Abstract

Previous research on segregation stresses things like urban form and racial preferences as primary causes. The author finds that an institutional force is more important: local land regulation. Using two datasets of land regulations for the largest U.S. metropolitan areas, the results indicate that anti-density regulations are responsible for large portions of the levels and changes in segregation from 1990 to 2000. A hypothetical switch in zoning regimes from the most exclusionary to the most liberal would reduce the equilibrium gap between the most and least segregated Metropolitan Statistical Areas by at least 35% for the ordinary least squares estimates.

Introduction

This paper sets out to explain two basic facts of segregation in the United States in recent decades. The segregation of Blacks remains everywhere higher than the segregation of immigrants, but the levels are converging. One potential explanation is anti-density zoning, which has two proposed effects on segregation. First, it increases interjurisdictional inequality and economic segregation in Metropolitan Statistical Areas (MSAs), and second, it curtails the exit of minority groups from segregated communities by limiting the supply of affordable housing in integrated areas.

Using two datasets of land regulation for the largest metropolitan areas, the results indicate that anti-density regulations are responsible for a large share of the observed patterns in segregation between 1990 and 2000. Minority groups are more segregated from Whites in metropolitan areas with prevalent exclusionary zoning no matter what their relative incomes and population sizes. This is the first paper to do a comparative analysis of how zoning affects metropolitan segregation across racial groups. It is also the first study to use metropolitan area fixed effects, albeit in a reduced sample of the twenty-five largest MSAs. The estimated effects are large enough that a hypothetical switch in zoning regimes from the most exclusionary to the most liberal would reduce the gap between the most and least segregated MSAs by at least 35% for the ordinary least squares (OLS) estimates. In addition to zoning, Tiebout-style incentives and relative income differences are strong predictors of segregation, though these are not robust across specifications and populations.

Figure 1 shows this basic trend looking at the three largest minority racial groups. Asian and Latino segregation from Whites has increased since 1980 in the largest metropolitan areas even as Black segregation has declined. Behind the rise of Asian and Latino segregation lie the experiences of new generations of immigrants. Segregation declined for American immigrants from 1920 to 1960 but has risen during the second half of the 20th century (Cutler, Glaeser, and Vigdor, 2006). Though lower than Blacks, the segregation of Asians and Hispanics from Whites is high at low and moderate levels of income, especially in the East and Midwest (Massey and Fischer, 1999). These trends are worrisome for many reasons. One is that teenagers and young adults who grow up in segregated communities fair much worse on a variety of measures (Massey and Denton, 1993; Cutler and Glaeser, 1997). Borjas (1998) finds strong evidence of neighborhood spillovers in “ethnic” (or nonnative) communities in the United States. More specifically, he finds that the skill level of one's ethnic group and the frequency of contact with that group in one's neighborhood have important and potentially detrimental consequences on the accumulation of education. Cutler, Glaeser, and Vigdor (2006) also find that segregation is strongly associated with a variety of negative outcomes for immigrant groups that have below average levels of formal education, though they find that the consequences may even be positive for highly skilled immigrant groups.

Figure 1.

Segregation Trends 1980–2000 for Fifty Largest Metros.

Figure 1.

Segregation Trends 1980–2000 for Fifty Largest Metros.

Review of Literature on Black Segregation

There are few papers that attempt to explain the segregation of Asian and Latin Americans and many that attempt to explain the segregation of Blacks.1 Before the 1960s civil rights era, Black segregation was easily explained by the laws and norms of the Jim Crow system, which, it must be noted, was prevalent in the North as well as the South (Massey and Denton, 1993). The segregation of Blacks has always been higher than the segregation of immigrants, and the intensity of anti-Black housing policy and practices has always been more pronounced (Massey and Denton, 1993). Black segregation reached a peak in the 1970s and was high in every large metropolitan area of the United States. It could be explained by what Cutler, Glaeser, and Vigdor (1999) call “collective action” on the part of Whites. Blacks paid more to live in segregated areas because housing there was scarce, and they were not allowed to leave, by force or law (Cutler, Glaeser, and Vigdor, 1999).

Racial zoning was used to directly exclude Blacks in many cities until the U.S. Supreme Court ruled it unconstitutional in Buchanan v. Warley in 1917 (Fogelson, 2005). Covenant agreements, which were restrictions on housing deeds that forbid their sale to Blacks (and other minorities), replaced racial zoning and were upheld in the Supreme Court case of Corrigan v. Buckley in 1925 (Fogelson, 2005). These were finally ruled unenforceable in Shelly v. Kraemer in 1948, but they persisted for a time thereafter (Massey and Denton, 1993).

Government policy exacerbated segregation by building segregated public housing, revitalizing potentially integrated neighborhoods to keep Blacks out, and by setting mortgage lending standards based on the racial composition of the neighborhood (Hirsch, 1983/1998; Massey and Denton, 1993).

Finally, real estate agents and their brokers refused to show homes to Blacks in White areas, and Blacks who moved into White neighborhoods were harassed (Hirsch, 1983/1998; Massey and Denton, 1993). This persisted, albeit at a lower level, until 1988 amendments to the Fair Housing Act in 1968 finally led to enforcement against housing discrimination (Massey and Denton, 1993).

Turning to the causes of post-civil rights era segregation, a number of explanations have been advanced. Some studies have entertained the possibility that Blacks prefer to self-segregate among other Blacks, as King and Mieszkowski (1973) argued, using data from New Haven. Galster (1982), however, found no evidence of this in two other cities, and a review article by Galster (1988) finds many reasons to doubt that Black racial preferences explain segregation. In a more recent study, Ihlanfeldt and Scafidi (2002) use survey responses from three large cities to find that Black preferences could explain, at most, 8% of the deviation from perfect integration in Atlanta and <2.8% and 4.4% in Detroit and Los Angeles. A more fundamental problem with using survey data is that hypothetical survey questions about neighborhood proportions ignore all the other attributes of that neighborhood, including its safety, the quality of its amenities, and its proximity to public goods, desirable businesses, and jobs. In other words, neighborhood racial preferences are strongly endogenous.

Galster (1982) and Cutler, Glaeser, and Vigdor (1999) find evidence that White discrimination is a significant source of segregation. Cutler, Glaeser, and Vigdor (1999) test this directly by analyzing individual-level data. They find that Whites pay significantly more to live away from Blacks, but Blacks do not pay more to live with Blacks. These results could be explained by White racial prejudice or omitted variables related to buying power, neighborhood quality, or zoning.

In a thorough review of the literature on housing discrimination, Ross (2008) cites a number of studies that have shown strong and convincing evidence that private housing market discrimination—that is, informal noninstitutionalized (or decentralized) discrimination—is still prevalent and that it played a large role in the rise of segregation. And yet, Ross (2008) also finds that there is no empirical evidence linking contemporary patterns of segregation to variation in housing discrimination, and indeed, what evidence exists suggests that it is no longer one of the most important contributing factors, but he discusses the possibility that the presence of minorities is used as a signal to negatively stereotype neighborhood quality, which leads to segregation if Whites are more sensitive to this signal than Blacks or Latinos. Ross (2008) concludes that asymmetric information based on historical neighborhood patterns is preventing further integration.

A related literature, which is also discussed by Ross (2008), tries to identify an integration “tipping point” beyond which Whites will eventually vacate a neighborhood after it reaches a minority threshold. Card, Mas, and Rothstein (2008) do find evidence that for every metropolitan area there is some minority share beyond which Whites move over subsequent decades. In a follow-up paper, they conclude that integration is only possible when there are few minorities. They write: “policies that are oriented toward maintaining stable neighborhoods can derive some justification from this result; these efforts need not contend forever against market forces that are pushing neighborhoods toward perfect segregation” (Card, Mas, and Rothstein, forthcoming). In other words, exclusionary zoning, which is designed to maintain stability and keep out poor people, may be the only way to maintain (low levels of) racial integration, which, when left to the market, would skyrocket.

While Card, Mas, and Rothstein (2008, forthcoming) never test the hypothesis that zoning causes racial segregation and explains its discontinuity, further research on tipping fails to find any evidence of it after 1970. Easterly (2009) explores the same dataset as Card, Mas, and Rothstein (2008, forthcoming) but concludes that Whites exited neighborhoods where there were initially more White shares than they did neighborhoods with more mixed shares. Easterly (2009) uses a more flexible functional form to reach this conclusion, showing that the results of Card et al. were driven by a misspecified relationship between racial composition and neighborhood dynamics. Overall, Easterly's work shows movement toward integration that is consistent with observed declines in Black–White segregation. Ross (2008) also reviews evidence inconsistent with Card, Mas, and Rothstein (2008).

There are also macroeconomic explanations advanced to explain segregation. Black desegregation in recent decades has accompanied metropolitan growth (Glaeser and Vigdor, 2001). The evidence suggests that the places where most Blacks live saw the least desegregation, but both Black population growth and overall population growth were highest in integrating areas. This is in sharp contrast to the experience of Asians and Latinos for which overall population growth and race-specific population growth were both correlated with higher levels of segregation from 1990 to 2000.2 Clearly, urban growth is not itself the key to understanding desegregation.

A different tradition, exemplified by Massey and Denton (1993), provides evidence that public policies, in addition to private discrimination, have continued to maintain barriers to Black integration. Pendall (2000) analyzes the consequences of local land-use regulations on the racial composition of jurisdictions. Using data from 1,168 jurisdictions across the twenty-five largest metropolitan areas, Pendall (2000) found that local governments employing low-density-only zoning had significantly fewer Hispanics and Blacks in 1980 and a decrease in Hispanics and Blacks from 1980 to 1990. A more comprehensive survey covering the fifty largest metropolitan areas was reported in Pendall, Puentes, and Martin (2006), with suggestive evidence that segregation is correlated with exclusionary zoning. Nelson, Sanchez, and Dawkins (2004) present evidence that containment regulations, which create an urban growth boundary, decrease Black segregation, but Rothwell and Massey (2009) cannot reject the null hypothesis that containment has no effect on segregation except through density zoning.

Berry (2001), on the other hand, rejects the hypothesis that zoning causes segregation. Using no measure of zoning, his method is to compare Houston, which uses covenants but not zoning, to Dallas, which uses zoning. He finds that Dallas is somewhat less segregated. However, his analysis is predicated on zoning per se rather than on anti-density regulation. He concludes that liberalizing zoning would not decrease segregation because alternative systems of exclusion such as covenants would take their place. A less narrow analysis draws the opposite conclusion. Two surveys of exclusionary land regulations—by Pendall, Puentes, and Martin (2006) and Gyourko et al. (2008)—both show that Dallas uses less exclusionary regulation than Houston. In the Pendall survey, the average jurisdiction in Houston's metropolitan area is 36–44% more exclusionary depending on whether one adjusts for land area. In the Gyourko survey, Houston is 120% more exclusionary in terms of anti-density regulation. The 2000 Census Bureau puts Houston as having 11–20% more Black–White segregation depending on whether one uses the dissimilarity index or the isolation index.

There is also judicial evidence, from a variety of sources, that zoning causes segregation (Orfield, 2002). In a seminal U.S. Supreme Court case, Village of Arlington Heights v. Metropolitan Housing Corporation (1977), the developer argued that it was denied rezoning for higher density racially integrated construction because of racial bias in the municipality's anti-density zoning laws—in violation of the 14th Amendment and the Fair Housing Act of 1968. In a 5–3 decision that reversed the Court of Appeals decision, the Supreme Court ruled that “protecting property values and the integrity of the Village's zoning plan” was a legitimate rationale and that the defendants must show intent to discriminate as to race.3 The Court of Appeals had disagreed, not finding the rational compelling or the means free from discrimination. It pointed out that the population in Arlington Heights would only be 5% Black if price alone were the deciding factor, and in reality the share was even lower, as the court stated:

According to the 1970 census, only 27 of the Village's 64,000 residents were Black. More specifically, the population of Arlington Heights in 1970 was 64,884, but only twenty-seven residents were Black. The four-township northwest Cook County area, of which Arlington Heights is a part, had a population increase from 1960 to 1970 of 219,000 people, but only 170 of these were Black. Indeed, the percentage of Blacks in this area actually decreased over this ten-year span while the percentage of the population in the entire Chicago metropolitan area that was Black increased from fourteen percent to eighteen percent (Arlington Heights v. Metropolitan Housing Corporation 1975).

Another illustrative example from the state courts is the New Jersey Supreme Court's Mt. Laurel decision (Southern Burlington County NAACP v. Township of Mount Laurel, 336, A. 2d 713 [N.J. 1975]), which ruled that jurisdictions must provide their fair share of affordable housing. The court reasoned that anti-density-only zoning precluded the residence of “low and moderate income” residents and thereby harmed the general welfare. The Mt. Laurel opinion detailed the varieties of exclusionary zoning, including, inter alia, minimum house or lot sizes, single-family-only zoning, and limitations on the number of school-aged children that a developer may allow to live in a multifamily complex. The court noted that in the outer-ring counties of Southern New Jersey, only 2.7% of residential land was zoned for multifamily housing and in some counties the share was <1%.

Surprisingly, although the local consequences of zoning have been studied a great deal, efforts to aggregate the analysis have been much more limited (see Fischel, 1985, for review) and almost never applied to segregation. Aggregation to the metropolitan level is advantageous theoretically and empirically. In the first instance, MSAs are proxies for economies, and so housing markets spillover across jurisdictions and regulations cluster together (Brueckner, 1998; Byun, Waldorf, and Esparza, 2005; Nguyen, 2009). Second, studies of the local estimates of zoning are going to have a selection bias unless the complicated dynamics of regulatory clustering and spillovers are modeled and measured appropriately.

Aggregation to the MSA level mitigates these selection issues. Rothwell and Massey (2010) use Pendall's dataset to show that density zoning is strongly associated with higher concentrations of poverty and greater interneighborhood median incomes (using a neighborhood Gini coefficient), and Rothwell and Massey (2009) show that Black segregation is higher in metropolitan areas with more stringent anti-density zoning and that zoning inhibited desegregation from 1980 to 2000. This paper builds on their analysis by assessing the segregation of other groups in relation to Blacks and directly controlling for the relative incomes of minority groups, as well as prejudice directed against them.

Literature on the Causes of Immigrant Segregation

One explanation for increasing Asian and Latino segregation is simply the recent wave of immigration. Using data from 1990 to 2000, Iceland and Scopilitti (2008) find little evidence that native-born Asians and Hispanics saw increasing levels of segregation, but in both cases foreign-born Asians and Hispanics were less likely to be living in neighborhoods with non-Hispanic Whites. For both groups, foreign-born residents were considerably more segregated than native-born residents, and older immigrants were less segregated than younger immigrants.

Another explanation is that the characteristics of new immigrants are different, including human capital levels and culture. Using detailed San Francisco data, Bayer, McMillan, and Rueben (2004) find that income and education strongly predict Hispanic segregation and Asian segregation. Similarly, Massey and Fischer (1999) find that income categories predict Asian and Hispanic segregation, and, especially in the Eastern and Midwestern United States, segregation is very high for Asian and Hispanics. In a comprehensive empirical study, Cutler, Glaeser, and Vigdor (2008) find that linguistic origins, social–economic status, years since migrating, and region of origin can explain roughly two-thirds of the increase in immigration. They find that only one-third can be explained by unchanging MSA effects or general year effects. These explanations, however, do not explain the increase in segregation that started before 1970.

Cutler, Glaeser, and Vigdor (2008) explore a third hypothesis that newer immigrant groups may prefer segregation more than older groups. According to their analysis, up until 1970, immigrants paid, on average, more to live in segregated metropolitan areas than nonsegregated ones, but by 2000 there is almost no evidence of this. Moreover, recent immigrants pay less in rent and housing to live in segregation areas, suggesting that if they could choose, immigrants would prefer integrated areas.

Cutler, Glaeser, and Vigdor (2008) also suggest a fourth hypothesis. They propose that changes in urban form, in terms of suburbanization and public transportation, partly explain the increase in segregation. They show evidence that immigrants rely more heavily on public transportation and are more likely to occupy older more urban areas than new suburbs. In general, the evidence suggests that immigrants are moving into areas that are undesirable to Whites.

The analysis of Cutler, Glaeser, and Vigdor (2006) is rich with analysis going back to 1920. Still, before concluding that the characteristics of immigrants explain most of the increase in segregation after 1970 and the changes in public transportation and suburbanization explain increases since the 1960s, two critical ideas come to mind. Cutler, Glaeser, and Vigdor (2006) use MSA fixed effects that control for unchanging MSA characteristics since 1920. Much has changed between 1920 and 2000 within MSAs that this will not capture, such as the structure of industry or, more broadly, economic opportunities for immigrants. Most relevantly perhaps, access to the suburbs has changed since the 1960s when civil rights legislation culminated in the Fair Housing Act of 1968. Glaeser and Gyourko (2005) document the rise of zoning since 1970 and the divorce between land prices and housing prices, even accounting for changes in housing quality and structure. This trend is also well documented by Fischel (2004) who posits that exclusionary zoning was widespread in the 1950s but became more restrictive in the 1970s. The emergence of exclusionary zoning could explain the finding of Cutler, Glaeser, and Vigdor (2006) that recent immigrants prefer integrated areas but cannot afford them.

There is another critical point that needs to be made with respect to the public transportation aspect of urban form. Fogelson (2003) and Jackson (1985) provide in-depth histories of public transportation and suburbanization. Both suggest that public transportation systems encouraged central city business concentration and the suburbanization of the middle class. Essentially, affluent Whites wanted to work and shop in central cities, but as central cities were becoming more industrialized, they did not want to live there. Public transportation encouraged White suburbanization by creating links to cities, and this facilitated the eventual erection of barriers to affordable housing in suburban jurisdictions. Zoning, therefore, is an omitted variable that is correlated with urban form, biasing the regressions of Cutler, Glaeser, and Vigdor (2008).

Theoretical Relationship between Density Zoning and Segregation

Density zoning is a function of the political economy of population growth. For analytic purposes, it is enough to consider two competing coalitions. One group, consisting of business owners and renters, is in favor of development. Population growth benefits local businesses by increasing demand for their products and benefits renters by increasing the supply of housing, thereby lowering rents. The other group, which consists of homeowners, is opposed to development. Population growth raises their property taxes and has an ambiguous effect on the value of their home. If housing population growth exceeds housing supply growth, homes appreciate, but if housing supply growth exceeds population growth, home values depreciate.

The relative strength of each group's preferences is conditioned by a third factor: the density of the area's population. In low-density areas, property taxes necessarily increase as population increases because there are decreasing returns to scale in the provision of public goods like education, roads, and sewers, but in high-density areas, such as cities, population growth can actually lower tax rates because the marginal costs of new residents are less than the gains in tax revenue from the new residents. In other words, with high density there are increasing returns. To summarize, in low-density areas, homeowners always prefer to restrict density, but for high-density areas, they either have no preference or lose out to business interests and renters. Business owners and renters always prefer a free market with respect to density. It follows from this that density zoning should be tighter in metropolitan areas with more low-density or rural settlements. As we will see, this is indeed the case.

Having established the motivation for density zoning, we need to consider its effect. Developers maximize profits by building wherever the expected return to development is highest. They can build only two types of housing, affordable housing or affluent housing, where the former is low density (e.g., one unit per acre) and the latter is high density (e.g., greater than or equal to eight units per acre). The profit to development from each housing unit can be modeled in the following simple way: 

(1)
graphic
This states that profits, π, conditional on the regulatory regime, are determined by revenue less costs, where costs consist of land and permit approval. Here p is the per-unit price of the property, with u representing the number of units, l representing the price per unit for land, and i representing the price per unit of influencing the town's regulators to permit the development. i is a measure of the relative strength of the two coalitions, and as such it is a function of population density and the relative size of the competing interest groups. R is the overall regulatory regime. When R equals 1, the jurisdiction allows only affluent housing or no development at all. When R = 0, there is a free market for housing in that developers can build high-density or low-density housing. The political costs of development, ui, can be thought of as a lump-sum bribe to homeowners, and they are relatively small when R = 0 because homeowners do not have sufficient power or motivation to block development.

