ demographics and geography in order to estimate the public opinion of each geographic subunit. One way of thinking about an MRP model is to compare it to a weighted survey estimate that applies very finely tuned weightings, based on Census data, of specific demographic-geographic types. It estimates each individual’s response as a function of both demographic and geographic predictors. MRP models have been found to produce very accurate estimates of public opinion by state and congressional district with national samples of just a few thousand respondents (Lax and Phillips Reference Lax and Phillips2009; Warshaw and Rodden Reference Warshaw and Rodden2012).

To validate our estimates of city policy conservatism, we compare them with estimates of presidential vote share in each city derived from precinct-level election returns (Harvard Election Data Archive 2012). Our estimates of city policy conservatism are correlated with presidential vote share in the 2008 election at .77. This suggests that our estimates are accurately capturing cities’ policy preferences on a left-right continuum.Footnote ⁴

Figure 1 shows the policy preferences of the 51 cities with a population larger than 250,000 people. We find significant variation in the policy preferences of cities. Not surprisingly, we find that San Francisco, Washington, DC, and Seattle are three of the most liberal cities in the country. Mesa, AZ, Oklahoma City, OK, and Virginia Beach, VA are three of the most conservative cities.

FIGURE 1.

Mean Policy Conservatism of Large Cities

Note: This graph shows the mean policy preferences of citizens in cities with more than 250,000 people.

Explaining Variation in City Policies

In order to examine the relationship between city conservatism and policy outcomes, we construct a multilevel regression model that controls for a variety of political, economic, and legal factors. In particular, we are concerned about the effects of factors that contribute to a city’s governing capacity. Cities may be constrained by the resources available to them and the extent of the duties they can reasonably be expected to perform. Large cities or rich cities can be expected to be involved in more areas of public life than cities that lack the resources to engage in as many projects. In contrast, smaller, poorer cities may have simpler tax systems, fewer environmental regulations, and lower expenditures. For this reason we include controls for city population, the median income, and median housing value. We also control for the percentage of each city’s population that is African-American (percent black). (See Table 1 for descriptive statistics of all our variables).

TABLE 1.

Summary Statistics of Key Variables

It is also important to account for heterogeneity in the constraints facing municipal policymakers across states. Indeed, city governments only have powers allocated to them by state laws and constitutions. As a result, the state legal and political context can exert an important influence on municipal policy decisions. We account for state heterogeneity by including random effects for each state in our multilevel models.

While our multilevel model is well suited to examine the association between city conservatism and policy outcomes, it is difficult to interpret the results for the effect of institutions on representation. As we will see, responsiveness of policy to the mass public’s conservatism is nonlinear, often due to what seems to be a conservative “floor” (cities can only spend so little, for instance). A very simple linear model does an adequate job of summarizing responsiveness, but does not capture the potential differences across different institutional settings. As a result, a quadratic model would be preferred. However, this introduces five more interaction terms into our model, and the squared terms have to be interpreted in conjunction with the nonsquared terms. A simpler approach is to use a preprocessing method to approximate balance between institutional conditions, and then use simple nonparametric smoothing methods to show the effects in each institutional condition.

Preprocessing methods are attractive for analyses involving large datasets where balance can plausibly be established by reweighing the data to achieve balance between the treatment and control groups. Ho et al. (Reference Ho, Imai, King and Stuart2007) make the case that preprocessing reduces model dependence compared to standard ordinary least-squares methods. The large set of cities in our dataset allows us to achieve a high level of balance by reweighting our data on all our covariates (Hainmueller Reference Hainmueller2012).

For this analysis, we use entropy balancing matching (Hainmueller Reference Hainmueller2012).Footnote ⁹ Entropy balancing matching applies a maximum entropy reweighting scheme that calibrates unit weights so that the reweighted treatment and control group satisfy a potentially large set of prespecified balance conditions that incorporate information about known sample moments. In our analysis, for each institutional hypothesis, we balance our covariates’ first moments for cities with one institution, and cities with the other.Footnote ¹⁰ The “control” group may then be compared to the “treatment” group. This enables us to simply compare the slope of the relationship between outcomes and policy preferences for the two groups.Footnote ¹¹ We balance on median income, median home values, population, the presence of each institution, and city conservatism.

ARE CITY POLICIES RESPONSIVE TO THEIR CITIZENS?

