Learn to build a world where everyone belongs. Take free classes at OBI University.   Start Now

Zoning parcel map of the bay area

As part of our series on racial segregation in the San Francisco Bay Area, we examined the relationship between restrictive zoning (and single-family zoning in particular) and racial residential segregation. That investigation led us to a broader understanding of the effects of single-family zoning, which we now document. This report describes the characteristics of communities in the San Francisco Bay Area in relation to the degree of restrictive zoning within each jurisdiction. The report also contributes towards zoning reform by identifying high single-family zoned cities in the Bay area that could potentially be upzoned.

We find that cities with high levels of single-family zoning have greater resources (even relative to the generally wealthy and expensive Bay Area) in virtually every statistic we are able to measure. These cities have higher incomes, higher home values, better-performing schools, and our evidence indicates they are high opportunity in the broadest sense: children who were raised in these cities 30 years ago have better outcomes in their adulthoods. However, this is also consistent with a troubling pattern of social, economic, and racial exclusion in cities with high levels of single-family zoning.

The origins of zoning is colored by classist and even racist history.1 Although no longer racially explicit, exclusionary zoning such as single-family zoning is explicitly classist, designed to exclude lower-income residents and more affordable housing options, and can be implicitly racist, designed to keep out certain groups of people based upon racist stereotypes. These exclusionary roots are evident in the composition of heavily single-family zoned cities today, and our findings are not trivial: home values in these cities are $100,000 higher, incomes are $34,000 higher, and these areas are nearly 20 percent whiter than the rest of the Bay Area’s cities. Furthermore, these areas don’t simply provide their own residents with these resources; they also hoard them from the rest of the Bay Area, especially people of color and people with low incomes. As we detail in this report, people who are excluded from these neighborhoods have fewer well-performing schools nearby, have lower incomes, and have less access to opportunity.

To conduct this research, we created original color-coded municipal maps displaying single-family zoning, other residential zoning, and non-residential zoning for 101 municipalities in the Bay Area across nine counties. Unlike zoning maps used by many other researchers, our maps were constructed from the parcel level upward and accounts for residential land use not publicly attributed to specific housing types, allowing us to assess the effects (or at minimum the correlates) of restrictive zoning with a far greater level of precision than has generally been done by others. Our goal is to convey a broader appreciation of the ramifications of restrictive zoning.

In addition to presenting our findings, we now make the underlying parcel-level data we collected to create the original bay area zoning maps available to other researchers to conduct their own analyses. The Othering & Belonging GitHub page hosts the parcel-by-parcel geographic and data files used in this report and its analysis. We hope that this unprecedented repository of detailed zoning information is used by researchers in future projects and are excited to see what other insights might be gleaned from the data we’ve produced.

We undertook this labor-intensive research in the hopes that the full significance of our efforts would lay not just in our own findings, but also in how other researchers use our data. This report merely reflects the first attempt to utilize this original dataset to draw inferences and conclusions about the characteristics of communities with restrictive zoning.

Difficulties with Zoning Research and Limitations of Existing Zoning Datasets

In Part 5 of our series on racial segregation in the Bay Area, we summarized prior research, much of it conducted in recent years, investigating the connections between restrictive zoning and racial residential segregation, jurisdictional fragmentation, and housing prices.2 Unfortunately, this research has been hampered by a lack of available datasets, especially precise zoning data.

The main problem impeding the development of this research is that, to date, there is no single comprehensive database or repository of zoning in the United States, nor is there a universal standard for zoning classifications or base zoning formulae from which they may be derived.3 While there may be similarities in zoning designations or taxonomies across jurisdictions with land-use authority, the particulars are often idiosyncratic to the jurisdiction, which often use their own designations. Furthermore, there is no national or even state-based reliable database or index of zoning in the United States to compare zoning codes. This makes systemic research on zoning time-intensive, laborious, and difficult, especially at a larger scale. But this is not for lack of effort.4