R is determined by i such that if i reaches threshold x, R turns from 0 to 1. It follows from this that profits (πu) from affordable housing are always higher than profits for affluent housing, when land prices are equal and pu | affordable = pu | affluent. In reality, revenues are likely to be considerably higher for denser developments.4

In a metropolitan area with low-density jurisdictions, a dual market emerges and pronounced interjurisdictional inequality results. Towns zoned R = 1 become exclusively affluent. Towns zoned R = 0 become enclaves of affordable housing (or mostly affordable housing, conditional in part on i). Because of the profit-making opportunities for developers to build affordable housing when R = 0, especially when demand for affluent homes is low, new development in affordable enclaves will tend to be multifamily, and formerly affluent homes or industrial buildings will be converted into multifamily homes. Exacerbating the bifurcation of housing markets, affluent homeowners that remain in affordable enclaves will see property taxes increase to pay for schooling for families or other public services, and they will flee to exclusively low-tax affluent towns where R = 1. In these ways, anti-density zoning creates MSA-wide economic segregation. Racial segregation will follow economic segregation wherever incomes are negatively correlated with minority status.

There is also a dynamic channel connecting anti-density zoning to racial segregation, which operates by suppressing exit from ghettos (or segregated enclaves). It reduces the probability of exiting the ghetto by reducing the quantity and increasing the price of nonghetto housing. There are also more indirect channels by which segregation is likely to follow. Anti-density zoning suppresses economic growth and job creation making it more difficult for residents of poor neighborhoods to accumulate high enough incomes and savings to live in more affluent areas. Lower population growth is both a goal and a consequence of anti-density zoning, so employment opportunities in demand-sensitive industries (such as construction, retail, food service, and other service sector occupations) will be limited. These are also occupations that are disproportionately characteristic of minorities and immigrants. Finally, segregation is self-perpetuating. As cited above, minorities are less likely to increase their educational attainment and incomes when living in a segregated neighborhood or city, making it all the more difficult to live in an affluent neighborhood.

Data Sources and Methods

Table 1 lists the main variables by their data source. To measure segregation, the two most widely cited formulas are used: the dissimilarity index and the isolation index, as developed by Iceland, Weinberg, and Steinmetz (2002). Both are measured at the Census track level and reported for every metropolitan area by the Census Bureau. These measures of segregation, and others, are discussed in detail in Massey and Denton (1988). The published results will focus mostly on the dissimilarity index to conserve space, but the corresponding tables with the isolation index are available upon request.

Table 1.

Summary Statistics and Sources of Main Variables for Fifty Metropolitan Areas

 Source Mean Standard Deviation Minimum Maximum 
Average permitted density in MSA Pendall, Puentes, and Martin (2006) 3.39 0.68 2.17 4.67 
Infrastructure regulation Pendall, Puentes, and Martin (2006) 6.49 2.89 0.13 11.55 
Contain regulation Pendall, Puentes, and Martin (2006) 0.27 0.25 0.00 0.94 
Growth controls Pendall, Puentes, and Martin (2006) 0.08 0.09 0.00 0.38 
Inclusionary zoning Pendall, Puentes, and Martin (2006) 0.22 0.26 0.00 1.00 
Density restrictions index Gyourko et al. (2008) 0.22 0.19 0.00 1.00 
Wharton Land Use Regulation Index Gyourko et al. (2008) 0.05 0.55 −1.35 1.16 
Density zoning liberalized since 1994a Pendall, Puentes, and Martin (2006) 0.10 0.08 0.00 0.29 
Density zoning tightened since 1994a Pendall, Puentes, and Martin (2006) 0.08 0.07 0.00 0.33 
No. of general governments in 1962 Cutler and Glaeser (1995) 80.48 88.32 4.00 339.00 
Year of statehood U.S. Mint 1,825.84 35.44 1,788.00 1,912.00 
Rural units per acre 1990 U.S. Census 20.84 12.39 0.18 48.02 
Dissimilarity Latino U.S. Census 0.46 0.10 0.26 0.67 
Dissimilarity Black U.S. Census 0.62 0.11 0.36 0.85 
Dissimilarity Asian U.S. Census 0.39 0.06 0.26 0.51 
Isolation Latino U.S. Census 0.31 0.20 0.02 0.79 
Isolation Black U.S. Census 0.52 0.17 0.05 0.83 
Isolation Asian U.S. Census 0.14 0.12 0.04 0.52 
Latino/White income 2000 U.S. Census 0.50 0.09 0.35 0.75 
Black/White income 2000 U.S. Census 0.57 0.08 0.35 0.85 
Asian/White income 2000 U.S. Census 0.80 0.13 0.55 1.24 
Property tax rate (MSA) U.S. Census 0.01 0.00 0.00 0.03 
Suburban/central city housing units HUD 3.85 5.31 0.14 34.55 
Suburban/central city rents HUD 1.12 0.12 0.82 1.38 
% Local revenue from own tax base (state) U.S. Census of Governments 0.53 0.07 0.41 0.66 
Log state per capita taxation U.S. Census of Governments 7.52 0.22 7.18 8.00 
% Local revenue from property tax (state) U.S. Census of Governments 43.67 13.83 0.34 83.60 
Median rent/median housing price U.S. Census 0.49% 0.10% 0.22% 0.72% 
Log median household income 2000 U.S. Census 10.73 0.12 10.47 11.04 
Log housing price index 2000 Federal Housing Finance Agency 4.83 0.08 4.64 5.06 
Average July temperature National Climatic Data Center 77.45 5.97 61.30 92.80 
Average January temperature National Climatic Data Center 37.84 13.74 11.80 68.10 
% Adults with BA degree 1990 U.S. Census 0.23 0.04 0.14 0.32 
Log median rent 1990 U.S. Census 6.53 0.19 5.94 6.97 
 Source Mean Standard Deviation Minimum Maximum 
Average permitted density in MSA Pendall, Puentes, and Martin (2006) 3.39 0.68 2.17 4.67 
Infrastructure regulation Pendall, Puentes, and Martin (2006) 6.49 2.89 0.13 11.55 
Contain regulation Pendall, Puentes, and Martin (2006) 0.27 0.25 0.00 0.94 
Growth controls Pendall, Puentes, and Martin (2006) 0.08 0.09 0.00 0.38 
Inclusionary zoning Pendall, Puentes, and Martin (2006) 0.22 0.26 0.00 1.00 
Density restrictions index Gyourko et al. (2008) 0.22 0.19 0.00 1.00 
Wharton Land Use Regulation Index Gyourko et al. (2008) 0.05 0.55 −1.35 1.16 
Density zoning liberalized since 1994a Pendall, Puentes, and Martin (2006) 0.10 0.08 0.00 0.29 
Density zoning tightened since 1994a Pendall, Puentes, and Martin (2006) 0.08 0.07 0.00 0.33 
No. of general governments in 1962 Cutler and Glaeser (1995) 80.48 88.32 4.00 339.00 
Year of statehood U.S. Mint 1,825.84 35.44 1,788.00 1,912.00 
Rural units per acre 1990 U.S. Census 20.84 12.39 0.18 48.02 
Dissimilarity Latino U.S. Census 0.46 0.10 0.26 0.67 
Dissimilarity Black U.S. Census 0.62 0.11 0.36 0.85 
Dissimilarity Asian U.S. Census 0.39 0.06 0.26 0.51 
Isolation Latino U.S. Census 0.31 0.20 0.02 0.79 
Isolation Black U.S. Census 0.52 0.17 0.05 0.83 
Isolation Asian U.S. Census 0.14 0.12 0.04 0.52 
Latino/White income 2000 U.S. Census 0.50 0.09 0.35 0.75 
Black/White income 2000 U.S. Census 0.57 0.08 0.35 0.85 
Asian/White income 2000 U.S. Census 0.80 0.13 0.55 1.24 
Property tax rate (MSA) U.S. Census 0.01 0.00 0.00 0.03 
Suburban/central city housing units HUD 3.85 5.31 0.14 34.55 
Suburban/central city rents HUD 1.12 0.12 0.82 1.38 
% Local revenue from own tax base (state) U.S. Census of Governments 0.53 0.07 0.41 0.66 
Log state per capita taxation U.S. Census of Governments 7.52 0.22 7.18 8.00 
% Local revenue from property tax (state) U.S. Census of Governments 43.67 13.83 0.34 83.60 
Median rent/median housing price U.S. Census 0.49% 0.10% 0.22% 0.72% 
Log median household income 2000 U.S. Census 10.73 0.12 10.47 11.04 
Log housing price index 2000 Federal Housing Finance Agency 4.83 0.08 4.64 5.06 
Average July temperature National Climatic Data Center 77.45 5.97 61.30 92.80 
Average January temperature National Climatic Data Center 37.84 13.74 11.80 68.10 
% Adults with BA degree 1990 U.S. Census 0.23 0.04 0.14 0.32 
Log median rent 1990 U.S. Census 6.53 0.19 5.94 6.97 

Note: a This is the percentage of jurisdictions that reported changing density regulations by at least 10%.

Since the Census Bureau does not report results separately for immigrant groups, this study did not distinguish native-born Hispanics and Asians from immigrants when analyzing trends in segregation. The 2000 Census data show that 69% of Asians and 40% of Latinos are foreign born, as opposed to only 6% of Whites and Blacks. To adjust for potentially different effects, I simply control for the percentage of residents who are foreign born and the percentage of residents in the appropriate group.

The measure of zoning comes from a 2003 survey conducted by Pendall (see Pendall, Puentes, and Martin, 2006, for details) in fifty metropolitan areas. Pendall et al. mailed survey questions to every jurisdiction in the fifty largest metropolitan areas that had 2000 Census population >10,000. In total, 3,177 jurisdictions were surveyed and 58% responded. After dropping jurisdictions that did not answer a question related to density zoning, the final sample contained 1,677 local governments. The metropolitan areas contained in this final sample cover an average of 71% of their jurisdictions, with a range of 20–100%.

Maximum allowable density is the focus of this study because it has been identified as having the most important effects on housing supply growth and racial demographics at the local level (Pendall, 2000; Glaeser and Ward, 2006) and the MSA level (Rothwell and Massey, 2009). On the relevant Pendall survey question, the choices ranged from less than four units per acre, which earned a score of 1, to more than thirty units per acre, which earned a score of 5. A score of 2 allowed four to seven units per acre; a score of 3 allowed eight to fifteen units per acre; and a score of 4 corresponded to sixteen to thirty units per acre. Within each metropolitan area, the average score was simply computed across jurisdictions. Despite the fact that each one has a population larger than 10,000, some readers may question the decision to weigh each jurisdiction equally. The problem with alternative weighting strategies is that zoning will itself affect the boundary and population density of the jurisdiction (in so far as voter preferences on annexation are correlated with voter preferences for land-use policy). For comparison, the main results discussed below are reexamined conditional on the percentage of surveyed suburbs included in the metropolitan aggregation. In a second robustness check, the main results are replicated weighing each metropolitan area by the share of land included in the sample of jurisdictions. In other words, this alternative strategy gives more weight to observations for which a larger share of the metropolitan area's land mass was sampled. Finally, this is compared to a jurisdiction-level weighting strategy that aggregated zoning scores by the share of land area. These results are discussed below.

Alternative Land Regulations

Since density zoning is but one aspect of the local regulatory environment, other regulations in the Pendall data were also coded according to the classification system in Pendall, Puentes, and Martin (2006). First, a variable was created for the percentage of jurisdictions that use “growth controls,” in the form of a moratorium on permits and caps on permits. Six percent of jurisdictions used these measures. Second, for each jurisdiction, an infrastructure tax was calculated and then aggregated to the MSA level. This was computed by adding up the different types of infrastructure that are subject to either impact fees or adequate public facilities ordinances (APFOs) in each jurisdiction. These regulations essentially tax developers for some fraction of the capital costs of development. The average jurisdiction levied fees or provisional ordinances on seven different categories of infrastructure, including water, waste, schools, parks, public safety, storms, and transportation. These “taxes” reflect expectations that tax revenue from new developments and their residents will not cover the costs of new infrastructure, especially in small towns.

Third, the percentage of jurisdictions in the MSA that implement “inclusionary” zoning laws was calculated. These included affordable housing bonuses, density bonuses, and expedited permitting for affordable housing production. Other research has found that these regulations do not increase housing production (see Schuetz, Been, and Meltzer, 2008). Finally, a fourth variable was created to reflect “containment regulations.” These refer to the presence of growth boundaries, greenbelts, or sprawl control regulations, all of which attempt to limit the spread of development away from central cities in order to increase density in the central city and abutting suburbs (see Pendall, Puentes, and Martin, 2006, for more details about these regulations).

In addition to Pendall's survey, the analysis is replicated with a newer version of the Wharton Index used by Malpezzi (1996). The newer version was developed by Gyourko et al. (2008) and includes information about local political pressure, approval authority, the degree of local democracy, permit delays, development fees, open space criteria, state regulations, and supply restrictions. Gyourko et al. (2008) combine all these measures into a single comprehensive index called the Wharton Land Use Regulation Index. Although these data may lead to many important insights, strict approval authority and voting on zoning may be irrelevant for multifamily developers if the project is precluded by known anti-density regulation. Nonetheless, this variable is a useful control for regulations that are correlated with one another.

The survey of Gyourko et al. (2008) also includes specific questions about density regulations. These are compared to the measurement of anti-density from the Pendall survey. They are not measured the same way. For density, Pendall's survey asks “What is the Average Permitted Density in MSA in the jurisdiction?” Gyourko et al. (2008) take the opposite approach, asking if there are minimum lot size requirements above one acre anywhere in the jurisdiction. Pendall's survey may therefore understate regulation, while the survey of Gyourko et al. may overstate it. In practice, they are only moderately correlated (−0.30), but the relationship is statistically significant.

Three reasons explain the decision to primarily use the Pendall data. Its sample population was jurisdictions in MSAs rather than jurisdictions in general; so, aggregating to MSAs with the Pendall data yields representative MSA averages, but aggregating from the data of Gyourko et al. (2008) does not. Second, the Pendall data offer five distinct degrees of regulation, while the Gyourko et al. density restrictions index is a binary variable. Finally, the Pendall data can be merged with 1994 survey data for a subset of twenty-five large MSAs, allowing one to see how density zoning has changed over time. Moreover, even for the full set, the 2003 survey asked respondents to choose one of three options: that their density restrictions in 1994 were “about the same,” at least 10% more permissive, or at least 10% more restrictive. Rothwell and Massey (2009) discuss these answers in more detail and conclude that density zoning was very stable from 1988 to 2003. Most jurisdictions did not change at all, and those that did changed only slightly. This is true of the aggregated measures as well. The average MSA saw its jurisdictions increase the severity of their density regulations by roughly 0.31 points from 1994 to 2003, and only 25% (five out of twenty-five) of large MSAs saw density liberalized. Of the changes, 76% of MSAs saw the absolute value of their change fall within a standard deviation of all 1994 MSA regulations. These data and other key variables are summarized and their sources listed in Table 1.

Empirical Model

The first step in the empirical process will be to identify the sources of 2000 levels of segregation in a cross section of MSAs for Hispanics, Asians, and Blacks. The regressions will allow us to test some key aspects of the theory suggested above. Segregation will be modeled as a function of density regulation, minority income relative to income to Whites, demographic characteristics, and variables that could be correlated with both density regulation and segregation. Specifically, the model will look as follows: 

(2)
graphic

where Srmt indicates the level of segregation for racial group r, in metropolitan region m at time t; Zm is the density zoning regime for that metropolis; Rm is a vector of other zoning regulations; forumla is the ratio of average minority group income to White income; and Xmt is a vector of other variables that could cause sorting such as the number of general governments in the MSA and the degree of local government variation in finance. Tiebout (1956) famously argued that residents express their preferences for a given tax and service bundle by changing jurisdictions. This sorting effect is more pronounced (and efficient in his model) if there are many jurisdictions but is mitigated if states regularly transfer large shares of revenue to local governments. The vector also includes controls for the degree of suburbanization, and commuting patterns, which is what Cutler, Glaeser, and Vigdor (2008) call urban form.

To measure government fragmentation, the analysis relies on the number of general governments in the metropolitan area, using data organized from the 1962 U.S. Census of Governments by Cutler and Glaeser (1997). They use this variable as an instrument for segregation, and the older data are preferred to a contemporary measure because fragmentation is not neutral with respect to racial composition. Alesina, Baqir, and Hoxby (2004) find that the annexation of jurisdictions within counties is strongly affected by the presence of Black residents, suggesting that anti-Black motivations play a role in fragmentation. The older measure of fragmentation is highly correlated with a 2002 measure in any case.

Since relative income could be endogenous to segregation, the results will distinguish between parsimonious models that control only for demographic characteristics and fuller models that attempt to guard against omitted-variables bias. In other words, the regressions acknowledge that there is a trade-off between the problems of endogeneity and omitted variables.

In the OLS estimations, it is assumed that E(μt|Zm), but this will be violated if segregation, or omitted variables correlated with it, also influence density zoning. To account for this possibility, density zoning is modeled as a function of rural settlements. Rural settlements are most common in areas that have had the longest experience with English colonization and settlement; a proxy for this is the year of statehood. As people gradually diffused out from urban agglomerations into suburbs and small towns around cities, new rural jurisdictions were created. Jackson (1985) describes how transportation technology accelerated this process in the late 19th century. Figure 2 shows the very tight relationship between rural settlement and year of statehood (the correlation is −0.68). Either variable can be used successfully as an instrument so long as they have no direct effect on segregation. Since year of statehood is less likely to affect 2000 metropolitan characteristics than a 1990 measure of rural settlements, the models will give preference to the former. In support of the theory discussed above, Figure 3 shows the strong relationship between year of statehood and anti-density zoning. A simple regression of the two yields an F-statistic of 35.4 and an adjusted R2 of 0.42. The older a metropolitan area, the more restrictive is its zoning regime.

Figure 2.

Rural Settlement and Year of Statehood.

Figure 2.

Rural Settlement and Year of Statehood.

Figure 3.

Maximum Permitted Density and Rural Settlement.

Figure 3.

Maximum Permitted Density and Rural Settlement.

In the final section of this paper, I will also predict changes from 1990 to 2000 in segregation. This model will focus on covariates that are associated with population growth and demographic changes. 

(3)
graphic
The outcome variable is the log of the segregation index in 2000 for metropolitan area m and race r, and it is modeled as a function of the metropolitan area's zoning regime, previous segregation, changes in P, the minority population share, changes in the immigrant population share of any race (captured in F), climate variables that predict growth in the 1990s, represented by G, and demand characteristics associated with housing markets such as median income and rent, captured in D.

One problem is that this model neglects unobserved metropolitan area effects. Say φm represents a set of unobserved time-invariant metropolitan characteristics that affect segregation. These could be cultural or historic legacies of segregation and discrimination or the presence of geographic, governmental, or infrastructural impediments to integration. To add this term to an abbreviated version of Equation (2) would look like this: 

(4)
graphic
Here Xm,2000 is a vector of all time-varying control variables. Notice that now Z has a time subscript in Equation (4). Since Pendall provided his survey from 1994, I can measure how changes in zoning affect changes in segregation. These data use a similar method for the twenty-five largest MSAs, all of which are included in the 2003 survey (see Pendall, 2000).5 This technique relaxes the assumption that density is stable over time and allows a “before and after” specification that eliminates φm. Since φm is constant over time, it is subtracted out when a 1990 version of Equation (4) is subtracted from the 2000 equation. The new dynamic equation is displayed in Equation (5) and is robust to the presence of significant metropolitan effects that do not vary between 1994 and 2003: 
(5)
graphic

Results

Figure 4 plots segregation against an anti-density index. A perusal of the positive slopes goes a long way toward summarizing the main argument of this paper. Black segregation is significantly higher everywhere expect where anti-density zoning is the lowest or most liberal. In these MSAs, Black segregation is not significantly higher than Latino segregation, which overlaps with Asian segregation. On the other hand, all groups see an increase in segregation as density zoning increases, but the slope for Blacks and Latinos is steeper reflecting their lower relative incomes. In these MSAs, the no-exit effect is particularly harmful to Blacks since their previous levels of segregation were established during the Jim Crow era.

Figure 4.

Dissimilarity Index and Anti-density Zoning.

Figure 4.

Dissimilarity Index and Anti-density Zoning.

Table 2 shows a basic correlation matrix of some of the variables. The population growth of Hispanics and Asians was very weakly correlated with density zoning, but Blacks moved to areas with more liberal (higher) density zoning scores. This would be consistent with the finding of Cutler, Glaeser, and Vigdor (2008) that immigrants are moving to MSAs with depressed rather than booming markets (the growth in foreign-born population was similar). Latino population growth was associated with greater inequality between White and Latino incomes, while Asian and Black population growth occurred in MSAs where the relative incomes of Asians and Blacks were higher. Likewise, density zoning was positively or negatively correlated with relative income depending on the group. Overall, these raw correlations shed little light on the theoretical explanations for the divergent segregation patterns depicted in Figure 4.