What is the relationship between city policy outcomes and city conservatism? Figure 2 shows the relationship between citizens’ conservatism and four different city policies for the 1,600 cities in our dataset. Each panel shows scatter plots of a policy outcome on the y axis, and mean policy preferences of a given city on the x axis. The panel on the top left is our general policy scale. The panel on the top right shows the relationship between city conservatism and taxes per capita. On the bottom left, we show the association between city conservatism and the share of taxes that come from sales taxes.Footnote ¹² Finally, the bottom right plot shows the relationship between policy policy preferences and expenditures per capita. The top ten most populous cities are labeled in each panel with their official abbreviation. We also label Washington, DC, which is a notable outlier on most policy outcomes. The size of the circles representing each city are proportional to population. It is important to note, however, that the analyses are not weighted by population. Loess curves are fitted to each scatter plot.

FIGURE 2.

Responsiveness of City Policy to Public Ideology

Notes: Each circle represents a city, with diameter scaled to log population. The lines are unweighted loess curve fits. Two-letter abbreviations are shown for cities that are in in the top 10 most populous in America and the District of Columbia.

These simple bivariate relationships go consistently in the direction we would expect. On the general policy scale, cities with more liberal populations tend to get more liberal policy, and this relationship is remarkably strong and linear. Likewise, liberal cities collect more taxes per capita and have substantially higher expenditures per capita. Moreover, liberal cities have less regressive tax systems. The share of revenues that comes from sales taxes is lower in liberal cities and higher in conservative cities.

These effects are all large, with upward slopes that cover most of the span of the policy outcome. However, in the case of expenditures per capita and taxes per capita, the relationship is difficult to see due to a small number of high values that stretch the scale of the y axis. In Figure 3, we replot these policy outcomes, censoring the y axis far below the maximum values. In both cases the effect is large, but seems to taper off on the right side of the spectrum, perhaps indicating a minimum level of taxes and service provision that is supported by conservatives.

FIGURE 3.

Responsiveness—Zoomed

Notes: This figure repeats the panels on the right side of Figure 2 with a restricted y axis. A linear regression line and loess curve is shown in each panel.

Despite a strong bivariate relationship, a number of factors appear to moderate this relationship, as well they should. It appears from Figures 1 and 2 that more populous places have a tendency to be closer to the liberal side of the policy and public preferences, and that there may be a tendency for larger cities to adopt more liberal policies irrespective of preferences. We expect this to be the case, because liberal policy is usually associated with more government activity and larger cities have more capacity for activity. This may also be the case for richer cities.

In order to account for these possibilities, we run a multilevel regression model that includes possible confounders, such as the size, wealth, and ethnic diversity levels of each city (Table 2). Controlling for other factors that influence city policy outcomes leaves the core relationships that we find in our bivariate analysis relatively unchanged. City policy conservatism has a robust, statistically significant, and substantively important relationship with the type of policy that cities implement. These relationships are similar in models that account for possible confounders.

TABLE 2.

Association Between City Liberalism and Policy Outcomes

**p < 0.05; *p < 0.1.

In order to understand the substantive significance of these effects, it is important to consider the scale of the outcome policies under examination. How strong should we expect these relationships to be if democracy is very strong? One way to look at this question is to examine the size of the “errors” from our model for each policy outcome. Figure 4 shows the estimated distribution of errors from a Bayesian implementation of our multilevel model with uninformative priors. These histograms represent the estimated posterior distribution of the residuals for the model—in other words, the distribution of differences from the predictions of the model and the observed outcome variables. The median error in predicting our policy scale is 0.54 standard deviations. The median error in the predicted share of taxes from sales tax is just 12%. The median error for taxes per capita is only 195 dollars, and the median error for expenditures is 392 dollars. How should these errors be interpreted? Quite simply, they suggest that citizens could expect their per capita tax levy to be within 195 dollars of the taxes in similar cities with the same political preferences. In substantive terms, the relationship between preferences and outcomes is tight. This suggests that city governments are responsive to the preferences of their citizens.

FIGURE 4.

Error Distribution of Responsiveness Residuals

Note: These figures graph the posterior distributions of the prediction errors from four Bayesian models that correspond to the models 2, 4, 6, and 8 of Table 3.

DO INSTITUTIONS AFFECT RESPONSIVENESS IN CITIES?