The most widely used data source for zoning is probably The Wharton Residential Land Use Regulatory Index (WRLURI), a survey tool that collected land use data from across the country.5 This Index has many limitations. Firstly, it only covers 2,600 jurisdictions out of possibly more than 30,000 local governments with zoning authority.6 Secondly, it has not been updated in over a decade. Thirdly, it is a survey, with all of the gaps, inaccuracies, and human errors attendant from municipal staff responding to a survey instrument. A similar survey response was gathered by the Urban Institute (in coordination with Fannie Mae), called the “2019 National Longitudinal Land Use Survey.”7 This survey instrument suffers many of the same problems, including that it only includes 1,700 jurisdictions in the 50 largest metropolitan statistical areas. This effort appears to be modeled after a similar survey (of 1,800 jurisdictions in the 50 largest metros) conducted by the Brookings Institution in 2003, a decade and half earlier.8

Another useful tool is the dataset collected in response to the Terner Center for Housing Innovation’s “Residential Land Use Survey,” which surveyed several hundred municipal jurisdictions in California.9 This repository is more recent, but suffers from many of the same problems of imprecision and limited scope. One of the studies published by the Terner Center itself found significant discrepancies and inconsistencies between survey responses and case study data that observed the zoning and land use directly.10 For example, the case study data showed developments were generally approved more slowly than was suggested by the survey responses. Similarly, the case studies suggested that survey respondents underestimated approval timelines, public opposition to developments, and the degree of restrictive zoning in their jurisdiction.11

Most critically for our purposes, however, the Terner survey asked local jurisdictions to respond within broad categorical answers rather than precise answers to the survey questions. For example, the survey asks cities “how much land is zoned to allow single-family detached housing?,” and provides respondents options for “Almost None (0-6%),” “Little (6-25%),” “Some (26-50%),” “A Lot (51-75%),” “Most (76-95%),” or “Almost All (96-100%).” These categorical answers are necessarily less precise than directly measuring the degree of single-family zoning in each jurisdiction.

Virtually all of the previously referenced studies examining the effects of restrictive zoning rely upon these imperfect datasets and survey instruments.12 Researchers that have attempted to directly observe municipal zoning encounter byzantine ordinances. Zoning codes are sometimes long, technical, and difficult to access. Moreover, actual codes do not necessarily indicate actual zoning, especially for planned developments. This is why land use tech tools are unable to provide this information systematically either. Rather, what they are doing is mapping a handful or limited number of cities, or attempting clever work-arounds, with limited results so far.13

Our hope is that the dataset we created for this research and now make available will greatly accelerate and facilitate the development of further research on this important subject, especially for researchers examining the Bay Area. With support, we hope to expand this research and our dataset statewide, and perhaps beyond.

Our Methodology

In order to study the characteristics of communities by level of restrictive zoning, we first needed to assess the types of zoning by jurisdiction and their relative extent. We analyzed the current municipal zoning across the nine Bay Area counties (Alameda, Contra Costa, Marin, Napa, San Francisco, San Mateo, Santa Clara, Solano, and Sonoma)—a total of 100 municipalities. Because we were specifically interested in the effects or correlates of restrictive zoning, and most of the zoning codes were idiosyncratic across municipalities, we simplified zoning designations into three broad categories: Single Family Residential, Multi-family Residential, and Non-Residential.

  • Single Family Residential is land designated for detached, single-family residential land use (one or two dwelling unit per parcel of land) in both low and high density. This includes single-family homes and two-family detached dwellings.
  • Multi-family Residential is land designated to allow for multiple dwelling units per parcel of land. This includes apartment buildings, duplexes, townhouses, mobile home parks, and two-family attached dwellings.
  • Non-Residential is land that is designated for non-residential uses. This includes parks and open space, commercial, industrial, commercial mixed-use, and public land. Non-developable land was removed from the overall land area, including water, waterways, marshland, and highways or streets.

Note:  The built environment in all categories may not perfectly match the zoning designations due to discrepancies between zoning and the built environment, changes in land use since analysis was conducted, and other anomalies. See our Appendix for more details.

Following this taxonomy, we categorized land parcels using local general plan land use map designations, based on definitions published in municipal general plans and zoning codes. We primarily relied upon municipal general plan land use shapefiles and land use designations as our primary source for categorizing zones. General plan land use and zoning map shapefiles were provided by the Association of Bay Area Governments and directly from municipal planning departments, or downloaded from ESRI’s ArcGIS HUB.

In instances where general plan land use and municipal zoning maps conflicted, municipal zoning geographic boundaries and designations were used. When land use and zoning definitions for land parcels did not delineate between single- and multi-family zones, such as in mixed-use residential, planned development, and specific plan zoned land areas, we defined the type of residential land by visually examining the current built environment through satellite imagery. Land use and zoning maps were designated and modified using ArcGIS and QGIS, land use category data was processed in R, and satellite imagery was accessed through Google Earth.

