The City Has a Large Legislative Delegation

Perhaps big cities lost as much as they gained in the apportionment revolution of the 1960s. Indeed, the discrepancy in passage rates has grown with time. As Figure 1 shows, passage rates for big-city bills were lower in 1981 and 1997 than in any previous year in the study, and in 1981 passage rates were more than twice as high for bills affecting other places in the state than for those affecting big cities. The increased representation of metropolitan interests in state legislatures—while empowering city residents on an array of statewide issues and resulting in fairer transfers of wealth to cities and suburbs—may have harmed big-city delegations seeking passage of district bills.

Large delegations have more difficulty speaking with a single voice than do small, compact delegations. “In the case of a small city represented by a single legislator in each house, what passes for the will of the entire house is actually the will of the particular member,” according to McGoldrick (Reference McGoldrick1933, 2). “His party colleagues stand ready not only to accept his judgment as to all matters relating solely to his district but to enact it into law.” The observation that there are costs to increasing a delegation's size fits with patterns seen in congressional studies, including Kanthak and Krause's (2012) work on the representation of minority interests. Voting “cues” (Kingdon Reference Kingdon1973) presumably assume special importance in state legislatures, as Burns et al. (Reference Burns, Evans, Gamm and McConnaughy2009) argue, especially on low-information matters such as district bills. Legislators habitually support the district bills of their colleagues, but deference becomes impossible when the city's own legislators are at odds. According to this theory, size itself leads to division—and division can fatally undermine efforts to persuade others in the legislature to support a district bill. Disagreement within a delegation may come from a simple lack of coordination among the members themselves, from true conflict over policy or along party lines, or from the difficulty of distinguishing bills affecting small parts of a city from those dealing with the city as a whole.

Larger delegations, our data show, are more likely to be internally divided, and their bills are less likely to pass. Drawing on data we present in the next section, we considered two measures of delegation cohesion—party unity and roll-call votes. By both measures, as Figure 3 demonstrates, delegations grow more divided as they grow larger in size. One-person delegations are, of course, perfectly united on roll calls and in party identification. These measures of cohesion trend downward as delegations grow, falling sharply once there are at least 11 legislators in the delegation. The largest delegations are most sharply divided along party lines. Roll-call data, too, suggest greater division as delegation size grows. While the typical bill has passed or failed without a roll-call vote, and while most roll-call votes have been unanimous, delegation unity on roll calls corrodes steadily as delegation size becomes larger. As Figure 3 shows, not only do delegations become more divided as they grow in size but passage rates also fall dramatically. The changes in passage rates are smooth and consistent across each step of increasing delegation size—from 95% for bills introduced by one-person delegations to 29% for bills introduced by the largest delegations. Looking only at the roll-call votes, we find that division in the delegation predicts higher levels of opposition in the whole chamber (with a statistically significant correlation of 0.58). When a delegation divides, the entire floor is more likely to divide. But we need to look beyond roll-call analysis: the vast majority of bills fail because they never come to the floor for a vote, and the great majority that come to a vote pass, most of them unanimously, without roll calls. Because we predict that the effect of shifting from a one-person delegation to a two-person delegation will have a much greater impact on the unity and visibility of the group than the shift from a 51-person to a 52-person delegation, we use the natural log of delegation size in our regression models to test this hypothesis.

FIGURE 3.

Delegation Size Leads to Delegation Division and then to Defeat

Notes: Roll-call analysis based on 163 bills that came up for a roll-call vote; all other analyses based on the full sample of 747 district bills from the biggest city in each state. The “Largest Party Share of Biggest-City Delegation” reports the share of seats in the city's legislative delegation that is held by the largest faction in that delegation (either the Republican Party, the Democratic Party, or a third party).