Table 2.

Correlation Matrix of Permitted Zoning and Main Variables for Fifty Metropolitan Areas

 Permitted Density Dissimilarity Latino Dissimilarity Asian Dissimilarity Black Latino:White Income Asian:White Income Black:White Income Latino Population Increase Asian Population Increase 
Dissimilarity Latino −0.1882        
Dissimilarity Asian −0.3171 0.2608       
Dissimilarity Black −0.5415 0.3616 0.484      
Latino:White income −0.0838 −0.7608 −0.1205 0.0541     
Asian:White income −0.374 −0.3473 −0.0851 0.1956 0.6245    
Black:White income 0.1489 −0.2495 −0.3157 −0.4846 0.2923 0.4261   
Latino population increase 0.4926 0.274 −0.2507 −0.3799 −0.5119 −0.3721 −0.0667  
Asian population increase 0.3453 0.2013 0.3172 −0.1496 −0.4621 −0.5026 −0.1245 0.2462 
Black population increase −0.3407 −0.2307 −0.1173 0.1997 0.3623 0.0647 −0.1702 −0.4117 −0.3359 
 Permitted Density Dissimilarity Latino Dissimilarity Asian Dissimilarity Black Latino:White Income Asian:White Income Black:White Income Latino Population Increase Asian Population Increase 
Dissimilarity Latino −0.1882        
Dissimilarity Asian −0.3171 0.2608       
Dissimilarity Black −0.5415 0.3616 0.484      
Latino:White income −0.0838 −0.7608 −0.1205 0.0541     
Asian:White income −0.374 −0.3473 −0.0851 0.1956 0.6245    
Black:White income 0.1489 −0.2495 −0.3157 −0.4846 0.2923 0.4261   
Latino population increase 0.4926 0.274 −0.2507 −0.3799 −0.5119 −0.3721 −0.0667  
Asian population increase 0.3453 0.2013 0.3172 −0.1496 −0.4621 −0.5026 −0.1245 0.2462 
Black population increase −0.3407 −0.2307 −0.1173 0.1997 0.3623 0.0647 −0.1702 −0.4117 −0.3359 

OLS Results for 2000 Cross Sections

Table 3 reports a more systematic attempt to estimate the effect of density regulation, conditional on covariates that may be correlated with both density regulation and segregation. Using the dissimilarity index with respect to Whites, the results show that the higher the permitted density score, meaning the more units of housing are permitted for a given acre in the average jurisdiction of a metropolitan area, the less segregation there is for all three racial groups. This result is robust across parsimonious and robust sets of controls.

Table 3.

OLS Regression of 2000 Dissimilarity Index on Density Zoning

 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0882*** (0.0180) −0.0464*** (0.0136) −0.0696*** (0.0193) −0.0599** (0.0257) −0.0500*** (0.0168) −0.0606** (0.0239) 
% Minority population 0.355*** (0.127) 0.980*** (0.244) 0.241 (0.149) 0.373 (0.234) 1.143*** (0.243) 0.0909 (0.214) 
Population density 2000 4.92 × 10−5 (5.97 × 10−5−3.38 × 10−7 (3.61 × 10−59.85 × 10−5 (6.41 × 10−50.000100** (4.18 × 10−54.84 × 10−5 (4.30 × 10−58.16 × 10−5 (4.99 × 10−5
% Foreign-born 2000 0.253 (0.217) −0.00583 (0.102) 0.324** (0.140) −0.271 (0.261) −0.0700 (0.121) −0.103 (0.154) 
General governments 1962 −0.000133 (0.000195) 0.000163** (7.58 × 10−50.000427*** (0.000158) −3.74 × 10−6 (8.31 × 10−50.000174** (7.75 × 10−50.000275 (0.000173) 
Property tax rate 2000    0.181 (2.287) 3.066 (2.305) 7.906** (3.258) 
Local government's share of revenue    0.0480 (0.133) −0.0414 (0.150) 0.176 (0.161) 
Log per capita state taxation    0.0853 (0.0712) −0.0227 (0.0574) −0.0654 (0.0935) 
Property tax share of local revenue    −0.000756 (0.000919) −0.00148* (0.000837) −0.00219** (0.00107) 
Median home rent/median home value    −1.828 (16.52) 2.055 (16.84) −52.59** (23.93) 
Log median household income 2000    1.28 × 10−6 (2.22 × 10−6−1.79 × 10−6 (1.35 × 10−65.65 × 10−7 (1.99 × 10−6
Log housing price index    0.0547 (0.115) 0.00227 (0.119) 0.0529 (0.176) 
Minority income/White income    −0.836*** (0.151) −0.0755 (0.0872) −0.533* (0.287) 
Suburban housing/central city housing    0.000588 (0.00128) 0.000528 (0.00159) 0.00411 (0.00256) 
Suburban rent/central city rent    0.0566 (0.0812) −0.157*** (0.0578) 0.117 (0.0986) 
% Minorities using public transportation or carpooling    −0.0135 (0.182) 0.0248 (0.206) −0.171 (0.220) 
Constant 0.679*** (0.0734) 0.497*** (0.0499) 0.697*** (0.0777) 0.0661 (1.043) 0.914 (0.911) 1.361 (1.608) 
Joint F-test on urban form    0.19* 3.12** 1.61 
Joint F-test on Tiebout variables    0.44 2.98** 3.62*** 
Observations 50 50 50 50 50 50 
Adjusted R2 0.323 0.390 0.533 0.739 0.416 0.644 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0882*** (0.0180) −0.0464*** (0.0136) −0.0696*** (0.0193) −0.0599** (0.0257) −0.0500*** (0.0168) −0.0606** (0.0239) 
% Minority population 0.355*** (0.127) 0.980*** (0.244) 0.241 (0.149) 0.373 (0.234) 1.143*** (0.243) 0.0909 (0.214) 
Population density 2000 4.92 × 10−5 (5.97 × 10−5−3.38 × 10−7 (3.61 × 10−59.85 × 10−5 (6.41 × 10−50.000100** (4.18 × 10−54.84 × 10−5 (4.30 × 10−58.16 × 10−5 (4.99 × 10−5
% Foreign-born 2000 0.253 (0.217) −0.00583 (0.102) 0.324** (0.140) −0.271 (0.261) −0.0700 (0.121) −0.103 (0.154) 
General governments 1962 −0.000133 (0.000195) 0.000163** (7.58 × 10−50.000427*** (0.000158) −3.74 × 10−6 (8.31 × 10−50.000174** (7.75 × 10−50.000275 (0.000173) 
Property tax rate 2000    0.181 (2.287) 3.066 (2.305) 7.906** (3.258) 
Local government's share of revenue    0.0480 (0.133) −0.0414 (0.150) 0.176 (0.161) 
Log per capita state taxation    0.0853 (0.0712) −0.0227 (0.0574) −0.0654 (0.0935) 
Property tax share of local revenue    −0.000756 (0.000919) −0.00148* (0.000837) −0.00219** (0.00107) 
Median home rent/median home value    −1.828 (16.52) 2.055 (16.84) −52.59** (23.93) 
Log median household income 2000    1.28 × 10−6 (2.22 × 10−6−1.79 × 10−6 (1.35 × 10−65.65 × 10−7 (1.99 × 10−6
Log housing price index    0.0547 (0.115) 0.00227 (0.119) 0.0529 (0.176) 
Minority income/White income    −0.836*** (0.151) −0.0755 (0.0872) −0.533* (0.287) 
Suburban housing/central city housing    0.000588 (0.00128) 0.000528 (0.00159) 0.00411 (0.00256) 
Suburban rent/central city rent    0.0566 (0.0812) −0.157*** (0.0578) 0.117 (0.0986) 
% Minorities using public transportation or carpooling    −0.0135 (0.182) 0.0248 (0.206) −0.171 (0.220) 
Constant 0.679*** (0.0734) 0.497*** (0.0499) 0.697*** (0.0777) 0.0661 (1.043) 0.914 (0.911) 1.361 (1.608) 
Joint F-test on urban form    0.19* 3.12** 1.61 
Joint F-test on Tiebout variables    0.44 2.98** 3.62*** 
Observations 50 50 50 50 50 50 
Adjusted R2 0.323 0.390 0.533 0.739 0.416 0.644 

Notes: Robust standard errors in parentheses, clustered on state. Urban form variables: % using public transportation/carpooling; % minorities using public transportation/carpooling; suburban–city housing ratio; suburban–city rent ratio. Tiebout variables: number of general governments, log property tax rate, local government's share of local revenue (by state); log per capita state income, local revenue from property tax.

*P < 0.1, **P < 0.05, ***P < 0.01.

The first three columns use only a parsimonious set of demographic controls, including the relevant minority share of the population, population density, the foreign-born share of the population, and a measure of government fragmentation. These variables alone explain a remarkably high share of the variation: 32–52%.

Columns 4–6 add to the regressions a number of control variables including a measure of relative socioeconomic status, housing price variables, a set of urban form variables, and a set of Tiebout-related variables. The association between permitted density and segregation remains negative and significant. To put the effect in context, imagine that the most restrictive MSA (Buffalo with a score of 2.17) liberalized density enough to equal the most liberal (San Diego with a score of 4.67). On a 0–1 scale, the model would predict segregation scores 0.14 points lower for Asians and Blacks and 0.15 points lower for Latinos. This would, in turn, reduce the gap between the most segregated and the least segregated MSAs by 58%, 29%, and 36%, respectively.

For Latinos, relative socioeconomic status—measured as a per capita income ratio—is strongly correlated with segregation. The relationship is marginally significant for Blacks and insignificant for Asians. The effects for Latinos and Blacks are considerably larger than the zoning effect. For example, in Pittsburgh, Latinos per capita income was 75% of White per capita income, but in Los Angeles it was just 34%. If Latinos in Los Angeles were exogenously brought up to the same relative standing as those in Pittsburgh, the segregation score would fall by 0.34 points, accounting for 84% of the observed gap. Blacks in Salt Lake City earned 85% of what Whites earned but only 34% in West Palm Beach. If this income gap were closed, Black segregation would fall by 0.26 points, accounting for 47% of the observed segregation gap. For Asians, the association with relative income is insignificant, but the effects from deregulating zoning would be greater in any case. Before concluding, however, that income explains a large share of segregation, we must consider that segregation also causes income inequality between minorities and Whites. In other words, reverse causality is biasing the coefficient on income, making it larger in absolute value.

Table 3 also includes controls for what Cutler, Glaeser, and Vigdor (2008) call urban form. These variables are the percentage of minority commuters (for the minority group in question) using public transportation or carpooling, the extent of suburbanization, and the price differences between rents in the suburbs and central cities. Using either public transportation or carpooling requires living in a specific location, and this could lead to segregation in areas where minorities disproportionately rely on public transportation or their neighbors to get to work, assuming that this form of collective action is easier within ethnic groups (which is certainly debatable). As mentioned above, public transportation systems are also caused by segregation because they were proposed by downtown business interests to counteract White flight (Fogelson, 2003). The evidence presented here does not support the idea that urban form is associated with segregation. The use of public transportation by a minority group does not predict that group's level of segregation. Indeed, the sign is negative for Blacks and Latinos. Suburban–central city rent differentials, likewise, do not explain the segregation of Blacks and Latinos, but higher suburban rents are associated with less Asian segregation. Overall, a joint F-test shows no relationship to segregation for any of the groups.

The Tiebout (1956) variables are collectively associated with higher segregation of Asians and Blacks. These variables include the degree of government fragmentation in the MSA, the property tax rate for the MSA, and state variables for total taxation per capita, the average local government's reliance on property taxation, and the degree that local government's rely on their own tax sources for revenue. Government fragmentation is significantly predictive of higher segregation in both models for Asians and Blacks, but it is not significant for Latinos. The effective property tax rate was strongly associated with segregation for Blacks but no other groups.

Other variables included in the model account for housing market characteristics such as prices and a measure of competitive housing prices (rents divided by prices). These variables had no consistent relationships across racial groups, but a measure of competitive housing was negatively associated with minority segregation generally, and significantly so for Blacks. Inflated markets, which may be caused by supply restrictions like zoning or the overdevelopment of desirable land, may price minorities out of mostly White neighborhoods.

Adjusting for Alternative Regulations

The previous regressions focused on anti-density regulation in isolation from other land regulations, though they are correlated with density zoning and may plausibly cause segregation by increasing (or decreasing) the cost of housing. Tables 4 and 5 address this omission by first including the various regulatory categories discussed and documented in Pendall, by Puentes, and Martin (2006). Second, a comprehensive regulatory index by Gyourko et al. (2008) is included to proxy for the regulatory environment by generally.

Table 4.

OLS Regressions of 2000 Segregation on Land Regulations

 White Minority Dissimilarity Index 2000 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.101*** (0.0299) −0.0443*** (0.0162) −0.0721** (0.0274) −0.0451* (0.0263) −0.0246 (0.0222) −0.0218 (0.0357) 
Average no. of impact fees and APFOs (MSA) −0.00426 (0.00436) 0.00229 (0.00292) 0.00219 (0.00499) −0.00435 (0.00656) −0.00923* (0.00518) −0.00562 (0.00930) 
% Juris with urban boundaries, greenbelts, or sprawl controls 0.00200 (0.0668) −0.0615 (0.0412) −0.0348 (0.0510) −0.0630 (0.0582) −0.112*** (0.0415) −0.153** (0.0715) 
% Juris with limits on housing permits 0.118 (0.120) 0.112 (0.137) 0.0796 (0.178) 0.183* (0.107) 0.145 (0.134) 0.0491 (0.155) 
Inclusionary zoning 0.0830 (0.0774) −0.000702 (0.0464) 0.0148 (0.0552) 0.0164 (0.0629) 0.0938 (0.0660) 0.00533 (0.0641) 
% Minority population 0.564*** (0.193) 1.013*** (0.295) 0.243 (0.177) 0.343 (0.265) 0.753** (0.295) 0.0363 (0.255) 
Population density 2000 5.09 × 10−5 (6.25 × 10−5−8.66 × 10−6 (3.73 × 10−59.44 × 10−5 (6.83 × 10−59.31 × 10−5* (5.22 × 10−51.00 × 10−5 (3.74 × 10−55.46 × 10−5 (6.64 × 10−5
% Foreign-born 2000 0.00919 (0.336) −0.0580 (0.0996) 0.280** (0.123) −0.284 (0.287) −0.132 (0.117) −0.284* (0.153) 
No. of general governments, 1962 −2.98 × 10−5 (0.000201) 0.000142* (8.16 × 10−50.000421** (0.000178) 2.57 × 10−5 (0.000113) 0.000212* (0.000117) 0.000168 (0.000165) 
Property tax rate 2000    2.549 (2.185) 7.788*** (2.454) 9.393** (3.778) 
Local government's share of revenue    −0.0314 (0.132) −0.234 (0.167) 0.0380 (0.198) 
Log per capita state taxation    0.00786 (0.0912) −0.166** (0.0678) −0.161 (0.126) 
Property tax share of local revenue    −0.00101 (0.000981) −0.00215*** (0.000797) −0.00277** (0.00133) 
Median home rent/median home value    −6.655 (19.29) −11.18 (17.04) −66.50** (25.71) 
Log median household income 2000    9.09 × 10−7 (2.32 × 10−6−2.41 × 10−6* (1.31 × 10−6−3.65 × 10−7 (1.83 × 10−6
Log housing price index    0.0508 (0.113) 0.0127 (0.104) 0.0261 (0.167) 
Minority income/White income    −0.871*** (0.146) −0.0928 (0.0817) −0.663** (0.321) 
Suburban housing/central city housing    0.000633 (0.00149) 0.000169 (0.00172) 0.00433 (0.00276) 
Suburban rent/central city rent    0.0563 (0.0762) −0.112* (0.0640) 0.152 (0.103) 
% Minorities using public transportation or carpooling    −0.0138 (0.172) 0.146 (0.205) −0.0182 (0.259) 
Constant 0.716*** (0.0841) 0.494*** (0.0512) 0.698*** (0.0829) 0.717 (1.233) 2.036** (0.919) 2.344 (1.761) 
Observations 50 50 50 50 50 50 
Joint F-test of regulations 8.07*** 4.20*** 3.71** 1.42 5.15*** 2.57** 
Joint F-test of urban form    0.23 2.12 1.11 
Joint F-test of Teibout    0.53 3.60*** 2.18* 
Adjusted R2 0.301 0.392 0.497 0.742 0.511 0.661 
 White Minority Dissimilarity Index 2000 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.101*** (0.0299) −0.0443*** (0.0162) −0.0721** (0.0274) −0.0451* (0.0263) −0.0246 (0.0222) −0.0218 (0.0357) 
Average no. of impact fees and APFOs (MSA) −0.00426 (0.00436) 0.00229 (0.00292) 0.00219 (0.00499) −0.00435 (0.00656) −0.00923* (0.00518) −0.00562 (0.00930) 
% Juris with urban boundaries, greenbelts, or sprawl controls 0.00200 (0.0668) −0.0615 (0.0412) −0.0348 (0.0510) −0.0630 (0.0582) −0.112*** (0.0415) −0.153** (0.0715) 
% Juris with limits on housing permits 0.118 (0.120) 0.112 (0.137) 0.0796 (0.178) 0.183* (0.107) 0.145 (0.134) 0.0491 (0.155) 
Inclusionary zoning 0.0830 (0.0774) −0.000702 (0.0464) 0.0148 (0.0552) 0.0164 (0.0629) 0.0938 (0.0660) 0.00533 (0.0641) 
% Minority population 0.564*** (0.193) 1.013*** (0.295) 0.243 (0.177) 0.343 (0.265) 0.753** (0.295) 0.0363 (0.255) 
Population density 2000 5.09 × 10−5 (6.25 × 10−5−8.66 × 10−6 (3.73 × 10−59.44 × 10−5 (6.83 × 10−59.31 × 10−5* (5.22 × 10−51.00 × 10−5 (3.74 × 10−55.46 × 10−5 (6.64 × 10−5
% Foreign-born 2000 0.00919 (0.336) −0.0580 (0.0996) 0.280** (0.123) −0.284 (0.287) −0.132 (0.117) −0.284* (0.153) 
No. of general governments, 1962 −2.98 × 10−5 (0.000201) 0.000142* (8.16 × 10−50.000421** (0.000178) 2.57 × 10−5 (0.000113) 0.000212* (0.000117) 0.000168 (0.000165) 
Property tax rate 2000    2.549 (2.185) 7.788*** (2.454) 9.393** (3.778) 
Local government's share of revenue    −0.0314 (0.132) −0.234 (0.167) 0.0380 (0.198) 
Log per capita state taxation    0.00786 (0.0912) −0.166** (0.0678) −0.161 (0.126) 
Property tax share of local revenue    −0.00101 (0.000981) −0.00215*** (0.000797) −0.00277** (0.00133) 
Median home rent/median home value    −6.655 (19.29) −11.18 (17.04) −66.50** (25.71) 
Log median household income 2000    9.09 × 10−7 (2.32 × 10−6−2.41 × 10−6* (1.31 × 10−6−3.65 × 10−7 (1.83 × 10−6
Log housing price index    0.0508 (0.113) 0.0127 (0.104) 0.0261 (0.167) 
Minority income/White income    −0.871*** (0.146) −0.0928 (0.0817) −0.663** (0.321) 
Suburban housing/central city housing    0.000633 (0.00149) 0.000169 (0.00172) 0.00433 (0.00276) 
Suburban rent/central city rent    0.0563 (0.0762) −0.112* (0.0640) 0.152 (0.103) 
% Minorities using public transportation or carpooling    −0.0138 (0.172) 0.146 (0.205) −0.0182 (0.259) 
Constant 0.716*** (0.0841) 0.494*** (0.0512) 0.698*** (0.0829) 0.717 (1.233) 2.036** (0.919) 2.344 (1.761) 
Observations 50 50 50 50 50 50 
Joint F-test of regulations 8.07*** 4.20*** 3.71** 1.42 5.15*** 2.57** 
Joint F-test of urban form    0.23 2.12 1.11 
Joint F-test of Teibout    0.53 3.60*** 2.18* 
Adjusted R2 0.301 0.392 0.497 0.742 0.511 0.661 

Notes: Robust standard errors in parentheses, clustered on state. Urban form variables: % using public transportation/carpooling; % minorities using public transportation/carpooling; suburban–city housing ratio; suburban–city rent ratio. Tiebout variables: number of general governments, log property tax rate, local government's share of local revenue (by state); log per capita state income, local revenue from property tax.