Next, we examine our hypotheses about the impact of city institutions on responsiveness to public opinion. Table 3 shows the results of a simple multilevel regression with random effects for states. The key variables are the interactions between city conservatism and each institution. These interactions measure whether each institution is making cities more responsive to the preferences of their citizens. Our approach to estimating responsiveness rests on a simple premise: that the differences in responsiveness should produce differences in the slope of the relationship between policy preferences and policy outcomes. Greater slopes indicate greater responsiveness (Lax and Phillips Reference Lax and Phillips2012). However, due to the fact that the policy preferences measures and policy outcomes measures are not in the same space, higher slopes do not necessarily imply greater congruence or proximity between the preferences of citizens and city policy conservatism (Achen Reference Achen1977 Reference Achen1978; Matsusaka Reference Matsusaka2001). Instead, greater slopes could indicate overreactions to constituent preferences (Erikson, Wright, and McIver Reference Erikson, Wright and McIver1993, 93–4).Footnote ¹³ As a result, we will not be able to assess the degree to which any given city policy is a good match for the preferences of the citizens of that city. We also cannot say for certain whether an increase in slope is normatively good for representation. It could be that a particular institution causes an increase in responsiveness, but this increase in responsiveness reflects an overreaction to constituent preferences.

TABLE 3.

Effect of Institutions on Responsiveness

**p < 0.05; *p < 0.1.

Across all five institutions, we find no consistent statistically significant interactive effect between institutions and policy conservatism. Only two of these 20 coefficients are significant. We have few a priori reasons to suspect that these effects should vary over different policy outcomes. This suggests that institutions are having little effect on representation in municipal governments.Footnote ¹⁴

Next, we examine the results of our analysis of the impact of institutions on representation using data that is preprocessed using entropy balancing (Hainmueller Reference Hainmueller2012). For each institution, the unit weights are calibrated to ensure balance on all covariates. In our main text, we present nonparametric results graphically where sets of points are color coded, grey for cities with in one institutional condition, and black for cities in the other. In Appendix D, we show parametric results using the preprocessed data with fixed effects for each state.

Figures 5–9 show the results of this analysis. Each of the four panels has as its y axis the same four policy outcomes from Figure 2. Each plot shows a loess curve fitted to the resulting scatter plots of the matched data, with a separate curve for each institutional condition. The scatter plots for the relationship between each city policy outcome and mean city policy preferences are shown. For each institutional condition, a loess curve is drawn through the corresponding points, weighted to balance institutional conditions with unequal numbers of matches.Footnote ¹⁵

FIGURE 5.

Type of Government

Notes: The black points in each panel are municipalities with elected mayors and the grey points are municipalities with council manager governments. Each black line is the loess curve for the mayoral cities. The grey line is a weighted loess curve for the council-manager cities that uses entropy balancing matching weights. These weights are chosen to achieve balance between the two institutional conditions on median income, median home values, population, city conservatism and the use of partisan elections, ballot initiatives, term limits, and at-large districts.

A few preliminary observations are in order. First, the number of points in the top left panel is much smaller than the number of points in the other panels due to the relatively small number of cities that respond to the ICMA policy survey. As a result, there is a tendency for the curves in this panel to be the most different from each other due to random error. Likewise, within each panel, the curves are likely to differ most due to random error where the data is sparse, such as at the far right and far left of the graph.

Figure 5 shows the results of the matching analysis for the type of government: mayoral or council-manager. Cities with elected mayors are drawn in black, and cities with city managers are drawn in grey, as are the corresponding loess curves. Figure 5 shows a remarkable lack of difference between the curves across three of the four policy outcomes. The policy scale in the upper-left panel is the only outcome that shows any difference between cities with elected mayors and council-manager systems. Here, the black line is above the grey line, indicating that policy is slightly more conservative on average in cities with mayors. This is a statistically significant difference. However, this relationship does not hold up in the other analyses, and it does not indicate a difference in responsiveness per se. Other apparent differences, such as the far left side of the top right panel with expenditures as the dependent variable, are supported by very few data points. Overall, the responsiveness curves for mayor and council-manager cities are practically indistinguishable for per capita taxes, per capita expenditures, and the percentage of city revenues that come from sales taxes. Thus, our results provide little evidence for H2, our hypothesis that cities with directly elected mayors are more responsive. City manager systems, designed to be more professional and less political, appear to be just as responsive to public opinion as their mayoral counterparts. Given the same set of public policy preferences, a city with a mayor looks almost exactly the same as a city with a city manager for most policy outcomes.