The City Is Big and Threatening

City size alone may be sufficient to explain why big-city bills fail at high rates. Especially in states dominated by a very large city, “a bipolarization of attitudes may tend to develop,” V. O. Key (Reference Key1956, 230) wrote in American State Politics. “The outstate has its anxieties about metropolitan domination.” Sitting in the rural hinterland of a state—or in the suburban fringe or in another smaller city—legislators might look at the concentration of population and resources in the big city with jealousy and concern. Some cities easily dominated all other places in their states. San Francisco, for example, was home to 27% of all Californians in 1880, and a majority of all New Yorkers (7.5 million people, or 55% of the state) lived in New York City in 1940. Meanwhile, in other states, even their largest cities were tiny. Walla Walla, Washington, had a population of 3,588 in 1880, and Burlington, Vermont, had a population of 11,365 that year; even in 2000, Burlington's population was only 38,889. We drew on U.S. Census Bureau data reported in Gibson and Jung (Reference Gibson and Jung2005) for cities and in Haines (Reference Haines, Carter, Gartner, Haines, Olmstead, Sutch and Wright2006, 1-180–1-365) for states to compute each city's percentage of the state population, and expect that as it rises, legislative support for its bills will fall.

The City Is Alien and Different

Many studies demonstrate the propensity of individuals to favor members of their own ethnic, racial, or tribal group and to discriminate against outsiders (Bernhard, Fischbacher, and Fehr Reference Bernhard, Fischbacher and Fehr2006; Choi and Bowles Reference Choi and Bowles2007; Duggan Reference Duggan2006; Habyarimana et al. Reference Habyarimana, Humphreys, Posner and Weinstein2007; Kristof Reference Kristof2008; Putnam Reference Putnam2007). Trust and cooperation, this scholarship suggests, flourish more in homogeneous than heterogeneous environments. In many states, the populations of big cities look different from those in the rest of the state, not just larger—an emphasis of classic studies of the urban machine such as Bryce (Reference Bryce1891) and Banfield and Wilson (Reference Banfield and Wilson1963)—and we hypothesize that this difference brings distinctive policy demands and may even breed hostility from outstate representatives.

We measured this difference, first, by calculating the number of foreign-born people in each city relative to their proportion in the rest of the state. We gleaned data on nativity directly from statewide census volumes for each of the years under study. The demographic divergence between states and their leading cities could be great: foreign-born whites made up 30% of Chicago's white population in 1920, but just 11% of the rest of the state's white population, whereas modern-day Birmingham, like the rest of Alabama, is barely more than 1% foreign born. Figure 4 displays the demographic distances between metropolitan populations and states in 1900, arraying states by the size of the gap between the state and the big city in their foreign-born populations. This gap is small—and, we predict, big-city passage rates should be relatively high—in low-immigrant states with low-immigrant cities, such as Virginia (Richmond) and Alabama (Mobile), as well as in uniformly high-immigrant states such as Minnesota (Minneapolis) and Massachusetts (Boston). By contrast, far more residents of Chicago and San Antonio were foreign born than the residents of Illinois and Texas, respectively. We expect that this demographic divide makes the jobs of legislators from those cities especially difficult. Thus we hypothesize that the differences in nativity—between the relative numbers of foreign-born and native-born residents in the big city and the rest of the state—will result in lower passage rates for big-city bills, because we expect difference to lead to animosity.Footnote ²

FIGURE 4.

Foreign-Born Population Percentages in Biggest Cities and in States, 1900

To test another form of demographic difference, we calculated the number of African Americans in each city, again in proportion to the rest of the state. Although race is more salient than country of birth in defining difference in American society, we recognize that a simple count of numbers does not necessarily indicate political power. African Americans were disfranchised in the South for most of the period of our study, and their numbers in most northern cities were relatively small. Through the 1970s, the cities in our sample with the largest black populations were all located in the South—which was not a region where big-city representatives could be expected to represent black interests or, at least as important, be regarded by fellow legislators as tribunes for black interests. In incorporating African American urban populations in our model, we consider the possibility that African American populations matter only in the North and in the post-1965 South and even the possibility that only after 1965 do they matter politically in state legislatures anywhere. Separately, we identify every city in our sample with an African American mayor to test the hypothesis that bills from these cities might face greater hurdles to passage.