*P < 0.1, **P < 0.05, ***P < 0.01.

Table 5.

OLS Regressions of 2000 Dissimilarity Index on Permitted Density and Wharton Index

 White Minority Dissimilarity Index 2000 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0927*** (0.0182) −0.0451*** (0.0123) −0.0680*** (0.0186) −0.0557** (0.0265) −0.0437** (0.0164) −0.0508** (0.0232) 
2006 Wharton Survey Land Use Regulation Index 0.0348 (0.0218) −0.0368** (0.0160) −0.0276 (0.0233) −0.0161 (0.0183) −0.0397** (0.0163) −0.0444* (0.0239) 
% Minority population 0.398*** (0.125) 1.142*** (0.254) 0.234 (0.151) 0.349 (0.233) 1.165*** (0.238) 0.124 (0.208) 
Population density 2000 4.37 × 10−5 (6.04 × 10−51.34 × 10−5 (3.86 × 10−50.000103 (6.66 × 10−50.000103** (4.22 × 10−55.21 × 10−5 (4.84 × 10−58.19 × 10−5 (5.24 × 10−5
% Foreign-born 2000 0.134 (0.232) 0.0367 (0.0989) 0.389** (0.155) −0.221 (0.270) −0.00962 (0.109) −0.0282 (0.155) 
General governments 1962 −0.000148 (0.000182) 0.000172** (7.50 × 10−50.000440*** (0.000146) 7.98 × 10−7 (9.42 × 10−50.000172** (6.85 × 10−50.000277* (0.000162) 
Property tax rate 2000    0.428 (2.257) 3.519* (1.971) 8.562*** (3.052) 
Local government's share of revenue    0.0116 (0.148) −0.148 (0.131) 0.0581 (0.195) 
Log per capita state taxation    0.0893 (0.0733) −0.0165 (0.0532) −0.0532 (0.0952) 
Property tax share of local revenue    −0.000616 (0.000953) −0.00109 (0.000665) −0.00174 (0.00111) 
Median home rent/median home value    −0.366 (17.51) 4.933 (14.73) −51.11** 23.28) 
Log median household income 2000    1.64 × 10−6 (2.25 × 10−6−1.21 × 10−6 (1.26 × 10−61.19 × 10−6 (1.97 × 10−6
Log housing price index    0.0714 (0.120) 0.0409 (0.112) 0.105 (0.169) 
Minority income/White income    −0.847*** (0.154) −0.0822 (0.0782) −0.545** (0.254) 
Suburban housing/central city housing    0.000469 (0.00126) 0.000424 (0.00133) 0.00372 (0.00226) 
Suburban rent/central city rent    0.0650 (0.0849) −0.135** (0.0544) 0.130 (0.0892) 
% Minorities using public transportation or carpooling    0.00158 (0.180) 0.0975 (0.188) −0.125 (0.194) 
Constant 0.705*** (0.0765) 0.476*** (0.0489) 0.684*** (0.0749) −0.0679 (1.092) 0.638 (0.785) 0.983 (1.544) 
Observations 50 50 50 50 50 50 
Joint F-test of urban form    0.21 2.97** 1.78 
Joint F-test of Teibout    0.49 5.21*** 4.61*** 
Adjusted R2 0.337 0.456 0.539 0.738 0.505 0.68 
 White Minority Dissimilarity Index 2000 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0927*** (0.0182) −0.0451*** (0.0123) −0.0680*** (0.0186) −0.0557** (0.0265) −0.0437** (0.0164) −0.0508** (0.0232) 
2006 Wharton Survey Land Use Regulation Index 0.0348 (0.0218) −0.0368** (0.0160) −0.0276 (0.0233) −0.0161 (0.0183) −0.0397** (0.0163) −0.0444* (0.0239) 
% Minority population 0.398*** (0.125) 1.142*** (0.254) 0.234 (0.151) 0.349 (0.233) 1.165*** (0.238) 0.124 (0.208) 
Population density 2000 4.37 × 10−5 (6.04 × 10−51.34 × 10−5 (3.86 × 10−50.000103 (6.66 × 10−50.000103** (4.22 × 10−55.21 × 10−5 (4.84 × 10−58.19 × 10−5 (5.24 × 10−5
% Foreign-born 2000 0.134 (0.232) 0.0367 (0.0989) 0.389** (0.155) −0.221 (0.270) −0.00962 (0.109) −0.0282 (0.155) 
General governments 1962 −0.000148 (0.000182) 0.000172** (7.50 × 10−50.000440*** (0.000146) 7.98 × 10−7 (9.42 × 10−50.000172** (6.85 × 10−50.000277* (0.000162) 
Property tax rate 2000    0.428 (2.257) 3.519* (1.971) 8.562*** (3.052) 
Local government's share of revenue    0.0116 (0.148) −0.148 (0.131) 0.0581 (0.195) 
Log per capita state taxation    0.0893 (0.0733) −0.0165 (0.0532) −0.0532 (0.0952) 
Property tax share of local revenue    −0.000616 (0.000953) −0.00109 (0.000665) −0.00174 (0.00111) 
Median home rent/median home value    −0.366 (17.51) 4.933 (14.73) −51.11** 23.28) 
Log median household income 2000    1.64 × 10−6 (2.25 × 10−6−1.21 × 10−6 (1.26 × 10−61.19 × 10−6 (1.97 × 10−6
Log housing price index    0.0714 (0.120) 0.0409 (0.112) 0.105 (0.169) 
Minority income/White income    −0.847*** (0.154) −0.0822 (0.0782) −0.545** (0.254) 
Suburban housing/central city housing    0.000469 (0.00126) 0.000424 (0.00133) 0.00372 (0.00226) 
Suburban rent/central city rent    0.0650 (0.0849) −0.135** (0.0544) 0.130 (0.0892) 
% Minorities using public transportation or carpooling    0.00158 (0.180) 0.0975 (0.188) −0.125 (0.194) 
Constant 0.705*** (0.0765) 0.476*** (0.0489) 0.684*** (0.0749) −0.0679 (1.092) 0.638 (0.785) 0.983 (1.544) 
Observations 50 50 50 50 50 50 
Joint F-test of urban form    0.21 2.97** 1.78 
Joint F-test of Teibout    0.49 5.21*** 4.61*** 
Adjusted R2 0.337 0.456 0.539 0.738 0.505 0.68 

Notes: Robust standard errors in parentheses. Urban form variables: % using public transportation/carpooling; % minorities using public transportation/carpooling; suburban–city housing ratio; suburban–city rent ratio. Tiebout variables: number of general governments, log property tax rate, local government's share of local revenue (by state); log per capita state income, local revenue from property tax.

*P < 0.1, **P < 0.05, ***P < 0.01.

In addition to permitted density, four other variables were used from the Pendall data. First, to measure infrastructure taxes, the number of fees used by each jurisdiction in the form of impact fees or APFOs was averaged by MSA. For both regulations, the goal is to displace the fixed costs of development onto developers, and so this could discourage development, increase housing costs, and thereby increase segregation. These regulations are unlikely to have the same enclave-generating effects of anti-density zoning, but they could, on the margins, make multifamily housing more expensive than single-family housing in so far as the former requires a heavier investment in infrastructure.

The second variable measures the prevalence of containment regulations. As mentioned, Nelson, Sanchez, and Dawkins (2004) find evidence that these urban growth boundaries or greenbelts reduce segregation. The theoretical logic behind this is that containment policies limit sprawl and encourage rural areas within the boundary to accept multifamily housing by increasing the value of land. These regulations also could limit isolated single-family housing in the outer suburbs.

A third variable captures limits on housing or building permits commonly referred to as growth controls. This was obtained by calculating the average number of jurisdictions in the metropolitan area that use growth controls. Finally, “inclusionary” zoning was included in the analysis. These regulations purport to subsidize affordable housing, but there is no evidence that they make housing less expensive, and they seem to be put in place for cosmetic political reasons when housing becomes unaffordable (Meltzer and Schuetz, 2008; Schuetz, Been, and Meltzer, 2008).

The results of the parsimonious specifications are reported in Columns 1–3 of Table 4. None of the regressions reveal a significant association between any of the alternative land regulations and segregation. Yet permitted density remains negative and highly significant, conditional on the other regulatory categories. The joint F-test on the regulations is significant for all three groups. The fully specified columns yield less convincing results. Controlling for everything, urban boundaries are associated with significantly less Asian and Black segregation, but no other regulation has consistently significant results across racial groups. The F-tests still reject the null hypothesis that the regulations have no effect for Asians and Blacks but not for Latinos. As other scholars of land regulation have found, multi-collinearity makes it very difficult to find significant results from a single regulation while controlling for all regulatory types.

Table 5 attempts to get around this problem by using a comprehensive regulatory index from Gyourko et al. (2008). Their Wharton Index captures all the concepts mentioned above and others, including the degree of local democracy. Entered alongside the permitted density measure from Pendall, one can interpret these regressions as testing the effect of permitted density on segregation, conditional on all other forms of land regulation. In all cases, including the fully specified models, permitted density is highly significant and the coefficients are of similar magnitude as those in Table 3.

In terms of the relative effects of containment and permitted density, Rothwell and Massey (2009) present evidence that previous density zoning predicts future adoption of containment, but the reverse is not true. They conclude that density zoning is more fundamental than containment, which is essentially a pro-density form of regulation. Theoretical logic is consistent with their empirical result. Containment would result in housing shortages if high-density developments were barred within the urban growth boundary. The regressions from Table 4 were repeated while adding a measure of containment that excludes from the index urban growth boundaries and greenbelts put in place after 1994. In those specifications, density zoning was still significant but containment was not. Overall, the results from this section provide substantial support for the contention that anti-density zoning is strongly associated with increased segregation.

Although the comprehensive Wharton Index is not significant in these specifications, Gyourko et al. (2008) have also made its components available for study. The one that most closely resembles density zoning is their density restrictions index. In Table 6, this is used in place of Pendall's maximum density zoning and the full specifications are reexamined. The density restrictions index is highly significant in all cases, and the comprehensive measure of regulation is negatively associated with segregation.6 Here a change from the maximum amount of regulation (1) to the minimum (0) would reduce segregation by eight to seventeen points on a 100-point scale. Taking into account the differences between the most segregated MSA in the sample and the least, that effect would explain 33–37% of the observed range in segregation. The size of this effect is about half the size as that in the income ratio.

Table 6.

OLS Regressions of 2000 Dissimilarity Index on Density Restrictions Index of Gyourko et al. and Wharton Index

 Latino Asian Black 
 
Density restrictions index 0.141*** (0.0496) 0.0950** (0.0374) 0.182*** (0.0504) 
2006 Wharton Land Use Regulation Index −0.0579** (0.0224) −0.0673*** (0.0209) −0.0947*** (0.0228) 
% Minority population 0.266 (0.219) 1.037*** (0.225) 0.167 (0.159) 
Population density 2000 0.000109*** (3.90 × 10−55.33 × 10−5 (5.47 × 10−58.87 × 10−5* (5.20 × 10−5
% Foreign-born 2000 −0.174 (0.284) −0.0436 (0.110) −0.0648 (0.150) 
General governments 1962 4.80 × 10−5 (0.000102) 0.000205*** (7.57 × 10−50.000328** (0.000158) 
Property tax rate 2000 4.239** (2.017) 5.826*** (2.121) 11.98*** (3.416) 
Local government's share of revenue −0.113 (0.160) −0.251* (0.127) −0.0804 (0.208) 
Log per capita state taxation 0.130 (0.0787) 0.0227 (0.0510) −0.00836 (0.0913) 
Property tax share of local revenue −0.000672 (0.00101) −0.00121* (0.000670) −0.00213* (0.00115) 
Median home rent/median home value 1.850 (19.48) 6.350 (15.13) −52.47** (23.46) 
Log median household income 2000 1.19 × 10−6 (2.47 × 10−6−2.32 × 10−6* (1.20 × 10−6−2.19 × 10−7 (2.14 × 10−6
Log housing price index 0.0982 (0.131) 0.0603 (0.112) 0.155 (0.164) 
Minority income/White income −0.783*** (0.153) −0.0510 (0.0682) −0.532** (0.236) 
Suburban housing/central city housing 6.32 × 10−5 (0.00144) 0.000145 (0.00115) 0.00320 (0.00202) 
Suburban rent/central city rent 0.0908 (0.0798) −0.112* (0.0618) 0.143* (0.0757) 
% Minorities using public transportation or carpooling 0.118 (0.149) 0.197 (0.164) −0.0576 (0.174) 
Constant −0.806 (1.208) 0.0444 (0.798) 0.203 (1.429) 
F-test on urban form 0.55 2.71* 2.05 
F-test on Tiebout 2.61** 7.94*** 7.21*** 
Adjusted R2 0.735 0.463 0.713 
 Latino Asian Black 
 
Density restrictions index 0.141*** (0.0496) 0.0950** (0.0374) 0.182*** (0.0504) 
2006 Wharton Land Use Regulation Index −0.0579** (0.0224) −0.0673*** (0.0209) −0.0947*** (0.0228) 
% Minority population 0.266 (0.219) 1.037*** (0.225) 0.167 (0.159) 
Population density 2000 0.000109*** (3.90 × 10−55.33 × 10−5 (5.47 × 10−58.87 × 10−5* (5.20 × 10−5
% Foreign-born 2000 −0.174 (0.284) −0.0436 (0.110) −0.0648 (0.150) 
General governments 1962 4.80 × 10−5 (0.000102) 0.000205*** (7.57 × 10−50.000328** (0.000158) 
Property tax rate 2000 4.239** (2.017) 5.826*** (2.121) 11.98*** (3.416) 
Local government's share of revenue −0.113 (0.160) −0.251* (0.127) −0.0804 (0.208) 
Log per capita state taxation 0.130 (0.0787) 0.0227 (0.0510) −0.00836 (0.0913) 
Property tax share of local revenue −0.000672 (0.00101) −0.00121* (0.000670) −0.00213* (0.00115) 
Median home rent/median home value 1.850 (19.48) 6.350 (15.13) −52.47** (23.46) 
Log median household income 2000 1.19 × 10−6 (2.47 × 10−6−2.32 × 10−6* (1.20 × 10−6−2.19 × 10−7 (2.14 × 10−6
Log housing price index 0.0982 (0.131) 0.0603 (0.112) 0.155 (0.164) 
Minority income/White income −0.783*** (0.153) −0.0510 (0.0682) −0.532** (0.236) 
Suburban housing/central city housing 6.32 × 10−5 (0.00144) 0.000145 (0.00115) 0.00320 (0.00202) 
Suburban rent/central city rent 0.0908 (0.0798) −0.112* (0.0618) 0.143* (0.0757) 
% Minorities using public transportation or carpooling 0.118 (0.149) 0.197 (0.164) −0.0576 (0.174) 
Constant −0.806 (1.208) 0.0444 (0.798) 0.203 (1.429) 
F-test on urban form 0.55 2.71* 2.05 
F-test on Tiebout 2.61** 7.94*** 7.21*** 
Adjusted R2 0.735 0.463 0.713 

Notes: Fifty observations of MSAs. Robust standard errors in parentheses. Urban form variables: % using public transportation/carpooling; % minorities using public transportation/carpooling; suburban–city housing ratio; suburban–city rent ratio. Tiebout variables: number of general governments, log property tax rate, local government's share of local revenue (by state); log per capita state income, local revenue from property tax.

*P < 0.1, **P < 0.05, ***P < 0.01.

Controlling for the Deeper Causes of Segregation

Readers may be concerned that the independent variables included here so far are all measured in 2000, except the number of local governments in the metropolitan area. This may be a problem if historical forces correlated with zoning shaped racial housing patterns during an epoch when anti-minority sentiment was ubiquitous, lawful, and institutionalized. In order to address this concern, this section includes the percentage of metropolitan employment in agriculture in 1970, the percentage of employment in manufacturing in 1970, the percentage of housing units built before 1960, population density in 1910, distance to Mexico, and year of statehood for the principal city of the metropolitan area. The 1960s and 1970s variables were chosen to capture the differences in economic structure and housing stock during the period just before major civil rights legislation was passed. Population density in 1910 captures the degree of early industrialization and urban development and year of statehood proxies for years of settlement by Europeans. Distance to Mexico is strongly associated with Latino settlement patterns and may be correlated with zoning.

Before interpreting these results too literally, many if not all of these variables could have a causal effect on zoning, and hence their inclusion may lead to the attenuation of zoning, even if the latter is a more proximate cause of contemporary segregation. As outlined in the theoretical model above, industrial development in central cities threatens the residents of low-density suburbs with population growth, higher taxes, and loss of political power, which will inspire anti-density zoning. Likewise, a short distance to Mexico may have made White suburban residents worry about immigrant-driven population growth in their jurisdictions.

The results, which are shown in Table 7, show that adding historical controls changes the results considerably for Latino and Asian segregation, in that zoning is no longer significantly associated with the segregation of these groups. Distance to the Mexican border and the share of units built before 1960 are the most consistently significant, but all the variables except the manufacturing share are significant in at least one specification. The results for Black segregation are more surprising: zoning remains significantly associated with Black–White segregation using both the dissimilarity index and the isolation index. Moreover, the marginal effect is only slightly lower than that in previous specifications. At least with Black–White segregation, zoning seems to be robust to a variety of potentially deeper causes of segregation.

Table 7.