This pattern of little institutional difference is continued in Figure 6, where cities with partisan elections are shown in black. The responsiveness curves are again very similar. Cities with partisan elections and cities without partisan elections appear to have roughly the same level of responsiveness. Differences in the curves are too small to attribute to systematic differences across institutions. Thus, we cannot conclude that whether cities have partisan or nonpartisan elections has an impact on the link between public opinion and public policy outcomes.

FIGURE 6.

Partisan Elections

Notes: The black points in each panel are municipalities with partisan elections and the grey points are municipalities with nonpartisan elections. Each black line is the loess curve for the partisan-election cities. The grey line is a weighted loess curve for the nonpartisan election cities that uses entropy balancing matching weights. These weights are chosen to achieve balance between the two institutional conditions on median income, median home values, population, city conservatism and the use of direct mayoral elections, ballot initiatives, term limits, and at-large districts.

Figure 7 shows the results for cities with (in black) and without (in grey) a popular initiative process. The overarching patterns are the same as for previous institutions. Across most of the support of the data, the estimated relationship between cities’ policy conservatism and their policy outcomes is nearly identical for cities with and without direct democracy. Overall, it appears to be the case that public views are about as well represented in cities where citizens are not able to vote on legislation at the ballot box as cities where they are.

FIGURE 7.

Direct Democracy

Notes: The black points in each panel are municipalities that allow ballot initiatives and the grey points are municipalities that do not. Each black line is the loess curve for the ballot initiative cities. The grey line is a weighted loess curve for the nonballot initiative cities that uses entropy balancing matching weights. These weights are chosen to achieve balance between the two institutional conditions on median income, median home values, population, city conservatism and the use of direct mayoral elections, partisan elections, term limits, and at-large districts.

Figure 8 shows the result of the matching analysis in the case of city council term limits. Cities with term limits are shown in black. Once again, the curves for cities with and without term limits are close to identical over most of the support of the data for each policy outcome. In this case the curves are particularly close, even at the extremes of the data. Overall, our findings provide no support for H5, that term limits lead to greater responsiveness.

FIGURE 8.

Term Limits

Notes: The black points in each panel are municipalities that have city council term limits and the grey points are municipalities that do not. Each black line is the loess curve for the term limit cities. The grey line is a weighted loess curve for the nonterm limit initiative cities that uses entropy balancing matching weights. These weights are chosen to achieve balance between the two institutional conditions on median income, median home values, population, city conservatism and the use of direct mayoral elections, partisan elections, ballot initiatives, and at-large districts.

Finally, Figure 9 shows the results of the matching analysis for cities with and without at-large districts. Cities with all at-large elections are shown in black, and cities with single-member districts are shown in grey. As with the other institutional conditions, there is little evidence of consistent variation in responsiveness across policy outcomes. The only notable results are that, contrary to H6, the slope of the relationship between citizens’ policy preferences and the share of taxes cities derive from sales taxes is different in at-large than district-based cities. In this case, cities with at-large elections are somewhat less responsive to their citizens’ policy preferences. But this difference in responsiveness is relatively small. There is also no difference in the relationship between public opinion and city policy conservatism in cities with and without at-large elections for our other three policy outcomes. Therefore, overall, our findings provide no support for H6, that at-large districts lead to better representation.

FIGURE 9.

At-Large Elections

Notes: The black points in each panel are municipalities that have at-large elections and the grey points are municipalities that do not. Each black line is the loess curve for the at-large cities. The grey line is a weighted loess curve for the mixed or single-member district cities that uses entropy balancing matching weights. These weights are chosen to achieve balance between the two institutional conditions on median income, median home values, population, city conservatism and the use of direct mayoral elections, partisan elections, ballot initiatives, and term limits.

Across five different policy outcomes, we find no evidence that any institution consistently affects representation in municipal government. These institutions, however, exist under conditions that are different from when many of them were formulated. Although the literature on representation in cities has emphasized their role in changing the politics of local government and breaking party monopolies, their influence on systematic representation today has been underexplored. Despite much attention to these institutions, whatever effect that they might have on policy responsiveness is too small for us to detect. Progressive reformers may not have been too surprised to learn that these institutions are not as influential today as they may have been in the past. After all, the party machines that these policies were designed to mitigate are long gone.