The City Is Harmed by Party Differences

Perhaps partisan disagreements disrupt the system of logrolls that characterizes the general treatment of district bills. Party tensions could emerge when the city's delegation is itself divided along party lines. They can also occur when the party controlling the city opposes the party controlling the rest of the state (McGoldrick Reference McGoldrick1933, 2). New York City, a Democratic metropolis, often confronted a hostile Republican state legislature, as Teaford (Reference Teaford1984, 95) and Bridges (Reference Bridges1984, 135) show, and Democrats in Chicago (Steiner and Gove Reference Steiner and Gove1960, 54) and Boston (Erie Reference Erie1988, 204) also faced antagonism from Republicans hailing from other parts of the state. Partisan differences can be vast. In 1941, when 88% of Boston's delegation was Democratic, 69% of the rest of the Massachusetts lower house was Republican. We expect that party differences—measured either at the individual or the aggregate level—will depress passage rates of big-city bills.

The City Introduces Too Many Bills

Perhaps the high failure rates of big-city bills are a function not of hostility but instead of the logistical challenges of managing a large body of legislation. In many cases, the largest cities make the greatest demands on the legislature, and the proliferation of bills from a great metropolis may itself be sufficient to undermine norms of deference to local legislation. Rather than seeking approval for a handful of targeted district bills, representatives from the big city may find their energies spread among many different priorities. At least at the extremes, in state legislatures with unusually high and low numbers of bills to consider, passage rates are negatively correlated with workload (Squire and Moncrief Reference Squire and Moncrief2010, 164). In some states, the number of big-city bills is immense—in 1981 the New York legislature considered 612 bills from New York City—whereas in other states the burden is much smaller. There were just 21 Detroit bills in the 1941 Michigan session and 5 Omaha bills in the 1961 Nebraska session. The great variance among cities on this dimension allows us to test the proposition that the number of big-city bills is negatively correlated with passage rates.

Do Big-City District Bills Really Fail at Higher Rates?

Although Figure 1 shows that bills from larger cities fare worse than those from smaller places, this apparent effect might be driven not by size but by other consequential differences between district bills from large and small cities. It might reflect a simple reaction against bills emanating from a state's largest city, no matter whether it was a major metropolis or not. Second, it could be driven by variation in the content of legislation. Finally, big-city bills may lose more often if they are more likely to be authored by members of the minority party or geographic faction in a statehouse. We outline each of these three alternative explanations of the pattern shown in Figure 1 and propose empirical measures (shown in boldface) to capture the causal factors they identify. Our multivariate model, which draws on our full set of 1,736 district bills, allows us to control for those factors and to hold constant other variables in order to isolate the impact of a city's population.

The first alternative explanation that we consider is that the poor performance of bills from big cities could be explained by a crude bias against the biggest city in each state. The city that looms the largest in a statehouse may stir resentment, even if it is relatively small in a national context. We take this into account by including in our analysis a dummy variable indicating that a bill comes from a state's largest city in a given session. Importantly, this variable does not perfectly correlate with city population. After all, in some states, such as Vermont and Montana, even the largest city is little more than a small town. By contrast, San Francisco (from 1921 onward), Los Angeles (before 1921), Dallas, St. Paul, and Buffalo are very large cities that do not rank as the largest in their state. This creates a divergence that allows us to distinguish between the effects of being a state's biggest city and a pure population effect.

Second, big-city bills may differ in some qualitative way from bills affecting smaller cities and towns. The needs of large cities are different not only in scale but also in substance from those of other places. Until recent decades, every large city supported large manufacturing centers and depended on a complex web of rail lines, ports, and power supplies to allow these industries to flourish. Large cities have also attracted great concentrations of people, requiring water and sewage lines on a scale unknown to the ordinary city or town in a state, as well as sophisticated systems of mass transit. Perhaps, then, big cities have historically sought bills that were substantively different from those sought by other localities, and perhaps this difference in bill content explains the high failure rate of bills from the biggest cities. To test this hypothesis, we analyze the content of bills introduced into the various state legislatures, distinguishing the bills from the biggest cities from all other district bills. The Online Appendix Table A1 lays out the categories that we developed to categorize each bill's content. We constructed 17 categories, read descriptions of each bill, and then identified its bill content according to one of these categories. As Online Appendix Figure A2 shows, we discovered marked similarities between the full set of district bills and the subsample of bills from the biggest cities. Therefore the poor performance of cities in these legislatures does not appear to be a product of differences in the substance of these bills, because there is little difference on average between the types of bills emanating from large and small cities. Still, to account for the impact that bill content might have on any individual piece of legislation, we include indicators for these content categories in our multivariate analysis.