OLS Regression of Permitted Density on Full Controls and Historical Characteristics of Metropolitan Area

 Dissimilarity Isolation 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0310 (0.0240) −0.00922 (0.0132) −0.0512*** (0.0183) −0.0178 (0.0256) −0.00337 (0.0175) −0.0603*** (0.0190) 
% Minority population 0.225 (0.257) 0.824*** (0.228) −0.353** (0.169) 1.212*** (0.265) 1.973*** (0.429) 0.546*** (0.175) 
Population density 2000 5.64 × 10−5 (5.04 × 10−51.55 × 10−5 (3.31 × 10−55.06 × 10−5 (4.24 × 10−55.71 × 10−5 (4.72 × 10−55.61 × 10−6 (3.44 × 10−54.64 × 10−5 (4.85 × 10−5
% Foreign-born 2000 −0.00573 (0.313) 0.0268 (0.0833) −0.0624 (0.119) 0.0591 (0.323) 0.337*** (0.118) 0.359* (0.189) 
General governments 1962 −9.39 × 10−5* (5.42 × 10−50.000124 (8.18 × 10−5−3.33 × 10−5 (0.000110) −6.66 × 10−6 (0.000169) 9.07 × 10−5 (6.69 × 10−53.68 × 10−5 (0.000134) 
Property tax rate 2000 −4.624** (2.239) 2.026 (2.049) −2.477 (3.064) −0.525 (3.154) −2.203 (2.347) −3.566 (3.148) 
Local government's share of revenue 0.0205 (0.141) −0.0856 (0.137) 0.222 (0.134) 0.0207 (0.175) 0.0669 (0.118) 0.253 (0.152) 
Log per capita state taxation 0.0716 (0.0610) 0.0177 (0.0477) −0.193*** (0.0624) 0.0839 (0.0969) 0.0243 (0.0447) −0.239*** (0.0818) 
Property tax share of local revenue 0.000288 (0.000883) −0.000502 (0.000516) −0.00139 (0.000871) 0.000148 (0.00107) −0.000310 (0.000542) −0.00140* (0.000778) 
Median home rent/median home value 42.28** (17.19) 25.36** (11.93) −9.600 (19.79) 33.16 (21.75) 19.64* (11.70) −0.667 (24.34) 
Log median household income 2000 8.68 × 10−7 (2.46 × 10−6−2.85 × 10−6** (1.08 × 10−6−1.56 × 10−6 (1.55 × 10−6−1.16 × 10−6 (2.74 × 10−6−1.36 × 10−6 (1.27 × 10−6−1.33 × 10−6 (1.76 × 10−6
Log housing price index 0.134 (0.128) 0.139 (0.106) −0.102 (0.138) 0.188 (0.168) 0.0424 (0.101) −0.258* (0.151) 
Minority income/White income −0.960*** (0.126) −0.0801 (0.0607) −1.064*** (0.179) −0.882*** (0.142) −0.217*** (0.0635) −1.162*** (0.206) 
Suburban housing/central city housing 0.00200 (0.00120) 0.00129* (0.000755) 0.00855*** (0.00153) 0.00280* (0.00156) 0.000712 (0.00113) 0.00843*** (0.00186) 
Suburban rent/central city rent 0.0302 (0.0715) −0.124** (0.0530) 0.117 (0.0731) 0.0332 (0.0843) −0.0507 (0.0478) 0.0569 (0.0710) 
% Minorities using public transportation or carpooling −0.0540 (0.148) 0.138 (0.170) −0.334* (0.196) −0.0960 (0.177) −0.0220 (0.208) −0.553*** (0.193) 
% Employment in agriculture in 1970 −0.463 (0.672) 0.136 (0.587) −2.756*** (0.916) −0.410 (0.876) 0.479 (0.496) −3.450*** (1.166) 
% Employment in manufacturing in 1970 −0.0495 (0.116) 0.0725 (0.0961) −0.0297 (0.125) −0.208 (0.139) −0.00505 (0.0976) 0.0177 (0.124) 
% 2000 units built before 1960 0.285** (0.126) 0.173 (0.125) 0.442*** (0.0879) 0.105 (0.146) 0.267** (0.116) 0.440*** (0.113) 
Population density in 1910 0.000127* (7.06 × 10−52.85 × 10−5 (7.04 × 10−50.000125 (7.59 × 10−50.000222** (0.000101) 9.54 × 10−5 (7.21 × 10−50.000133 (8.48 × 10−5
Year of statehood −0.000346 (0.000374) −0.00111*** (0.000278) −4.21 × 10−5 (0.000402) −0.000341 (0.000357) −0.000647** (0.000270) −0.000183 (0.000475) 
Thousands of miles to nearest Mexican border −0.0361 (0.0300) −0.0641*** (0.0194) −0.0572*** (0.0164) −0.0381 (0.0299) −0.0663*** (0.0225) −0.0578*** (0.0191) 
Constant 0.181 (0.936) 1.668*** (0.485) 3.353*** (1.089) −0.446 (1.577) 0.904 (0.577) 4.589*** (1.392) 
Observations 50 50 50 50 50 50 
Joint F-test on urban form 1.27 4.29*** 12.54*** 1.26 0.51 8.24*** 
Joint F-test on Tiebout variables 1.53 0.86 4.34*** 0.22 0.57 5.45*** 
Adjusted R2 0.804 0.682 0.834 0.936 0.899 0.914 
 Dissimilarity Isolation 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0310 (0.0240) −0.00922 (0.0132) −0.0512*** (0.0183) −0.0178 (0.0256) −0.00337 (0.0175) −0.0603*** (0.0190) 
% Minority population 0.225 (0.257) 0.824*** (0.228) −0.353** (0.169) 1.212*** (0.265) 1.973*** (0.429) 0.546*** (0.175) 
Population density 2000 5.64 × 10−5 (5.04 × 10−51.55 × 10−5 (3.31 × 10−55.06 × 10−5 (4.24 × 10−55.71 × 10−5 (4.72 × 10−55.61 × 10−6 (3.44 × 10−54.64 × 10−5 (4.85 × 10−5
% Foreign-born 2000 −0.00573 (0.313) 0.0268 (0.0833) −0.0624 (0.119) 0.0591 (0.323) 0.337*** (0.118) 0.359* (0.189) 
General governments 1962 −9.39 × 10−5* (5.42 × 10−50.000124 (8.18 × 10−5−3.33 × 10−5 (0.000110) −6.66 × 10−6 (0.000169) 9.07 × 10−5 (6.69 × 10−53.68 × 10−5 (0.000134) 
Property tax rate 2000 −4.624** (2.239) 2.026 (2.049) −2.477 (3.064) −0.525 (3.154) −2.203 (2.347) −3.566 (3.148) 
Local government's share of revenue 0.0205 (0.141) −0.0856 (0.137) 0.222 (0.134) 0.0207 (0.175) 0.0669 (0.118) 0.253 (0.152) 
Log per capita state taxation 0.0716 (0.0610) 0.0177 (0.0477) −0.193*** (0.0624) 0.0839 (0.0969) 0.0243 (0.0447) −0.239*** (0.0818) 
Property tax share of local revenue 0.000288 (0.000883) −0.000502 (0.000516) −0.00139 (0.000871) 0.000148 (0.00107) −0.000310 (0.000542) −0.00140* (0.000778) 
Median home rent/median home value 42.28** (17.19) 25.36** (11.93) −9.600 (19.79) 33.16 (21.75) 19.64* (11.70) −0.667 (24.34) 
Log median household income 2000 8.68 × 10−7 (2.46 × 10−6−2.85 × 10−6** (1.08 × 10−6−1.56 × 10−6 (1.55 × 10−6−1.16 × 10−6 (2.74 × 10−6−1.36 × 10−6 (1.27 × 10−6−1.33 × 10−6 (1.76 × 10−6
Log housing price index 0.134 (0.128) 0.139 (0.106) −0.102 (0.138) 0.188 (0.168) 0.0424 (0.101) −0.258* (0.151) 
Minority income/White income −0.960*** (0.126) −0.0801 (0.0607) −1.064*** (0.179) −0.882*** (0.142) −0.217*** (0.0635) −1.162*** (0.206) 
Suburban housing/central city housing 0.00200 (0.00120) 0.00129* (0.000755) 0.00855*** (0.00153) 0.00280* (0.00156) 0.000712 (0.00113) 0.00843*** (0.00186) 
Suburban rent/central city rent 0.0302 (0.0715) −0.124** (0.0530) 0.117 (0.0731) 0.0332 (0.0843) −0.0507 (0.0478) 0.0569 (0.0710) 
% Minorities using public transportation or carpooling −0.0540 (0.148) 0.138 (0.170) −0.334* (0.196) −0.0960 (0.177) −0.0220 (0.208) −0.553*** (0.193) 
% Employment in agriculture in 1970 −0.463 (0.672) 0.136 (0.587) −2.756*** (0.916) −0.410 (0.876) 0.479 (0.496) −3.450*** (1.166) 
% Employment in manufacturing in 1970 −0.0495 (0.116) 0.0725 (0.0961) −0.0297 (0.125) −0.208 (0.139) −0.00505 (0.0976) 0.0177 (0.124) 
% 2000 units built before 1960 0.285** (0.126) 0.173 (0.125) 0.442*** (0.0879) 0.105 (0.146) 0.267** (0.116) 0.440*** (0.113) 
Population density in 1910 0.000127* (7.06 × 10−52.85 × 10−5 (7.04 × 10−50.000125 (7.59 × 10−50.000222** (0.000101) 9.54 × 10−5 (7.21 × 10−50.000133 (8.48 × 10−5
Year of statehood −0.000346 (0.000374) −0.00111*** (0.000278) −4.21 × 10−5 (0.000402) −0.000341 (0.000357) −0.000647** (0.000270) −0.000183 (0.000475) 
Thousands of miles to nearest Mexican border −0.0361 (0.0300) −0.0641*** (0.0194) −0.0572*** (0.0164) −0.0381 (0.0299) −0.0663*** (0.0225) −0.0578*** (0.0191) 
Constant 0.181 (0.936) 1.668*** (0.485) 3.353*** (1.089) −0.446 (1.577) 0.904 (0.577) 4.589*** (1.392) 
Observations 50 50 50 50 50 50 
Joint F-test on urban form 1.27 4.29*** 12.54*** 1.26 0.51 8.24*** 
Joint F-test on Tiebout variables 1.53 0.86 4.34*** 0.22 0.57 5.45*** 
Adjusted R2 0.804 0.682 0.834 0.936 0.899 0.914 

Notes: Robust standard errors in parentheses, clustered on state. Urban form variables: % using public transportation/carpooling; % minorities using public transportation/carpooling; suburban–city housing ratio; suburban–city rent ratio. Tiebout variables: number of general governments, log property tax rate, local government's share of local revenue (by state); log per capita state income, local revenue from property tax.

*P < 0.1, **P < 0.05, ***P < 0.01.

Regression Analysis for 1990–2000 Changes in Segregation

This section examines whether or not zoning can be linked to the changing patterns of segregation rather than just its level. This strategy allows one to answer the question: which metropolitan characteristics in 1990 predicted desegregation from 1990 to 2000? Since the Pendall measure of zoning was taken in 2003, and the goal is to measure its effect from 1990 to 2000, this could introduce a biased estimate if changes in density regulation varied systematically with previous zoning or segregation. Fortunately, as mentioned above, Pendall's survey allows one to control for which MSAs increased or decreased their density zoning score since 1994 by ≥10%. Using these data, two variables were created for the percentage of jurisdictions in each MSA that increased or decreased their density zoning and overall regulations.

Table 8 runs the fully specified regressions with and without controlling for changes to the zoning regime. In so far as durable patterns of urban form and Tiebout sorting incentives affect segregation, they are absorbed into the measure of 1990 segregation and do not need to be included. In addition to demographic variables and the income ratios, the other controls added here are intended to capture predicted growth. Glaeser and Shapiro (2003) found that average summer and winter temperatures strongly predicted population growth during the 1990s as did the share of adults with a college degree. The log of median rent is also included to capture demand. Also added to the model were the changes in the minority groups’ shares and changes in the foreign-born shares. This is important because the goal is to assess the effects of density zoning when all things are equal, and MSAs that are becoming more Latino, Asian, or nonnative are more likely to see an increase in segregation.

Table 8.

OLS Regression of 1990–2000 Change in Segregation Conditional on Permitted Density in MSA

 Log of 2000 Minority Dissimilarity Index 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0760** (0.0352) −0.0341 (0.0254) −0.0268 (0.0176) −0.101*** (0.0311) −0.0442 (0.0268) −0.0374* (0.0199) 
%Foreign00−%foreign90 3.139 (1.995) −0.344 (0.809) −0.398 (0.564) 3.706 (2.316) 0.612 (0.944) 0.271 (0.511) 
%Minority00−%minority90 0.400 (1.513) 3.908 (3.155) −0.302 (0.685) 0.723 (1.674) 2.405 (2.456) 0.0429 (0.711) 
Log dissimilarity index 1990 0.567*** (0.119) 0.887*** (0.0808) 1.065*** (0.0837) 0.561*** (0.120) 0.874*** (0.0718) 1.077*** (0.0759) 
% Minority 1990 −0.0596 (0.471) −0.882 (0.673) 0.397*** (0.112) −0.0535 (0.461) −0.521 (0.530) 0.382*** (0.100) 
Population density 1990 −6.39 × 10−5 (7.53 × 10−5−8.05 × 10−5** (3.45 × 10−52.62 × 10−5 (3.63 × 10−5−3.28 × 10−5 (8.77 × 10−5−6.40 × 10−5* (3.25 × 10−53.39 × 10−5 (3.53 × 10−5
% Foreign-born 1990 −0.0525 (0.505) 0.582*** (0.205) 0.0720 (0.159) −0.189 (0.537) 0.436** (0.182) 0.0358 (0.152) 
Average July temperature −0.00391 (0.00348) −0.000381 (0.00261) 0.000913 (0.00240) −0.00467 (0.00384) −0.000803 (0.00238) 0.000339 (0.00239) 
Average January temperature −0.000104 (0.00193) −0.000151 (0.00140) 0.000876 (0.00125) 0.000365 (0.00191) −0.000104 (0.00129) 0.00110 (0.00124) 
% With BA 1990 0.00928 (0.551) −0.343 (0.343) 0.543** (0.221) −0.0648 (0.551) −0.260 (0.389) 0.598** (0.255) 
Average minority income/average White income −0.0965 (0.390) −0.0655 (0.0810) 0.0720 (0.234) −0.175 (0.408) −0.0657 (0.0723) 0.0936 (0.217) 
Log median rent 1990 −0.00647 (0.161) −0.0506 (0.0730) −0.00531 (0.0582) 0.0123 (0.166) −0.0154 (0.0718) 0.000644 (0.0612) 
% All residents using public transportation/carpooling −0.190 (0.514) 0.0710 (0.259) −0.285* (0.147) −0.295 (0.504) −0.0423 (0.257) −0.387** (0.192) 
% Juris that increased allowable density since 1994    −0.00808 (0.305) −0.224 (0.135) −0.112 (0.136) 
% Juris that decreased allowable density since 1994    −0.535* (0.292) −0.247 (0.183) −0.242* (0.135) 
Constant 0.366 (1.139) 0.474 (0.623) −0.189 (0.540) 0.450 (1.117) 0.320 (0.580) −0.155 (0.549) 
Observations 50 50 50 50 50 50 
Adjusted R2 0.744 0.871 0.922 0.752 0.879 0.926 
 Log of 2000 Minority Dissimilarity Index 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.0760** (0.0352) −0.0341 (0.0254) −0.0268 (0.0176) −0.101*** (0.0311) −0.0442 (0.0268) −0.0374* (0.0199) 
%Foreign00−%foreign90 3.139 (1.995) −0.344 (0.809) −0.398 (0.564) 3.706 (2.316) 0.612 (0.944) 0.271 (0.511) 
%Minority00−%minority90 0.400 (1.513) 3.908 (3.155) −0.302 (0.685) 0.723 (1.674) 2.405 (2.456) 0.0429 (0.711) 
Log dissimilarity index 1990 0.567*** (0.119) 0.887*** (0.0808) 1.065*** (0.0837) 0.561*** (0.120) 0.874*** (0.0718) 1.077*** (0.0759) 
% Minority 1990 −0.0596 (0.471) −0.882 (0.673) 0.397*** (0.112) −0.0535 (0.461) −0.521 (0.530) 0.382*** (0.100) 
Population density 1990 −6.39 × 10−5 (7.53 × 10−5−8.05 × 10−5** (3.45 × 10−52.62 × 10−5 (3.63 × 10−5−3.28 × 10−5 (8.77 × 10−5−6.40 × 10−5* (3.25 × 10−53.39 × 10−5 (3.53 × 10−5
% Foreign-born 1990 −0.0525 (0.505) 0.582*** (0.205) 0.0720 (0.159) −0.189 (0.537) 0.436** (0.182) 0.0358 (0.152) 
Average July temperature −0.00391 (0.00348) −0.000381 (0.00261) 0.000913 (0.00240) −0.00467 (0.00384) −0.000803 (0.00238) 0.000339 (0.00239) 
Average January temperature −0.000104 (0.00193) −0.000151 (0.00140) 0.000876 (0.00125) 0.000365 (0.00191) −0.000104 (0.00129) 0.00110 (0.00124) 
% With BA 1990 0.00928 (0.551) −0.343 (0.343) 0.543** (0.221) −0.0648 (0.551) −0.260 (0.389) 0.598** (0.255) 
Average minority income/average White income −0.0965 (0.390) −0.0655 (0.0810) 0.0720 (0.234) −0.175 (0.408) −0.0657 (0.0723) 0.0936 (0.217) 
Log median rent 1990 −0.00647 (0.161) −0.0506 (0.0730) −0.00531 (0.0582) 0.0123 (0.166) −0.0154 (0.0718) 0.000644 (0.0612) 
% All residents using public transportation/carpooling −0.190 (0.514) 0.0710 (0.259) −0.285* (0.147) −0.295 (0.504) −0.0423 (0.257) −0.387** (0.192) 
% Juris that increased allowable density since 1994    −0.00808 (0.305) −0.224 (0.135) −0.112 (0.136) 
% Juris that decreased allowable density since 1994    −0.535* (0.292) −0.247 (0.183) −0.242* (0.135) 
Constant 0.366 (1.139) 0.474 (0.623) −0.189 (0.540) 0.450 (1.117) 0.320 (0.580) −0.155 (0.549) 
Observations 50 50 50 50 50 50 
Adjusted R2 0.744 0.871 0.922 0.752 0.879 0.926 

Notes: Robust standard errors in parentheses, clustered on state.

*P < 0.1, **P < 0.05, ***P < 0.01.

None of the control variables are significant across racial groups, except for 1990 segregation. Even changes in relative minority incomes have no significant effect on changes in zoning in this specification. Overall, the results for zoning are not as strong as in previous regressions. Density zoning does seem to predict significant changes in the Latino dissimilarity index, but not the Asian index, where the sign is in the expected direction but low in relation to the standard errors. The effect on Black segregation is insignificant without the change variables and marginally significant with them. The change variables themselves slightly improve the goodness of fit.

There are a number of reasons why the results in Table 8 may suffer from an endogeneity bias, with respect to zoning and segregation. For example, Whites living in a more segregated metropolitan area may be compelled to choose exclusionary zoning regimes or tighten long-standing regulations on housing density. This could be called the fortification hypothesis, in that Whites fortify boundaries that already exist when living in segregated areas to preserve that arrangement. This would bias the effect of density zoning upward in (absolute value) magnitude and make it more likely to accept the hypothesis that zoning affects segregation. On the other hand, metropolitan areas in which minorities are becoming more integrated and rising population growth is becoming more of a threat to local tax bills arguably have the stronger incentive to tighten anti-density zoning laws. If that is true, then the coefficients from Table 8 were biased toward zero, understating the effect of zoning on segregation, and instrumented coefficients should be larger in absolute value. Call this the reactionary hypothesis.

To address these dynamics, year of statehood is used as the instrument. It is a valid instrument if it strongly predicts density zoning, which has been shown, and it is uncorrelated with the error term (or omitted variables) from the regression shown in Table 8. It would be correlated with that residual if year of statehood causes changes in segregation from 1990 to 2000 that do not work through zoning. To explore this further, the residuals were calculated for each metropolitan area using the regressions in Table 8. These residuals were regressed on density zoning and year of statehood, clustering the errors on states. The resulting t-statistics for statehood were 1.37, 0.74, and 0.89 for Latino, Asian, and Black segregation regressions, respectively. One cannot reject the null hypothesis that year of statehood is a valid instrument for zoning.

Theoretically, the link between year of statehood and zoning is that the oldest states were the first to have established settlements and the legal institutions to foster large cities. When combined with immigration and Black migration, this early urbanization compelled native Whites to seek political power and refuge from property taxes in low-density suburban jurisdictions. Table A1 shows the strong relationship between low density and low property taxes in the year 2000. The attraction of immigrants and Blacks to metropolitan areas in more developed states may have led to higher segregation, but since minority population shares and, indeed, previous segregation levels are controlled for directly, year of statehood should only affect changes in segregation through zoning. In any case, the high F-statistics in the first stage confirm that year of statehood is at least a relevant instrument in that it strongly predicts zoning, conditional on the other variables in the model.7

Assuming the instrument is valid, the results in Table 9 are consistent with the reactionary hypothesis. In each of the models, the coefficients are much larger (in absolute value), and in all cases they are significant, with standard errors clustered on states to account for unobserved correlations between metros in the same state. For Latino and Black segregation, the effects are more than double in the model that omits reported changes in zoning and >50% larger in the model that adjusts for reported changes. Moreover, jurisdictions that reportedly increased allowable density were associated with a significant decrease in segregation from 1990 to 2000 across racial groups.

Table 9.