The third alternative explanation we consider is that bills from the biggest cities fail because their representatives are distinctive in some fashion. Here we explore two possibilities: that the big-city representative introducing the bill is in the minority party or in the minority geographic faction. Whereas in the next section of this article, we analyze the impact of minority status on legislative success using our smaller sample of bills, in this section we consider the impact of this same factor in our full dataset of 1,736 district bills. Our strong prior assumption is that members of the majority party will give deference to each other's district bills, because they have little to gain by taking away the electoral tools of their co-partisans. By contrast, if they block district bills authored by members of the minority party, they could secure an electoral benefit for their party's challengers for these seats. If this logic prevails, majority status will lead to higher passage rates. Historically, many states have been fractured more by geographic differences than by partisan splits, so we also look at geographic factions to test the possibility that the identity of the person introducing the bill explains success or failure. Because geography creates a salient division within many legislatures, we expect that members of the largest geographic faction—urban, suburban, or rural—will discriminate against district bills introduced by their rivals for the same reason that the majority party might wish to block the minority. To test these two hypotheses, we used party rosters and information on district geography gleaned from an array of archival sources to code the introducer's party (and faction), determine which party (and faction) controlled the lower house of the legislature, and then identify whether the introducer is in the majority or minority party (and faction).

Our multivariate models include other characteristics of the states and sessions in which these district bills were introduced. Prior research (Gamm and Kousser Reference Gamm and Kousser2008) explores the causal linkages between these factors and the fate of all types of district bills in greater depth; here, we treat them as controls to be held constant. Chief among these are measures of the homogeneity of the legislature in partisan and geographic terms, and of the state in racial terms. Our analysis of all district bills found that they pass at sharply higher rates as the majority party's margin of control grows, with deference prevailing in one-party legislatures but collapsing when the parties are divided evenly. We also found that increases in the share of seats held by minor parties—another mark of division—lower passage rates. We use our partisan categorization of individual legislators to construct these aggregate measures. We also use our geographic coding of legislators to include the share of seats held by suburban members.Footnote ⁵ We draw on Census Bureau figures to measure the percentage of the state population that is African American in each year, because many state studies have found that diversity leads to more conservative policy outputs (Grogan Reference Grogan1994; Kronebusch Reference Kronebusch1993; Plotnick and Winters Reference Plotnick and Winters1985). To test the hypothesis that patterns of mutual deference collapse when legislative time horizons are shortened—an idea suggested by Axelrod and Hamilton's (Reference Axelrod and Hamilton1981) finding that cooperation in prisoner's dilemma games can be achieved when play is repeated—we include the level of legislative turnover (taken from Burns et al. Reference Burns, Evans, Gamm and McConnaughy2008, Table 1a) in our models. We control for the total number of bills introduced in a session, gathered from legislative archives, to account for the possibility that legislators forced to deal with many bills in a session will have less time to scrutinize each of them and be more likely to pass them all. We also control for the length of a legislative session, measured in legislative days using the archival sources first documented in the appendix to Gamm and Kousser (Reference Gamm and Kousser2007), as well as legislative salaries as measured in proportion to state per capita incomes in Burns et al. (Reference Burns, Evans, Gamm and McConnaughy2008, Table 4), expecting that more professional legislatures may have the time to pass more district bills of all types. Finally, we control both for whether or not a state grants broad home rule powers to its cities (Krane, Rigos, and Hill Reference Krane, Rigos and Hill2000) and for the level of malapportionment, measured by the ratio of the least populous counties’ voting power in the legislature divided by the power held by a state's biggest counties, averaged across a state's two houses in 1910 and calculated by David and Eisenberg (Reference David and Eisenberg1961, 15). Our expectations are that cities granted home rule will only come to the legislature with their more complex problems, and thus fare worse than cities without home rule, and that malapportionment should hurt big cities.

Table 1 reports the results of a logit model on our full sample of 1,736 district bills predicting whether a bill passes out of the lower house or fails at any prior point in the process. (In Online Appendix Table A2, we report very similar results from a parallel model explaining whether each bill became law.) Because the explanatory variables are measured at different levels—a bill's authorship, its content, and city size at the level of an individual bill; the legislature's party structure and other characteristics at the level of a legislative session—we use multilevel modeling techniques to acknowledge this data structure explicitly. Each bill is nested within the environment of a legislative session and is subject to the same subtle and unmeasured forces that affect other bills in the same session. We thus include random effects for the 87 sessions that we analyze here,Footnote ⁶ as well as state and year fixed effects to recognize the other hierarchies and potential for autocorrelation in our dataset, and estimate the model using the GLLAMM program (Rabe-Hesketh, Skrondal, and Pickles Reference Rabe–Hesketh, Skrondal and Pickles2004) in Stata 11.0.

The results of this estimation show that the performance gap between legislation from big and small cities, which is so clear in Figure 1, remains strong and significant when we hold constant three alternative explanations of this gap, along with variables capturing the systematic and idiosyncratic features of legislative sessions. Leaving the 513 bills from counties and other localities as the reference category, we look at how bills from cities of different sizes—whether or not they happen to be the biggest city in a state—fare. Bills from cities with populations less than 10,000 or between 10,000 and 100,000 residents do just as well as district bills from counties, the insignificant coefficients show. But bills from cities with more than 100,000 residents do much worse.Footnote ⁷ The coefficient on this category is negative and significant at the 99% confidence level. One way to get a sense of the substantive scale of this effect is to calculate an odds ratio, which reports the factor by which a one-unit increase in an independent variable either boosts or lowers the legislation's chances of passage.Footnote ⁸ The odds ratio for the City over 100,000 Residents indicator variable is 0.29, showing that bills from any city of this size are less than a third as likely to pass as bills from small cities or counties, all else equal. Perhaps paradoxically, but confirming the contention of a long literature in urban politics, major metropolises really do lose more often in state capitols when it comes to district bills.

Other results of our model reveal additional intriguing patterns. The insignificant coefficient on our indicator that a bill is from the state's biggest city shows that, for cities of a given size, a predominant position in a state poses no additional penalty. Bills from Burlington, the largest city in Vermont, do just as well as bills from other small cities, and bills from Houston (Texas's biggest city in the modern era) do no worse than bills from its rival city of Dallas. It is population size itself that matters. Using our measures of the policy areas addressed by each piece of legislation shows that some aspects of bill content matter. Bills giving power to a locality pass at higher rates, whereas bills sending resources either down to the city or county or back up to the state pass much less often. Still, because bills from the biggest cities propose the same sorts of policies contained in other district bills—as Online Appendix Figure A2 shows—holding constant the content of these bills does not explain away the big-city effect.

Neither does controlling for who introduces them. Surprisingly, district bills authored by a member of the legislature's majority party do no better than bills with a minority-party author.Footnote ⁹ Although one might expect the party cue to be powerful, this finding shows that the rule of universalism applies to district legislation in the states much as it does to district spending in the U.S. Congress (Fenno Reference Fenno1966; Ferejohn Reference Ferejohn1974; Schattschneider Reference Schattschneider1935). Similarly, bills authored by a member of the state's largest geographic group do no better, on average, than bills that come from other geographic factions, an additional form of universalism. Looking at the other control variables, we see that, consistent with earlier work on a more limited sample of district bills (Gamm and Kousser Reference Gamm and Kousser2008), district legislation is more likely to win when one party or geographic faction (urban, rural, or suburban) dominates a house, but lose more often when the percentage of seats held by minor parties rises. Other factors, such as the suburban seat share, the racial diversity of a state's population, legislative professionalism, membership turnover, home rule, and malapportionment, do not seem to matter. The primary lesson of this analysis is that the gap in performance between big-city bills and other district legislation displayed in Figure 1 persists, even with a myriad of controls in place.