2SLS Regression of 1990–2000 Change in Segregation Conditional on Permitted Density in MSA

 Log of 2000 Minority Dissimilarity Index 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.161** (0.0648) −0.214** (0.0910) −0.0711** (0.0322) −0.170*** (0.0536) −0.140*** (0.0515) −0.0587** (0.0233) 
%Foreign00−%foreign90 4.266** (1.984) 0.281 (0.730) −0.0733 (0.617) 4.524** (2.052) 1.485** (0.684) 0.571 (0.538) 
%Minority00−%minority90 −0.425 (1.274) 6.251** (3.050) −0.158 (0.719) 0.192 (1.130) 2.870 (1.944) 0.159 (0.598) 
Log of 1990 minority dissimilarity index 0.524*** (0.102) 0.546*** (0.193) 1.016*** (0.0748) 0.524*** (0.110) 0.682*** (0.116) 1.049*** (0.0601) 
% Minority population 0.567 (0.496) 0.680 (0.883) 0.355*** (0.0997) 0.443 (0.460) 0.442 (0.653) 0.371*** (0.0969) 
Average minority income/average White income −0.000413* (0.000219) 0.000373** (0.000160) 5.72 × 10−5 (9.68 × 10−5−0.000314 (0.000205) 0.000342*** (0.000127) 7.66 × 10−5 (7.55 × 10−5
No. of general governments, 1962 −7.70 × 10−5 (6.82 × 10−5−9.40 × 10−5** (4.21 × 10−51.72 × 10−5 (3.57 × 10−5−3.78 × 10−5 (7.83 × 10−5−6.33 × 10−5** (2.51 × 10−53.19 × 10−5 (3.08 × 10−5
Population density 1990 −0.527 (0.598) 0.521*** (0.170) 0.156 (0.172) −0.581 (0.533) 0.324** (0.148) 0.0690 (0.158) 
% Foreign-born 1990 −0.00616 (0.00417) −0.00661** (0.00308) −0.000206 (0.00209) −0.00646 (0.00407) −0.00461* (0.00242) −0.000394 (0.00206) 
Average July temperature −0.000165 (0.00215) 0.00581** (0.00262) 0.00220* (0.00129) 0.000386 (0.00210) 0.00344** (0.00171) 0.00183 (0.00127) 
Average January temperature −0.314 (0.496) −0.527 (0.459) 0.527** (0.242) −0.336 (0.513) −0.323 (0.420) 0.591** (0.244) 
% With BA 1990 −0.0267 (0.309) −0.304** (0.119) 0.146 (0.196) −0.136 (0.305) −0.198*** (0.0742) 0.130 (0.162) 
Average minority income/average White income 0.132 (0.170) −0.324** (0.143) −0.0273 (0.0466) 0.122 (0.147) −0.141* (0.0729) −0.00672 (0.0485) 
Log median rent 1990 −0.0174 (0.315) 0.0657 (0.405) −0.272** (0.135) −0.152 (0.338) −0.112 (0.263) −0.417*** (0.157) 
% All residents using public transportation/carpooling    0.0261 (0.218) −0.305** (0.127) −0.124 (0.101) 
% Juris. that increased allowable density since 1994    −0.598** (0.277) −0.399** (0.155) −0.294*** (0.0994) 
% Juris. that decreased allowable density since 1994 −0.0969 (−1.114) 2.909** (−1.146) 0.0683 (−0.41) 0.0949 (−0.944) 1.506** (−0.646) −0.0441 (−0.434) 
Observations 50 50 50 50 50 50 
F-statistics on instrument (year of statehood) 42.17*** 5.07** 15.22*** 80.69*** 12.07*** 20.88*** 
Adjusted R2 0.73 0.708 0.907 0.739 0.848 0.922 
 Log of 2000 Minority Dissimilarity Index 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.161** (0.0648) −0.214** (0.0910) −0.0711** (0.0322) −0.170*** (0.0536) −0.140*** (0.0515) −0.0587** (0.0233) 
%Foreign00−%foreign90 4.266** (1.984) 0.281 (0.730) −0.0733 (0.617) 4.524** (2.052) 1.485** (0.684) 0.571 (0.538) 
%Minority00−%minority90 −0.425 (1.274) 6.251** (3.050) −0.158 (0.719) 0.192 (1.130) 2.870 (1.944) 0.159 (0.598) 
Log of 1990 minority dissimilarity index 0.524*** (0.102) 0.546*** (0.193) 1.016*** (0.0748) 0.524*** (0.110) 0.682*** (0.116) 1.049*** (0.0601) 
% Minority population 0.567 (0.496) 0.680 (0.883) 0.355*** (0.0997) 0.443 (0.460) 0.442 (0.653) 0.371*** (0.0969) 
Average minority income/average White income −0.000413* (0.000219) 0.000373** (0.000160) 5.72 × 10−5 (9.68 × 10−5−0.000314 (0.000205) 0.000342*** (0.000127) 7.66 × 10−5 (7.55 × 10−5
No. of general governments, 1962 −7.70 × 10−5 (6.82 × 10−5−9.40 × 10−5** (4.21 × 10−51.72 × 10−5 (3.57 × 10−5−3.78 × 10−5 (7.83 × 10−5−6.33 × 10−5** (2.51 × 10−53.19 × 10−5 (3.08 × 10−5
Population density 1990 −0.527 (0.598) 0.521*** (0.170) 0.156 (0.172) −0.581 (0.533) 0.324** (0.148) 0.0690 (0.158) 
% Foreign-born 1990 −0.00616 (0.00417) −0.00661** (0.00308) −0.000206 (0.00209) −0.00646 (0.00407) −0.00461* (0.00242) −0.000394 (0.00206) 
Average July temperature −0.000165 (0.00215) 0.00581** (0.00262) 0.00220* (0.00129) 0.000386 (0.00210) 0.00344** (0.00171) 0.00183 (0.00127) 
Average January temperature −0.314 (0.496) −0.527 (0.459) 0.527** (0.242) −0.336 (0.513) −0.323 (0.420) 0.591** (0.244) 
% With BA 1990 −0.0267 (0.309) −0.304** (0.119) 0.146 (0.196) −0.136 (0.305) −0.198*** (0.0742) 0.130 (0.162) 
Average minority income/average White income 0.132 (0.170) −0.324** (0.143) −0.0273 (0.0466) 0.122 (0.147) −0.141* (0.0729) −0.00672 (0.0485) 
Log median rent 1990 −0.0174 (0.315) 0.0657 (0.405) −0.272** (0.135) −0.152 (0.338) −0.112 (0.263) −0.417*** (0.157) 
% All residents using public transportation/carpooling    0.0261 (0.218) −0.305** (0.127) −0.124 (0.101) 
% Juris. that increased allowable density since 1994    −0.598** (0.277) −0.399** (0.155) −0.294*** (0.0994) 
% Juris. that decreased allowable density since 1994 −0.0969 (−1.114) 2.909** (−1.146) 0.0683 (−0.41) 0.0949 (−0.944) 1.506** (−0.646) −0.0441 (−0.434) 
Observations 50 50 50 50 50 50 
F-statistics on instrument (year of statehood) 42.17*** 5.07** 15.22*** 80.69*** 12.07*** 20.88*** 
Adjusted R2 0.73 0.708 0.907 0.739 0.848 0.922 

Notes: Robust standard errors in parentheses, clustered at state level. Average permitted density is instrumented with year of statehood for the state of the primary city.

*P < 0.1, **P < 0.05, ***P < 0.01.

Table 10.

2SLS Regression of Permitted Density on 1990–2000 Change in Segregation and Historical Characteristics

 Log of 2000 Minority Dissimilarity Index 
 Latino Asian Black 
 
Average permitted density in MSA −0.290*** (0.0814) −0.136* (0.0812) −0.122** (0.0583) 
%Foreign00−%foreign90 4.950 (3.114) 0.964 (1.303) 0.973 (0.846) 
Log of 1990 minority dissimilarity index 0.297*** (0.114) 0.636*** (0.123) 0.956*** (0.0766) 
% Juris in MSA that increased allowable density since 1994 −0.0187 (0.222) −0.316** (0.132) −0.158* (0.0827) 
% Juris in MSA that decreased allowable density since 1994 −1.095*** (0.324) −0.448* (0.265) −0.567*** (0.158) 
Thousands of miles to nearest Mexican border −0.231*** (0.0624) −0630 (0.0456) −0936*** (0.0262) 
% 2000 housing units built before 1960 −0.141 (0.306) −0.0864 (0.171) −0.0242 (0.132) 
Population density in 1910 −0.000418* (0.000240) −0.000109 (0.000147) −2.33 × 10−5 (0.000172) 
% Jobs in agriculture in 1970 1.530 (1.616) −1.666** (0.802) −1.058 (0.819) 
% Jobs in manufacturing in 1970 0.545 (0.374) 0.165 (0.257) 0.0674 (0.189) 
Constant −2.288 (1.565) 1.187** (0.515) 0.180 (0.542) 
Observations 50 50 50 
F-statistics on instrument (year of statehood) 12.97*** 3.22* 3.36* 
Anderson canonical correlation statistic 13.40*** 2.72* 5.68** 
Adjusted R2 0.720 0.871 0.916 
 Log of 2000 Minority Dissimilarity Index 
 Latino Asian Black 
 
Average permitted density in MSA −0.290*** (0.0814) −0.136* (0.0812) −0.122** (0.0583) 
%Foreign00−%foreign90 4.950 (3.114) 0.964 (1.303) 0.973 (0.846) 
Log of 1990 minority dissimilarity index 0.297*** (0.114) 0.636*** (0.123) 0.956*** (0.0766) 
% Juris in MSA that increased allowable density since 1994 −0.0187 (0.222) −0.316** (0.132) −0.158* (0.0827) 
% Juris in MSA that decreased allowable density since 1994 −1.095*** (0.324) −0.448* (0.265) −0.567*** (0.158) 
Thousands of miles to nearest Mexican border −0.231*** (0.0624) −0630 (0.0456) −0936*** (0.0262) 
% 2000 housing units built before 1960 −0.141 (0.306) −0.0864 (0.171) −0.0242 (0.132) 
Population density in 1910 −0.000418* (0.000240) −0.000109 (0.000147) −2.33 × 10−5 (0.000172) 
% Jobs in agriculture in 1970 1.530 (1.616) −1.666** (0.802) −1.058 (0.819) 
% Jobs in manufacturing in 1970 0.545 (0.374) 0.165 (0.257) 0.0674 (0.189) 
Constant −2.288 (1.565) 1.187** (0.515) 0.180 (0.542) 
Observations 50 50 50 
F-statistics on instrument (year of statehood) 12.97*** 3.22* 3.36* 
Anderson canonical correlation statistic 13.40*** 2.72* 5.68** 
Adjusted R2 0.720 0.871 0.916 

Notes: Robust standard errors in parentheses, clustered by state. Instrumental variable is year of statehood. All variables from previous table are included in the regression; coefficients are available upon request. The Anderson statistic tests the null hypothesis that the instrument does not identify the endogenous variable.

*P < 0.1, **P < 0.05, ***P < 0.01.

For Asian segregation, the marginal effect of permitted density is six times larger in the first model and three times larger in the second. These figures should be interpreted with caution and may, in fact, be driven by the significant negative association between year of statehood and Asian segregation observed in Table 7. In other words, there may be something unique about young states and decreases in Asian segregation from 1990 to 2000 that is unrelated to Asian population shares and zoning.

To further guard against a potentially invalid instrument, the other historic variables used in Table 7 are brought back into the 2-stage least squares (2SLS) regression reported in Table 9. Now, this specification includes a host of deeper development and settlement-related causes of segregation. If year of statehood is still significantly associated with zoning, it is (potentially) because years spent under state institutions have a unique effect on the emergence of rural (low-density) settlements that cuts through aggregate metropolitan population density, agricultural or manufacturing production, the share of the housing stock built in the Jim Crow era, and Latino settlement patterns influenced by proximity to Mexico. Remarkably, the coefficient on instrumented permitted density retains its significance for all three minority groups. It must be noted, however, that the effect of zoning on segregation is now much weaker for Asians and only marginally significant, implying that young states have other historical characteristics that are associated with Asian desegregation, such as agricultural production, which may have prevented the development of significant rural towns in the suburbs of western cities. Overall, however, the results are consistent with the interpretation that liberal zoning laws significantly contributed to the desegregation of Latinos, Blacks, and even Asians, albeit to a lesser extent.

Predicting Changes in Zoning in a Metropolitan Fixed-Effects Model

In this section, the assumption that unobserved metropolitan characteristics (that do not change from 1990 to 2003) are exogenous to zoning can be relaxed. These characteristics could include state and local housing policies that are invariant over this period, historic legacies or cultures associated with racial sorting, urban form, government fragmentation, and regional or geographical characteristics. For the reduced sample of twenty-five metropolitan areas, the 1994 and 2003 measures of zoning were standardized to have a mean of 0 and a standard deviation of 1. The 1994 measure was then subtracted from the 2003 measure and used as the main independent variable.8 The average MSA became more exclusionary in absolute terms, and similarly, taking the difference in standardized indices shows that the average MSA became 0.01 points more exclusionary with a standard deviation of 0.65. Again, this signifies the stability of zoning regimes. New Haven and San Diego liberalized the most—increasing average density by 0.33 and 0.21, respectively, from 1994 to 2003. Using the standardized index, New Haven went from being −80% of a standard deviation below the mean to being just 24% below. San Diego, on the other hand, increased from 65% above the mean to 178%. Miami and Tampa also saw considerable liberalization. The MSAs that tightened regulations the most were Houston, Cincinnati, Kansas City, Detroit, and Cleveland. Each decreased permitted density by >0.6 points. There was no correlation between 1994 density scores and changes from 1994 to 2003 (0.001). The variables used as controls include changes in the following: the share that is foreign born, the share of the racial group in question, relative income, and population.

The results are presented in Table 11. A one-unit increase in the permitted density score is associated with a decrease in Black and Latino segregation of two percentage points and a decrease of Asian segregation equal to almost one percentage point. The results are only marginally significant for Asians and marginally insignificant for Latinos, but this is not surprising given the small sample size and lower levels of observed segregation for these groups. No other variable was consistently predictive of changes in segregation across groups, but income differences were significant in explaining changes in Latino segregation, and increases in immigration shares predicted increases in Asian segregation. A one standard deviation change in density zoning explained 25%, 25%, and 35% of a standard deviation change in the dissimilarity scores of Latinos, Asians, and Blacks, respectively.

Table 11.

Metropolitan Fixed Effects: OLS Regression of Desegregation on Changes from 1990 to 2000

 Dissimilarity Index 2000−Dissimilarity Index 1990 
 Latino Asian Black 
 
Permitted density 2003−permitted density 1994 −0.0253* (0.0134) −0.00856* (0.00438) −0.0177*** (0.00566) 
%Foreign-born2000−%foreign-born1990 1.496* (0.774) 0.322** (0.143) 0.428 (0.251) 
%Minority2000−%minority1990 −0.716 (0.566) −0.634* (0.312) 0.398 (0.289) 
Minority to White income 2000−minority to White income 1990 −0.434** (0.200) −0.0632 (0.0625) −0.212 (0.175) 
Population 2000−population 1990 −1.16 × 10−8 (1.71 × 10−87.72 × 10−9* (3.88 × 10−9−6.00 × 10−9 (9.92 × 10−9
Constant −0.0193 (0.0217) 0.00120 (0.00900) −0.0487*** (0.00955) 
Observations 25 25 25 
Adjusted R2 0.288 0.151 0.086 
 Dissimilarity Index 2000−Dissimilarity Index 1990 
 Latino Asian Black 
 
Permitted density 2003−permitted density 1994 −0.0253* (0.0134) −0.00856* (0.00438) −0.0177*** (0.00566) 
%Foreign-born2000−%foreign-born1990 1.496* (0.774) 0.322** (0.143) 0.428 (0.251) 
%Minority2000−%minority1990 −0.716 (0.566) −0.634* (0.312) 0.398 (0.289) 
Minority to White income 2000−minority to White income 1990 −0.434** (0.200) −0.0632 (0.0625) −0.212 (0.175) 
Population 2000−population 1990 −1.16 × 10−8 (1.71 × 10−87.72 × 10−9* (3.88 × 10−9−6.00 × 10−9 (9.92 × 10−9
Constant −0.0193 (0.0217) 0.00120 (0.00900) −0.0487*** (0.00955) 
Observations 25 25 25 
Adjusted R2 0.288 0.151 0.086 

Notes: Robust standard errors in parentheses. Permitted density scores are standardized indices. Regressions subtract out metropolitan area effects. See text.

*P < 0.1, **P < 0.05, ***P < 0.01.

Other Robustness Checks

Racial bias.

Given that anti-minority attitudes may be correlated with exclusionary regulation, some readers may wonder if a pervasive pattern of housing market discrimination or, alternatively, decentralized White racism is the fundamental cause behind the observed link between segregation and density zoning. One strategy to test these hypotheses would be to aggregate individual data into MSA indices of anti-minority sentiment or discrimination. Unfortunately, I know of no surveys that are representative at the MSA or city level for anti-minority sentiment, but the U.S. Department of Housing and Urban Development has sponsored auditors to systematically test housing markets for racial discrimination in general and how it relates to racial steering in particular (Turner et al., 2002; Turner and Ross, 2003). The results of this work are discussed in the Appendix below (see Table A2 in particular). To summarize, the trends in patterns of real estate market discrimination, while disturbingly pervasive, are at odds with the trends in segregation, and MSA markets with high levels of real estate market discrimination do not exhibit higher segregation.

Despite the absence of MSA indicators of racism, the 2000 General Social Survey (GSS) does ask respondents living in metropolitan areas relevant questions about housing preferences with respect to Asians, Latinos, and Blacks and also asks them to estimate the number of minorities living in their community (Davis and Smith, 2009). Table A3 summarizes the data using a question that asks respondents whether or not they opposed living in a hypothetical neighborhood that was half comprised of a particular minority group. White opposition to living around minorities was highest for Blacks (at 24%) and lowest for Asians (at 9%), which is consistent with overall segregation patterns. Seven percent of both Blacks and Latinos also said they opposed living in a half Black neighborhood, but no group expressed opposition to living in White neighborhoods (only 4% of Blacks said this). Clearly, if racial bias is a fundamental cause of segregation, this evidence suggests that it is a bias of Whites against minorities.

The author obtained permission to acquire the MSA codes for these data from the GSS and matched it to the thirty-seven metropolitan areas for which there were data on both zoning and racial bias. The model adjusts for the interaction between permitted density and the minority share of the metropolitan population, permitted density, the minority share of the metropolitan population, the total population, the land area, and per capita income. The model also adjusts for an extensive list of individual-level controls to adjust for socioeconomic status, family circumstances, gender, age, and whether or not the respondents live in the suburbs or city. Along with the racial bias variable, the interaction term between density zoning and the percentage of each minority group is the key independent variable since there is a mechanical relationship between the population share and neighborhood exposure and zoning tends to be more liberal in the west, where there are fewer Blacks but more Latinos and Asians. In order to improve the rather low goodness of fit, the model also adjusts for state fixed effects, so that the comparisons are between metropolitan areas in the same state.

The results are presented in Table 12. In general, these variables explain little of the variation in individual exposure of Whites to minorities compared to the metropolitan-level models used above. Still, for Latinos and Blacks, there is a positive and significant relationship between White exposure to minorities and permitted density interacted with the minority population share in the MSA. The effect, however, is insignificant for Asians. The results imply that permitted density significantly increases integration so long as the minority population share in the metro area is >11%. Using only thirty-seven MSAs, it is remarkable that these regressions were still able to identify a significant effect from density zoning on integration, even conditional on the individual attributes of Whites.

Table 12.