TABLE 2.

Explaining Probability of House Passage, Subsample with Richer Data

Notes: N = 747. Table entries are estimated coefficients and (standard errors) from models with random coefficients for each legislative session, estimated using GLLAMM in Stata 11.0. * = p < .05 and ** = p < .01 in two-tailed tests. This model includes all of the bill content categories shown in Table 1, but we do not report coefficients for the sake of clear exposition.

Why Do Big-City Bills Fail So Often?

Now that we have shown that district bills from cities with larger populations fare worse—and that there is nothing distinctive about the largest city in a state—we ask why big-city bills fail at these high rates, drawing on our subsample of 747 bills. This sample provides the wide variation in population size, levels of representation in the statehouse, demographics, and partisan control that we need to test Hypotheses H1–H7. To perform this analysis, we estimate the same sorts of multilevel models of house passage that we did in the prior section and report a parallel model predicting whether a bill became law in Online Appendix Table A2. Again, the model includes state and year fixed effects, session-level random effects,Footnote ¹⁰ measures of bill content, and a host of substantive control variables.

The results reported in Table 2 present a clear explanation of why district bills from cities such as New York, Chicago, and Detroit fare so poorly: legislation from a city is more likely to lose when that city's legislative delegation is large and when the city houses many more immigrants than the state. The first result strongly confirms Hypothesis H1. The first row in the table reports a negative coefficient on the (logged) delegation size that is significant at the 99% confidence level. It is not driven by an outlier; this finding is robust to the exclusion of any of the cities in our dataset. One way to get a sense of the magnitude of this effect is to see how increases in the size of a city delegation change the predicted probability of a district bill's passage in a typical session.Footnote ¹¹Figure 5 presents these effects graphically, across the full range of variation in delegation sizes in our sample. It shows that when only one legislator represents a city, a typical district bill is likely to pass 80% of the time. Yet when the delegation size increases to two members, allowing the possibility of internal disagreement, the predicted probability of passage declines sharply to 64%, and then falls again to 53% with a delegation size of three. Passage rates continue to drop off as delegation size increases all the way to its maximum value in our data (the 69 house members representing Chicago in 1961). Yet the declines are steepest moving from one legislator to a handful, holding all other factors equal.

FIGURE 5.

Substantive Impact of City Delegation Size on Passage Rates, Using the Subsample with Richer Data

Our multivariate finding is buttressed by a look at the impact of the reapportionment revolution on a state-by-state basis. Reapportionment brought an exogenous shift in delegation size, holding population largely constant, and all of the cities that saw their delegation sizes jump between 1960 and 1980 also saw their legislative success plummet over that time period.Footnote ¹²

What logic drives this result? We show in Figure 3 that bigger delegations create more opportunity for internal division and that internal division leads to legislative defeat. The tight link between delegation size and passage rates buttresses that argument. It also suggests an answer to the puzzle of why, even after the reapportionment revolution, bills from big cities continue to lose more often than other district bills. Although the elimination of malapportionment gave big cities and other urban areas greater representation, it also saddled them with more representatives and thus increased the chances that their delegations might be internally divided. Bigger delegations may have been more likely to split on city bills than in the past, when only one or a handful of legislators represented the area, which may explain why their district bills fared so poorly in the 1981 and 1997 sessions.