OLS Regression of White Exposure to Minorities on Zoning and Opposition to Living in Majority–Minority Neighborhoods

 % Latino in Community % Asian in Community % Black in Community 
 
Average permitted density in MSA × % minority in MSA 60.25** (24.55) −52.79 (59.61) 82.36** (33.36) 
Average permitted density in MSA −5.656 (5.812) 0.238 (2.294) −9.337** (3.978) 
% Minority in MSA −180.3 (112.7) 273.0 (268.6) −204.2** (98.39) 
Ln MSA population 4.225 (5.831) −2.130 (3.574) −4.805** (1.939) 
Ln Land area MSA −11.76** (4.328) 1.401 (3.531) 2.899 (3.397) 
Ln Per capita MSA income 28.13** (12.71) −3.160 (7.486) 13.76 (8.678) 
Individual-level covariates for White respondents 
Opposition to living in majority-minority neighborhood 0.150 (0.897) 0.513 (0.736) −1.578* (0.853) 
    Age −0.118 (0.338) −0.114 (0.272) 0.0971 (0.386) 
    Age squared 0.000108 (0.00330) 0.000514 (0.00277) −0.00191 (0.00384) 
    Married −1.410 (1.572) −3.092** (1.484) −3.143 (1.868) 
    Male −1.429 (1.796) −0.834 (1.354) −2.256 (2.344) 
    Number of children −0.610 (0.449) 0.455 (0.906) −0.106 (0.574) 
   Less than high-school education 5.817* (2.973) 1.507 (2.623) 4.883 (3.394) 
    College education −1.266 (1.454) −0.663 (0.710) −2.886 (1.990) 
    Ln of income −3.275*** (0.815) −0.486 (0.861) −4.206*** (1.522) 
    Works full time 0.566 (1.968) −1.389 (0.973) 3.184 (2.467) 
    Works part time −1.167 (3.436) 17.10 (11.60) 7.032 (4.524) 
    Homeowner 0.168 (1.523) 1.197 (1.283) −0.690 (1.765) 
    Lives in city −2.847* (1.664) −0.657 (1.751) 2.137 (2.272) 
    Lives in rural area −5.006** (2.058) −4.025** (1.847) −2.875 (3.518) 
    Lives in suburb −2.334 (2.732) 0.797 (1.422) 2.683 (2.102) 
    Constant −168.7 (144.5) 61.79 (89.04) −32.23 (97.43) 
    Observations 399 397 390 
    State fixed effects Yes Yes Yes 
    Adjusted R2 0.225 0.232 0.126 
 % Latino in Community % Asian in Community % Black in Community 
 
Average permitted density in MSA × % minority in MSA 60.25** (24.55) −52.79 (59.61) 82.36** (33.36) 
Average permitted density in MSA −5.656 (5.812) 0.238 (2.294) −9.337** (3.978) 
% Minority in MSA −180.3 (112.7) 273.0 (268.6) −204.2** (98.39) 
Ln MSA population 4.225 (5.831) −2.130 (3.574) −4.805** (1.939) 
Ln Land area MSA −11.76** (4.328) 1.401 (3.531) 2.899 (3.397) 
Ln Per capita MSA income 28.13** (12.71) −3.160 (7.486) 13.76 (8.678) 
Individual-level covariates for White respondents 
Opposition to living in majority-minority neighborhood 0.150 (0.897) 0.513 (0.736) −1.578* (0.853) 
    Age −0.118 (0.338) −0.114 (0.272) 0.0971 (0.386) 
    Age squared 0.000108 (0.00330) 0.000514 (0.00277) −0.00191 (0.00384) 
    Married −1.410 (1.572) −3.092** (1.484) −3.143 (1.868) 
    Male −1.429 (1.796) −0.834 (1.354) −2.256 (2.344) 
    Number of children −0.610 (0.449) 0.455 (0.906) −0.106 (0.574) 
   Less than high-school education 5.817* (2.973) 1.507 (2.623) 4.883 (3.394) 
    College education −1.266 (1.454) −0.663 (0.710) −2.886 (1.990) 
    Ln of income −3.275*** (0.815) −0.486 (0.861) −4.206*** (1.522) 
    Works full time 0.566 (1.968) −1.389 (0.973) 3.184 (2.467) 
    Works part time −1.167 (3.436) 17.10 (11.60) 7.032 (4.524) 
    Homeowner 0.168 (1.523) 1.197 (1.283) −0.690 (1.765) 
    Lives in city −2.847* (1.664) −0.657 (1.751) 2.137 (2.272) 
    Lives in rural area −5.006** (2.058) −4.025** (1.847) −2.875 (3.518) 
    Lives in suburb −2.334 (2.732) 0.797 (1.422) 2.683 (2.102) 
    Constant −168.7 (144.5) 61.79 (89.04) −32.23 (97.43) 
    Observations 399 397 390 
    State fixed effects Yes Yes Yes 
    Adjusted R2 0.225 0.232 0.126 

Notes: Standard errors in parentheses, clustered on metropolitan areas. Data from thirty-seven MSAs.

*P < 0.1, **P < 0.05, ***P < 0.01.

Overall stated White opposition to living in half minority neighborhoods is weakly predictive of the level of minorities in the respondents “community.” It is significant, but only marginally so, in predicting lower exposure to Blacks. White opposition to Black neighborhoods shaves roughly 3% points off the share of a respondent's community that is Black. Income is the most consistent predictor of exposure to Blacks. Poor Whites are much more likely to be living near Blacks and Latinos. These findings are comparable to those of Cutler, Glaeser, and Vigdor (1999) who, using a larger sample, found White (but not Black) attitudes toward segregation associated with higher Black–White segregation. They found no significant relationship between the neighborhood preferences question used here and segregation.

3.6.2. Aggregation bias.

The results reported above are based on simple metropolitan averages for permitted density in each jurisdiction within the metropolitan area, and this section will explore alternative aggregation strategies and efforts to address survey bias. The previous method assumes that the population of interest is the set of the fifty largest metropolitan areas and weighs each jurisdiction equally. This is arguably the most reasonable strategy since each jurisdiction will have its own regulations and Pendall, Puentes, and Martin (2006) surveyed in such a way to get a representative sample of jurisdiction types for each MSA.

Still suburban jurisdictions are different than large populous jurisdictions and more likely to have restrictive zoning. So, MSAs that surveyed more suburbs may have stricter zoning even if suburban zoning laws are everywhere more restrictive. Above, the regressions controlled for the suburban housing share, but this may not be sufficient. To directly capture a potential selection bias, Table A4 controls directly for the percentage of jurisdictions with populations smaller than 150,000 in the metropolitan area that were included in the survey. Doing so increases the magnitude and significance of the coefficient on permitted density. So, if anything, suburban versus city survey bias leads to a simple measurement error that understates the importance of zoning.

Another possibility is that the survey only captures a nonrepresentative amount of the metropolitan area's land. This is unlikely to be a problem given Pendall's survey strategy, but one can do two things to address it. First, metropolitan areas can be given more weight in the regression according to the share of land area captured by the survey. Using land share surveyed in a weighted least squares regression also increases the significance of the coefficients, implying that the standard errors reported above are too high.

Second, one could aggregate the jurisdictions so that larger jurisdictions are given more weight. The resulting metropolitan indicator would not be representative of jurisdictions but it would be more representative of land area. Doing this does, in fact, tend to drive the coefficients toward zero, as Table A4 shows. This means that the results reported in the main body of this paper may be driven by jurisdictional differences in regulation more so than by differences in the regulation of land area per se. This does not invalidate the results, and in some sense it is highly intuitive. The regulation, or lack thereof, of empty vast tracks of potentially undesirable land, which is especially common in the west, has a trivial effect on the level of segregation. What matters is the regulation of jurisdictions in desirable locations around central cities.

Regional effects in the 2000 levels model.

As a final cross-section robust check, regional effects are incorporated into the model. Some readers may be concerned that unobserved regional effects could be driving the results. Indeed, there is a strong regional pattern in land regulations as Pendall, Puentes, and Martin (2006) showed. While regional effects were “subtracted out” in the dynamic fixed-effects model, they may explain much of the level effect. With respect to density regulation, the west is the most liberal, followed by the south, and both are significantly more liberal than the Midwest and north-east, which are similar to one another. The pattern is so striking that using regional controls drowns out a large share of the variation and makes the zoning effect insignificant in the baseline regressions shown in Table 3. The reason is likely that intra-regional variation is only a fraction of interregional variation.9 Yet, for Black segregation, the effect of zoning remains robust using the isolation index and marginally significant using the dissimilarity index when the full spectrum of historical controls is included in the regression along with regional dummies. This implies that the regional effects were absorbing more fundamental historical patterns. None of the regions were significant for Black segregation, and the regional dummies had inconsistent associations with Latino and Asian segregation depending on the segregation indicator. These results are shown in Table A5.

The results from Table 10 showing that average permitted density, when instrumented with year of statehood, predicts significantly less segregation were also reinvestigated with regional effects. These results, which are available upon request, continue to show a strong relationship between predicted zoning and segregation for Latinos but not for Asian and Blacks. The coefficients are still negative but not distinguishable from zero. For Latinos, the instrument continues to be relevant (according to the Anderson correlation) but not for Asians and Blacks. In those models, the regional effects are too closely correlated with year of statehood for year of statehood to identify zoning.

Arguably, including regional dummies is not a desirable strategy. Explaining within regional variation requires many observations within regions, which is not available here. Historic regional differences with slavery and oppression are obviously relevant, and controlling for the south does make sense, but including only a southern variable does not change the results. Still the differences between regions should not be exaggerated. Slavery was permitted in northern states throughout most of the eighteenth century, and while Jim Crow laws were more formalized in the South, they pervaded the nation. For example, deed restrictions were widely used in northern cities to keep Blacks out of White neighborhoods and many northern unions barred Black membership formally (in their constitutions) until late into the 20th century (see Marshall, 1963; Massey and Denton, 1993; Quadagno, 1994). Other variables correlated with regions like agricultural activity, weather, age of housing, and manufacturing patterns were already controlled for in the model without regional effects (and union membership in 2000 was added as a robustness check without changing the results), so it is not clear whether or not the regional dummies pick up anything else of theoretical relevance. This issue does encourage caution in interpreting the results using year of statehood as an instrument, and the reader can decide what to make of the causal link between zoning and segregation based on this evidence.

Conclusion

In this paper I have made an effort to explain two stylized facts of segregation: the decrease in African American segregation and the rise of Asian and Hispanic segregation. The cross-sectional results find that density zoning and relative incomes explain most of the metropolitan variation and that they have roughly the same predictive power. Looking at other theories of segregation, in light of the evidence presented here, discriminatory attitudes on the part of Whites do not account for the variation in observed segregation, and density zoning was associated with the exposure of White respondents to Blacks and Latinos even after adjusting for respondents' attitudes about living near those groups. Although real estate market discrimination remains a pervasive problem throughout the United States, there is no persuasive evidence that it accounts for any of the variation in segregation. Greater interjurisdictional choice, in the context of local reliance on taxation à la Tiebout (1956), explains a small share of observed segregation patterns, and urban form, including the use of public transportation by minorities and suburbanization patterns, has no consistent effect on segregation.

The main result that zoning is strongly associated and perhaps causally linked to segregation is robust to a variety of different estimation strategies, including using year of statehood as an instrumental variable for density zoning, using an alternative measure of density zoning, controlling for all other aspects of land regulation, adjusting for individual characteristics, and using different aggregation strategies. Finally, changes from 1990 to 2000 were explained in large part by density zoning, using either the 2SLS estimates or metropolitan fixed effects with the reduced sample. A fairly conservative summary estimate could claim that somewhere between 25% and 50% of contemporary segregation is explained by anti-density zoning.

Histories of zoning, such as Fogelson (2005) and Fischel (2004), have acknowledged racial and class bias at the heart of the legal efforts to exclude affordable housing. Indeed, anti-density zoning can be seen as continuing a tradition of legal barriers, from racial zoning to covenants to urban renewal, that were part of a nationwide effort, both private and public, on the part of European Americans to live in homogenous communities. Aside from prejudice, Fischel (2004) argues that zoning should be thought of as homeowners insurance, and likewise Fogelson (2005) stresses its goals of permanence and stability; Rothwell and Massey (2009) point to fiscal incentives related to density. More empirical work needs to be done to understand the motivations, and greater effort to measure and document changes in zoning would strengthen the empirical case. Whatever the motivations, however, the disparate impacts of zoning are becoming clear. Anti-density zoning is strongly associated with the segregation of the three largest minority groups in the United States; moreover, evidence and straightforward logic suggest that its effect is causal. After so many years of enabling and protecting the elite local interests that create and enforce low-density regulatory regimes, liberalizing federal policy action will likely be necessary if this continuing barrier to racial equality is to be dismantled.

The author has benefited from the suggestions of Douglas Masssey, Jesse Rothstein, Jenny Schuetz, Mark Shroder, Matthew Steinberg, John Donohue, and anonymous referees.

The author would especially like to thank Rolf Pendall for generously supplying his regulation data and Joseph Gyourko and his colleagues at the Wharton School for making their zoning data publicly available. This paper was first written while the author was affiliated with Princeton University and does not necessarily reflect the views of the Brookings Institution or its researchers.

Appendix

Real Estate Market Discrimination

In one of the most comprehensive experimental studies to date, Housing and Urban Development (HUD) auditors were assigned equal incomes, debt levels, wealth, educational status, family circumstances, job characteristics, and housing preferences and sent out to twenty large metropolitan areas (Turner et al., 2002). Turner et al. (2002) find very little evidence that is consistent with the hypothesis that segregation is largely caused by contemporary discrimination. While discrimination that favored Whites over minorities was widespread—occurring in roughly one out of every five audits—the differences between minority groups were largely insignificant in terms of consistent adverse treatment (Turner et al., 2002; Turner and Ross, 2003).

A closer look reveals even less evidence that real estate market discrimination drives discrimination. Racial steering is clearly the most relevant form of real estate discrimination. Racial steering occurs when Whites are “steered” toward homes in neighborhoods that have a higher percentage of Whites than the homes shown to the minority group in question. Aggregating metropolitan and state-level data, the authors determined that Whites were favored geographically over Blacks, on balance, just 3.5% of the time in terms of homes inspected and 3.7% of cases in terms of homes recommended. This difference was statistically significant, but it is hardly widespread enough to explain the high levels of observed segregation. This form of discrimination actually increased significantly from 1989 to 2000 (by 5.9%) even though African Americans experienced, on average, considerably less segregation.

Even less persuasively for the housing markets hypothesis, non-White Latinos were often shown homes in “Whiter” neighborhoods than Whites, but in 2000 the difference was not statistically significant. By contrast in 1989, Latinos were shown homes in neighborhoods with significantly fewer Whites in 14.7% of cases (see Turner et al., 2002, 3.12–3.16). So the trend in real estate markets should have encouraged considerable Latino integration from 1990 to 2000; however, as we know, the opposite was true. At the MSA level, racial steering did not significantly favor Whites over Blacks in any MSA or state, and it only significantly favored Whites over Latinos in Pueblo but in no other MSA or state (Turner et al., 2002, Annex 6).

Finally, the racial steering of Asians and Pacific Islanders was not significantly different from the steering of African Americans in the MSAs where discrimination against both groups was tested (Turner and Ross, 2003, Annex 6). Moreover, in a few cases, dark-skinned Asians were actually less likely to be discriminated against than light-skinned Asians, including for measures of racial steering.

Table A2 pools the metropolitan indicators of discrimination against Blacks together and runs a simple regression with density zoning as an added control. In no case was discrimination—measured comprehensively or by racial steering—associated with significantly more Black segregation at the metropolitan level. Density zoning remained significant in all but one of the specifications, despite the small sample.

Table A1.

Regression of Property Tax Rate on Share of Population Living in Rural Areas and Other Characteristics of County for all U.S. Counties in 2000

 Effective Property Tax Rate 
% Living in rural areas −2.516*** (0.275)  −2.014*** (0.197)  
Population in millions  3.788*** (0.555)  1.248*** (0.471) 
(Population in millions)2  −4.69 × 10−7*** (8.00 × 10−8 −1.55 × 10−7*** (5.91 × 10−8
% Population enrolled in school   0.0659 (2.219) −2.840 (2.387) 
Homeowner rate   −3.320*** (1.003) −7.339*** (0.743) 
Median household income   0.000101*** (1.02 × 10−50.000136*** (1.15 × 10−5
Average property value   −2.34 × 10−5*** (2.40 × 10−6−2.66 × 10−5*** (2.58 × 10−6
% Black   1.446*** (0.555) 1.053* (0.563) 
% Non-White and non-Black   0.432 (0.557) 0.866 (0.548) 
Constant 11.33*** (0.186) 9.509*** (0.0882) 12.06*** (0.650) 13.31*** (0.568) 
Observations 3,060 3,137 3,060 3,137 
Adjusted R2 0.029 0.021 0.860 0.858 
 Effective Property Tax Rate 
% Living in rural areas −2.516*** (0.275)  −2.014*** (0.197)  
Population in millions  3.788*** (0.555)  1.248*** (0.471) 
(Population in millions)2  −4.69 × 10−7*** (8.00 × 10−8 −1.55 × 10−7*** (5.91 × 10−8
% Population enrolled in school   0.0659 (2.219) −2.840 (2.387) 
Homeowner rate   −3.320*** (1.003) −7.339*** (0.743) 
Median household income   0.000101*** (1.02 × 10−50.000136*** (1.15 × 10−5
Average property value   −2.34 × 10−5*** (2.40 × 10−6−2.66 × 10−5*** (2.58 × 10−6
% Black   1.446*** (0.555) 1.053* (0.563) 
% Non-White and non-Black   0.432 (0.557) 0.866 (0.548) 
Constant 11.33*** (0.186) 9.509*** (0.0882) 12.06*** (0.650) 13.31*** (0.568) 
Observations 3,060 3,137 3,060 3,137 
Adjusted R2 0.029 0.021 0.860 0.858 

Notes: Robust standard errors in parentheses. The effective property tax was calculated by economists at the National Association of Home Builders, http://www.nahb.org/generic.aspx?genericContentID=35450, using 2000 Census data on self-reported property taxes and home values.

*P < 0.1, **P < 0.05, ***P < 0.01.

Table A2.

OLS Regression of Segregation on Housing Market Discrimination and Zoning

 Black–White Dissimilarity Index 
 
Average permitted density in MSA −0.0432 (0.0292) −0.0658* (0.0340) −0.0832** (0.0305) −0.0847** (0.0309) 
% Black 0.503** (0.210) 0.356 (0.257) 0.167 (0.193) 0.260 (0.243) 
Consistent adverse treatment in rentals −0.0100** (0.00356)    
Consistent adverse treatment in home sales  0.00103 (0.00572)   
Racial steering in recommendations   0.00325 (0.00331)  
Racial steering in homes shown    0.000571 (0.00689) 
Constant 0.943*** (0.0935) 0.818*** (0.219) 0.913*** (0.127) 0.911*** (0.113) 
Observations 14 14 13 13 
Adjusted R2 0.592 0.206 0.305 0.255 
 Black–White Dissimilarity Index 
 
Average permitted density in MSA −0.0432 (0.0292) −0.0658* (0.0340) −0.0832** (0.0305) −0.0847** (0.0309) 
% Black 0.503** (0.210) 0.356 (0.257) 0.167 (0.193) 0.260 (0.243) 
Consistent adverse treatment in rentals −0.0100** (0.00356)    
Consistent adverse treatment in home sales  0.00103 (0.00572)   
Racial steering in recommendations   0.00325 (0.00331)  
Racial steering in homes shown    0.000571 (0.00689) 
Constant 0.943*** (0.0935) 0.818*** (0.219) 0.913*** (0.127) 0.911*** (0.113) 
Observations 14 14 13 13 
Adjusted R2 0.592 0.206 0.305 0.255 

Notes: Robust standard errors in parentheses. All discrimination measures are from Turner et al. (2002) and are MSA measures from 2000. Adverse treatment includes any form of discrimination, and racial steering refers to the net bias that Whites were given over Blacks in terms of being shown neighborhoods with higher percentages of White residents. Data are available at http://www.huduser.org/Publications/pdf/Phase1_Annex_8.pdf

*P < 0.1, **P < 0.05, ***P < 0.01.

Table A3.

Individual- and Metropolitan-Level Data on Racialized Housing Preferences from the 2000 GSS, Pendall, Puentes, and Martin (2006), and the U.S. Census Bureau

 Observations Mean Standard Deviation Minimum Maximum 
Whites opposed to living in half Latino neighborhood 2,213 0.13 0.34 
Whites opposed to living in half Asian neighborhood 2,213 0.09 0.29 
Whites opposed to living in half Black neighborhood 2,213 0.24 0.43 
Latinos opposed to living in half Black neighborhood 213 0.07 0.25 
Blacks opposed to living in half Black neighborhood 429 0.07 0.26 
Latinos opposing half White neighborhood 213 0.01 0.12 
Asians opposing half White neighborhood 0.00 0.00 
Blacks opposing half White neighborhood 429 0.04 0.19 
% Latino in community of White respondent 1,008 13.75 16.85 
% Asian in community of White respondent 997 7.03 10.94 
% Black in community of White respondent 1,038 15.00 16.54 
Metropolitan-level covariates (for thirty-six MSAs) 
    Permitted density zoning 1,376 3.36 0.66 2.17 4.67 
    % Latino population 2,090 0.14 0.14 0.01 0.57 
    % Asian 2,090 0.05 0.05 0.01 0.25 
    % Black 2,090 0.14 0.09 0.01 0.43 
 Observations Mean Standard Deviation Minimum Maximum 
Whites opposed to living in half Latino neighborhood 2,213 0.13 0.34 
Whites opposed to living in half Asian neighborhood 2,213 0.09 0.29 
Whites opposed to living in half Black neighborhood 2,213 0.24 0.43 
Latinos opposed to living in half Black neighborhood 213 0.07 0.25 
Blacks opposed to living in half Black neighborhood 429 0.07 0.26 
Latinos opposing half White neighborhood 213 0.01 0.12 
Asians opposing half White neighborhood 0.00 0.00 
Blacks opposing half White neighborhood 429 0.04 0.19 
% Latino in community of White respondent 1,008 13.75 16.85 
% Asian in community of White respondent 997 7.03 10.94 
% Black in community of White respondent 1,038 15.00 16.54 
Metropolitan-level covariates (for thirty-six MSAs) 
    Permitted density zoning 1,376 3.36 0.66 2.17 4.67 
    % Latino population 2,090 0.14 0.14 0.01 0.57 
    % Asian 2,090 0.05 0.05 0.01 0.25 
    % Black 2,090 0.14 0.09 0.01 0.43 

Notes: All averages were weighted by the inverse probability of selection, in the following way: forumla. The questions (liveblk, livehsps, liveasns, livewhts) are from the 2000 GSS and ask respondents how they would feel about living in a neighborhood that was 50% of group X. Answers that opposed or strongly opposed were coded as opposed.

Table A4.

Baseline Regressions under Various Sample Selection Assumptions and Weighting Strategies

 Latino Asian Black 
 Adjusting for responses from suburban jurisdictions 
Average permitted density in MSA −0.0704*** (0.0254) −0.0564** (0.0244) −0.0659** (0.0267) 
Adjusted R2 0.739 0.403 0.635 
 MSAs weighted by share of land surveyed 
Average permitted density in MSA −0.0764*** (0.0247) −0.0470** (0.0186) −0.0705** (0.0263) 
Adjusted R2 0.819 0.517 0.670 
 Land-area-weighted zoning 
Average permitted density in MSA, weighted by land area −0.0117 (0.0163) −0.0125 (0.0143) −0.000895 (0.0185) 
Adjusted R2 0.678 0.298 0.587 
 Latino Asian Black 
 Adjusting for responses from suburban jurisdictions 
Average permitted density in MSA −0.0704*** (0.0254) −0.0564** (0.0244) −0.0659** (0.0267) 
Adjusted R2 0.739 0.403 0.635 
 MSAs weighted by share of land surveyed 
Average permitted density in MSA −0.0764*** (0.0247) −0.0470** (0.0186) −0.0705** (0.0263) 
Adjusted R2 0.819 0.517 0.670 
 Land-area-weighted zoning 
Average permitted density in MSA, weighted by land area −0.0117 (0.0163) −0.0125 (0.0143) −0.000895 (0.0185) 
Adjusted R2 0.678 0.298 0.587 

Notes: Fifty observations. Robust standard errors in parentheses, clustered on state. Only the coefficients on permitted density are shown for nine separate regressions. The first row adjusts for the percentage of sampled jurisdictions that are in suburbs (defined as population <150,000). This variable, the percentage of jurisdictions in the MSA sample, was never significant. The second row weighted all metropolitan averages by the share of metropolitan land area sampled. The third row weighted zoning at the jurisdiction level. It calculates permitted density as the metropolitan average weighted by the share of land in each jurisdiction, giving more weight to larger jurisdictions. The controls are the same as in previous Table 3: population density, % minority, % foreign born, minority income/White income; the full set includes the property tax rate, local share of local revenue for each state, the log of per capita state taxes, the suburban housing units/central city housing, suburban rent/central city rent, median home prices/median rent, the average property tax share of local revenue for each state, the log of median household income, and the percent of the minority group using public transportation or carpooling.

*P < 0.1, **P < 0.05, ***P < 0.01.

Table A5.

OLS Regression of Permitted Density on Regional Effects, Full Controls, and Historical Characteristics of Metropolitan Area

 Dissimilarity Isolation 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.00110 (0.0224) −0.00196 (0.0164) −0.0317* (0.0185) −0.00127 (0.0271) −0.0113 (0.0166) −0.0490** (0.0201) 
% Employment in agriculture in 1970 −0.453 (0.570) 0.335 (0.547) −2.265*** (0.711) −0.418 (0.819) 0.994* (0.522) −2.900*** (1.039) 
% Employment in manufacturing in 1970 −0.0525 (0.118) 0.0818 (0.0958) −0.0170 (0.112) −0.196 (0.146) −0.0310 (0.113) 0.0312 (0.120) 
% 2000 units built before 1960 0.244* (0.143) 0.190 (0.122) 0.423*** (0.113) 0.0629 (0.151) 0.388*** (0.132) 0.471*** (0.158) 
Population density in 1910 5.95 × 10−5 (7.56 × 10−51.44 × 10−5 (6.81 × 10−50.000157* (8.24 × 10−50.000163 (0.000112) 0.000202*** (6.50 × 10−50.000200* (0.000101) 
Year of statehood −0.000459 (0.000424) −0.00105*** (0.000338) −0.000679* (0.000351) −0.000447 (0.000390) −0.00111*** (0.000275) −0.000773* (0.000451) 
Thousands of miles to nearest Mexican border −0.0395 (0.0280) −0.0737*** (0.0244) −0.0546*** (0.0153) −0.0406 (0.0299) −0.0650*** (0.0201) −0.0605*** (0.0170) 
Midwest 0.0991** (0.0469) 0.0351 (0.0496) 0.0491 (0.0506) 0.0488 (0.0481) −0.0862* (0.0507) 0.0326 (0.0676) 
Northeast 0.130* (0.0767) −0.0103 (0.0742) −0.0207 (0.0688) 0.0867 (0.0775) −0.189** (0.0705) −0.0732 (0.0916) 
South −0.00332 (0.0400) 0.0373 (0.0442) −0.0606 (0.0447) −0.0165 (0.0426) −0.0637* (0.0373) −0.0308 (0.0550) 
Constant 0.454 (0.827) 1.364** (0.664) 3.867*** (0.962) −0.257 (1.553) 0.842 (0.684) 4.899*** (1.543) 
Observations 50 50 50 50 50 50 
Adjusted R2 0.834 0.704 0.889 0.936 0.917 0.926 
 Dissimilarity Isolation 
 Latino Asian Black Latino Asian Black 
 
Average permitted density in MSA −0.00110 (0.0224) −0.00196 (0.0164) −0.0317* (0.0185) −0.00127 (0.0271) −0.0113 (0.0166) −0.0490** (0.0201) 
% Employment in agriculture in 1970 −0.453 (0.570) 0.335 (0.547) −2.265*** (0.711) −0.418 (0.819) 0.994* (0.522) −2.900*** (1.039) 
% Employment in manufacturing in 1970 −0.0525 (0.118) 0.0818 (0.0958) −0.0170 (0.112) −0.196 (0.146) −0.0310 (0.113) 0.0312 (0.120) 
% 2000 units built before 1960 0.244* (0.143) 0.190 (0.122) 0.423*** (0.113) 0.0629 (0.151) 0.388*** (0.132) 0.471*** (0.158) 
Population density in 1910 5.95 × 10−5 (7.56 × 10−51.44 × 10−5 (6.81 × 10−50.000157* (8.24 × 10−50.000163 (0.000112) 0.000202*** (6.50 × 10−50.000200* (0.000101) 
Year of statehood −0.000459 (0.000424) −0.00105*** (0.000338) −0.000679* (0.000351) −0.000447 (0.000390) −0.00111*** (0.000275) −0.000773* (0.000451) 
Thousands of miles to nearest Mexican border −0.0395 (0.0280) −0.0737*** (0.0244) −0.0546*** (0.0153) −0.0406 (0.0299) −0.0650*** (0.0201) −0.0605*** (0.0170) 
Midwest 0.0991** (0.0469) 0.0351 (0.0496) 0.0491 (0.0506) 0.0488 (0.0481) −0.0862* (0.0507) 0.0326 (0.0676) 
Northeast 0.130* (0.0767) −0.0103 (0.0742) −0.0207 (0.0688) 0.0867 (0.0775) −0.189** (0.0705) −0.0732 (0.0916) 
South −0.00332 (0.0400) 0.0373 (0.0442) −0.0606 (0.0447) −0.0165 (0.0426) −0.0637* (0.0373) −0.0308 (0.0550) 
Constant 0.454 (0.827) 1.364** (0.664) 3.867*** (0.962) −0.257 (1.553) 0.842 (0.684) 4.899*** (1.543) 
Observations 50 50 50 50 50 50 
Adjusted R2 0.834 0.704 0.889 0.936 0.917 0.926 

Notes: Robust standard errors in parentheses, clustered on state. Full list of controls are included.

*P < 0.1, **P < 0.05, ***P < 0.01.

References

Alesina
Alberto
Baqir
Reza
Hoxby
Caroline
“Political Jurisdictions in Heterogeneous Communities”
Journal of Political Economy
 , 
2004
, vol. 
112
 (pg. 
348
-
96
)
Bayer
Patrick
McMillan
Robert
Rueben
Kim S.
“What Drives Racial Segregation? New Evidence Using Census Microdata”
Journal of Urban Economics
 , 
2004
, vol. 
56
 (pg. 
514
-
35
)
Berry
Christopher
“Land Use Regulation and Residential Segregation: Does Zoning Matter?”
 , 
2001
 
3(2) American Law and Economics Review 251–74
Borjas
George J
“To Ghetto or Not to Ghetto: Ethnicity and Residential Segregation”
Journal of Urban Economics
 , 
1998
, vol. 
44
 (pg. 
228
-
53
)
Brueckner
Jan
“Strategic Interaction Among Local Governments: The Case of Growth Controls,”
 , 
1998
 
44 Journal of Urban Economics 438–67
Byun
Pillsung
Brigitte Waldorf
S
Adrian
Esparza
“Spillovers and Local Growth Controls: An Alternative Perspective on Suburbanization,”
 , 
2005
 
36(2) Growth and Change 196--219
Card
David
Mas
Alexandre
Rothstein
Jesse
“Tipping and the Dynamics of Segregation”
Quarterly Journal of Economics
 , 
2008
, vol. 
123
 (pg. 
177
-
218
)
———. 2010
“Are Mixed Neighborhoods Always Unstable? Two-Sided and One-Sided Tipping”
 
Harriet Newburger, Eugenie Birch, and Susan Wachter, eds., Reinventing Older Communities: How Does Place Matter? University of Pennsylvania Press
Cutler
David M
Glaeser
Edward L
Are Ghettos Good or Bad?
 , 
1995
 
Working Paper No. W5163, National Bureau of Economic Research, Inc
———
“Are Ghettos Good or Bad?”
Quarterly Journal of Economics
 , 
1997
, vol. 
112
 (pg. 
827
-
72
)
Cutler
David M
Glaeser
Edward L
Vigdor
Jacob
“The Rise and Decline of the American Ghetto”
Journal of Political Economy
 , 
1999
, vol. 
107
 (pg. 
455
-
50
)
———
When are Ghettos Bad? Lessons from Immigrant Segregation in the United States
 , 
2006
 
Working Paper, Duke University and NBER
———
“Is the Melting Pot Still Hot? Explaining the Resurgence of Immigrant Segregation”
Review of Economics and Statistics
 , 
2008
, vol. 
90
 (pg. 
478
-
97
)
Davis
James Allan
Smith
Tom W
General Social Surveys, 1972–2006
 , 
2009
 
[machine-readable data file] Principal Investigator, James A. Davis; Director and Co-Principal Investigator, Tom W. Smith; Co-Principal Investigator, Peter V. Marsden; Sponsored by National Science Foundation. NORC, ed. Chicago: National Opinion Research Center [producer]; Storrs, CT: The Roper Center for Public Opinion Research, University of Connecticut [distributor]
Easterly
William
“Empirics of Strategic Interdependence: The Case of the Racial Tipping Point”
B. E. Journal of Macroeconomics
 , 
2009
, vol. 
9
 (pg. 
1
-
33
)
Fischel
William A
The Economics of Zoning Laws: A Property Rights Approach to American Land Use Controls
 , 
1985
Baltimore, MD
The Johns Hopkins University Press
———
2004. “An Economic History of Zoning and a Cure for Its Exclusionary Effects”
41 Urban Studies
  
317–40
Fogelson
Robert M
Downtown: Its Rise and Fall, 1880–1950
 , 
2003
New Haven, CT
Yale University Press
———
Bourgeois Nightmares: Suburbia 1870–1930
 , 
2005
New Haven, CT
Yale University Press
Galster
George
“Black and White Preferences for Neighborhood Racial Composition”
AREUEA Journal
 , 
1982
, vol. 
10
 (pg. 
39
-
66
)
———
“Residential Segregation in American Cities: A Contrary Review”
Population Research and Policy Review
 , 
1988
, vol. 
7
 (pg. 
93
-
112
)
Glaeser
Edward L
Gyourko
Joseph
“Urban Decline and Durable Housing”
Journal of Political Economy
 , 
2005
, vol. 
113
 (pg. 
345
-
75
)
Glaeser
Edward L
Shapiro
Jesse
“Urban Growth in the 1990s. Is City Living Back?”
Journal of Regional Science
 , 
2003
, vol. 
43
 (pg. 
139
-
65
)
Glaeser
Edward L
Vigdor
Jacob
Racial Segregation in the 2000 Census: Promising News
 , 
2001
Washington
Brookings Institution, Center on Urban and Metropolitan Policy
Glaeser
Edward L
Ward
Bryce
The Causes and Consequences of Land Use Regulation: Evidence from Greater Boston
 , 
2006
 
Discussion Paper Number 2124, Harvard Institute of Economic Research
Hirsch
Arnold
Making the Second Ghetto: Race and Housing in Chicago, 1940–1960
 , 
1983
Chicago
University of Chicago Press
Iceland
John
Scopilitti
Melissa
“Immigrant Residential Segregation in U.S. Metropolitan Areas, 1990-2000,”
 , 
2008
 
45 Demography 79--94
Iceland
John
Weinberg
Daniel H
Steinmetz
Erika
Racial and Ethnic Residential Segregation in the United States 1980–2000
 , 
2002
Washington, DC
U.S. Bureau of the Census
Ihlanfeldt
Keith
Scafidi
Benjamin
“Black Self-segregation as a Cause of Housing Segregation: Evidence from the Multi-city Study of Urban Inequality”
Journal of Urban Economics
 , 
2002
, vol. 
51
 (pg. 
366
-
90
)
Jackson
Kenneth
Crabgrass Frontier: The Suburbanization of the United States
 , 
1985
New York
Oxford University Press
King
TA
Mieszkowski
P
“Racial Discrimination, Segregation, and the Price of Housing”
Journal of Political Economy
 , 
1973
, vol. 
81
 (pg. 
590
-
606
)
Malpezzi
Stephen
“Housing Prices, Externalities, and Regulation in the U.S. Metropolitan Areas”
 , 
1996
 
7(2) Journal of Housing Research 209--41
Marshall
Ray
“The Negro and Organized Labor”
Journal of Negro Education
 , 
1963
, vol. 
32
 (pg. 
375
-
89
)
Massey
D
Denton
N
“The Dimensions of Residential Segregation”
Social Forces
 , 
1988
, vol. 
67
 (pg. 
281
-
315
)
———
American Apartheid
 , 
1993
Cambridge, MA
Harvard University Press
Massey
Douglas
Fischer
Mary
“Does Rising Income Bring Integration? New Results for Blacks, Hispanics, and Asians in 1990”
Social Science Research
 , 
1999
, vol. 
28
 (pg. 
316
-
26
)
Meltzer
R
Schuetz
J
“The Most Popular Kids in the Class: Diffusion of Inclusionary Zoning across San Francisco Bay Area Governments.”
 , 
2008
 
Presented at the Association for Public Policy Analysis and Management Fall Conference, November 7, 2008: Los Angeles, CA
Nelson
A
T. Sanchez, C. Dawkins. 2004
“The Effect of Urban Containment and Mandatory Housing Elements on Racial Segregation in U.S. Metropolitan Areas, 1990-2000,”
  
26 Journal of Urban Affairs 339–50
Nguyen
Mai Thi
“Why Do Communities Mobilize Against Growth: Growth Pressures, Community Status, Metropolitan Hierarchy, or Strategic Interaction?”
 , 
2009
 
31(1) Journal of Urban Affairs 25–43
Orfield
Myron
American Metropolitics: The New Suburban Reality
 , 
2002
Washington, DC
Brookings Institution Press
Pendall
Rolf
“Local Land-Use Regulation and the Chain of Exclusion”
Journal of the American Planning Association
 , 
2000
, vol. 
66
 (pg. 
125
-
42
)
Pendall
Rolf
Puentes
R
Martin
J.
“From Traditional to Reformed: A Review of the Land Use Regulations in the Nation's 50 Largest Metropolitan Areas.”
 , 
2006
Washington
The Brookings Institution, Research Brief
Quadagno
Jill
The Color of Welfare: How Racism Undermined the War on Poverty
 , 
1994
New York
Oxford University Press
Ross
Stephen L
“Understanding Racial Segregation: What is Known About the Effect of Housing Discrimination”
 , 
2008
 
Working Paper 2008-15R, University of Connecticut
Rothwell
Jonathan
Massey
Douglas
“The Effect of Density Zoning on Racial Segregation in U.S. Urban Areas”
Urban Affairs Review
 , 
2009
, vol. 
44
 (pg. 
779
-
806
)
———
“Density Zoning and Class Segregation in U.S. Metropolitan Areas”
Social Science Quarterly
 , 
2010
, vol. 
91
 (pg. 
1123
-
43
)
Schuetz
J
Been
V
Meltzer
R.
Silver Bullet or Trojan Horse? The Effects of Inclusionary Zoning on Local Housing Markets
 , 
2008
New York
Furman Center for Real Estate and Urban Policy
Tiebout
Charles M
“A Pure Theory of Local Expenditures”
Journal of Political Economy
 , 
1956
, vol. 
64
 (pg. 
416
-
24
)
Turner
MA
Ross
S
Discrimination in Metropolitan Housing Markets: National Results from Phase II HDS 2000
 , 
2003
Washington, DC
U.S. Department of Housing and Urban Development
 
Contract No. C-OPC-21304
Turner
MA
Ross
S
Galster
G
Yinger
J.
Discrimination in Metropolitan Housing Markets: National Results from Phase I HDS 2000
 , 
2002
Washington, DC
U.S. Department of Housing and Urban Development
 
Contract No. C-OPC-21304
1
The term “Blacks” is used throughout to refer to people recently (within 500 years) descended in large part from Africa because, with these data, there is no way to distinguish between African Americans and immigrants of African descent.
2
In the sample used below, of large metropolitan areas, Black population growth is positively correlated with desegregation. The correlation coefficient is 0.40 for the dissimilarity index. The corresponding correlations for Latinos and Asians are −0.84 and −0.26. The results follow the same pattern using the isolation index.
3
The majority proceeded to establish a standard for determining racial discrimination: “A racially discriminatory intent, as evidenced by such factors as disproportionate impact, the historical background of the challenged decision, the specific antecedent events, departures from normal procedures, and contemporary statements of the decision makers, must be shown” (Arlington Heights v. Metropolitan Housing Corporation 1977).
4
Consider this thought experiment. In Princeton, NJ, median home values in 2000 were >$314,000 according to U.S. Census data. The median monthly cost for homeowners paying mortgages was $2,211. Meanwhile, the median gross rent was $846. This means that so long as density (measured as units per acre) was 2.6 times higher, rental units would yield higher profits than owner-occupied homes. Considering that affluent single-family homes are often on at least one acre of land, anything zoned multifamily (at least eight units per acre) would be over three times more profitable. Developing an apartment complex with thirty units per acre that charged the median rent could be more than eleven times more profitable than developing the median affluent home. In this model, however, these choices are only possible when R = 0, and we know that in many regions of the country, such as in New Jersey and Massachusetts, R = 1.
5
The only methodological difference is that permitted density zoning received the most restrictive score on the survey if the jurisdictions had a limit of less than eight units per acre. In 2003, the survey distinguished between less than eight and less than four. When combining the surveys, the 2003 scores were rescaled to make them comparable. The response rate to the 1994 survey was 77% (Pendall, 2000).
6
Given the problems mentioned above with aggregating the density restrictions index, and given its blunt focus on whether or not an area uses a minimum lot size requirement above one unit per acre, there is likely to be a great deal of measurement error. This will tend to attenuate the coefficients. Instrumenting with an alternative measure of density zoning (i.e., the Pendall measure) can correct measurement error by filtering out some of the noise. This should increase the size of the coefficients on the DRI. This is exactly what happens. The coefficients increase by 200% to nearly 300%. The effect of a one standard deviation increase in the DRI becomes associated with a dissimilarity score that is seven to nine points higher. These results are available upon request.
7
In the model with reported zoning changes, the increases and decreases of ≥10% from 1994 to 2003 are taken as exogenous, while the 2003 level is instrumented. The idea is to predict the prevailing zoning regime between 1990 and 2000 using the instrument, rather than trying to predict the exact changes, which are explored below more thoroughly. In fact, efforts to predict the reported changes proved futile, implying that they were at least somewhat random.
8
The standardization index is used to make the results more comparable and to ease interpretation. The results for Blacks and Asian are unchanged when 2003 is rescaled to match 1994 and the difference is used. However, for Latinos, the coefficient on zoning becomes marginally insignificant because the standard error increases.
9
In fact, the standard deviation of density zoning within regions ranges from 32% to 75% of the standard deviation of density zoning across the United States. The south had the most intra-regional variation, and using the south as a dummy variable makes very little difference to the results.