Delegation size appears to be the main avenue through which large populations hurt big-city bills. We find no support for Hypothesis H2. Controlling for delegation size, cities that make up a larger share of the state's population do no worse than cities that are relatively smaller. We also include a dummy variable indicating that the state had a rival big city in that era, coded as 1 if the state's second-largest city was both (1) among the nation's 100 largest cities in the prior census and (2) at least half the size of the state's biggest city. This variable's impact also falls well short of significance, showing that Houston is not harmed by the presence of Dallas or Minneapolis by St. Paul. Rivalries over population size alone do not appear to drive the failure of big-city bills; it is delegation size that creates weakness in numbers.

Hypotheses 3 and 4 posit that demographic differences galvanize hostility to cities: the larger the difference between the populations of a state's biggest city and the state itself, either in immigration patterns or in race, the greater the hostility. We find evidence that district bills lose at higher rates when they emerge from cities with more foreign-born residents relative to the state. This effect is statistically significant and substantively strong. Figure 6 displays its magnitude, again holding all other explanatory variables constant at their mean (for continuous) or median (for dichotomous) values. It shows that a bill from a city that had a smaller proportion of immigrants in its population than the state—as was the case in Minnesota in 1881, where the difference in proportions was 2.3 percentage points—would pass 23% of the time. By contrast, in the situation where the city is much more immigrant-rich than the state—as in Illinois in 1881, where there was a 21.8 percentage point difference between Chicago's immigrant proportion and the state's—our model produces a predicted probability of passage of only 2%. (Each of these baseline probabilities of passage is relatively low because we hold the city's delegation size constant at its mean value, which is 19 in our sample, thus predicting low passage rates regardless of demographic factors.) This variable helps explain why bills from cities such as New York and Chicago, hubs of immigration relative to the rest of their states, fare more poorly than bills from cities such as Richmond and Minneapolis, which are microcosms of their states.

FIGURE 6.

Substantive Impact of Foreign-Born Population on Passage Rates, Using the Subsample with Richer Data

A parallel measure of racial difference, however, exerts no such effect, yielding a coefficient that is far from statistically significant. We also sought to capture racial prejudice by coding a variable indicating that the city had an African American mayor, but this factor is similarly insignificant (and even when we remove the mayoral variable from our model, the measure of racial demographic difference still falls well short of significance). In models not reported here, we probed for other potential impacts of race or ethnicity,Footnote ¹³ but found no evidence that it was an important obstacle to big-city success in our sample over this time period. In addition to testing for an effect of differences between city and state black populations, we (separately) tested for the impact of racial diversity in the city and in the state. We also tested each effect in constrained contexts—after the civil rights era (our 1981 and 1997 sessions) or in the subset of our cases where blacks had effective suffrage (in the South after the civil rights era and in other regions in all years). In none of these models was any measure of black population statistically significant.

Hypotheses 5 and 6 also yield null findings. The presence of partisan divides between metropolitan delegations and the legislature as a whole does not seem to lead to district bill defeats. This is the case whether partisan difference is measured at the individual or the aggregate level. Just as we saw in our analysis of all district bills, legislation from the biggest city authored by majority party members does not pass at significantly higher rates than bills introduced by minority party legislators. Neither do bills emerging from cities with delegations controlled by the same party that controls the legislature. States appear to practice universalism in this sphere, just as they do in distributive politics at the federal level. Finally, we see no evidence in favor of Hypothesis 7: the fates of individual district bills are not affected by the aggregate number of bills introduced by a city delegation.

Overall, the major lessons of this analysis are that cities lose in the statehouse when their residents appear foreign to other state legislators and when their delegations are too large to speak with a unanimous voice. Although we lack a direct measure of ethnic animus, nativism may be relevant to understanding the lower passage rates of big-city bills. But other measures of hostility—party divisions, bills introduced by members of the minority political party, large African American populations, many rural legislators—have no effect on passage rates. Large numbers of observers have pointed to partisan divides between big cities and their states, and many others have focused on rural–urban cleavages, but the evidence we present in this article leads us to reject both sets of explanations. Instead we embrace the finding that deference, rather than antagonism, characterizes the treatment of district bills. That big-city bills fail in such large numbers is due, above all, to large delegation sizes. Division within big-city delegations is the central variable explaining low passage rates.