The zoning and land use information used reflects the most recent ratified zoning that was publicly available or provided by municipal planning departments as of March 2020. As the unit of analysis was municipal regulation and land use policy, the categorized maps in this report reflect policy at a single moment in time; they do not reflect the exact built environment if that differs from municipal policy, and they cannot capture the changes in zoning since the research was conducted. Further description of the methodology for categorization can be found in the Appendix at the bottom of this page.

In Part 5 of our racial segregation series, we published static maps for 101 jurisdictions based upon this methodology. We are now pleased to present our interactive mapping tool that contains the same information, but which allows users to more readily and closely examine zoning designations in the Bay Area. To the best of our knowledge, this is the first effort of this kind to systematically map zoning for Bay Area jurisdictions. Contributing an unprecedented level of detail, we determined whether each parcel in every jurisdiction in the Bay Area is single-family zoned.

Figure 1: Interactive map

Click for a full-sized version of this map.

Extent of Single-Family Zoning in the Bay Area

In our initial analysis, which only examined jurisdictions with more than 10,000 people, we found that 82 percent of residential land was single-family-only zoned. When extending our analysis to the entire region, we now find that 85 percent of residential land in Bay Area municipalities is exclusively reserved for single-family housing. The average amount of residential land in each city is 81 percent, with a median of 86 percent, which means that most land is available to residential development. The average amount of total land (including commercial areas and parks) exclusively reserved for single-family housing was 47 percent. Without a systemic analysis of jurisdictions, this calculation would not be possible.

Moreover, only two of the 101 cities in the Bay Area dedicated less than 40 percent of their land to single-family zoning.14 Single-family zoning predominates our residential areas. Denser housing options are only permitted in less than one-fifth of residential areas in this region. Table 1 below categorizes each city into categories of low, medium, and high levels of single-family zoning, which we will use for the rest of the report.

Table 1: Cities by Single-Family Zoning Percentage

Low (0-80%) Medium (80-90%) High (90-100%)
Alameda American Canyon Antioch
Albany Belmont Atherton
Benicia Brisbane Belvedere
Berkeley Burlingame Brentwood
Campbell Colma Calistoga
Cloverdale Concord Clayton
Cotati Corte Madera Cupertino
Dixon Daly City Danville
Dublin East Palo Alto Fairfax
Emeryville El Cerrito Hillsborough
Fairfield Fremont Lafayette
Foster City Gilroy Livermore
Hayward Half Moon Bay Los Altos
Hercules Healdsburg Los Altos Hills
Larkspur Menlo Park Los Gatos
Milpitas Millbrae Martinez
Mountain View Morgan Mill Valley
Oakland Napa Monte Sereno
Petaluma Newark Moraga
Pittsburg Palo Alto Novato
Redwood City Pleasant Hill Oakley
Richmond Rio Vista Orinda
Rohnert Park San Bruno Pacifica
San Francisco San Carlos Piedmont
San Mateo San Jose Pinole
San Pablo San Leandro Pleasanton
Santa Clara San Rafael Portola Valley
Sausalito San Ramon Ross
Sebastopol Santa Rosa San Anselmo
Sonoma St. Helena Saratoga
South San Francisco Tiburon Woodside
Suisun City Union City  
Sunnyvale Vacaville  
Yountville Vallejo  
  Walnut Creek  

Our initial findings on the relationship between single-family zoning and racial demographics and racial residential segregation were reported in Part 5 of our series on the latter. To situate this research in its original context, we briefly recapitulate those findings.

Single-Family Zoning and Race

The following table (Table 2) displays municipal differences in levels of racial segregation and racial composition based upon the degree of restrictiveness in land use, as measured by the percentage of single-family-only zoning as a percentage of all residential land by jurisdiction. The table also indicates the number of cities out of the 101 we examined that fall within each range.

Table 2: City Composition by Single-Family Zoning Percentage

  Low (0-80%) Medium (80-90%) High (90-100%) All cities with measured zoning Bay Area Total
Within-City Segregation 0.13 0.12 0.04 0.13 NA
Divergence from Bay Area 0.12 0.14 0.21 0.13 NA
% White 36% 35% 55% 38% 40%
% Black 8% 4% 5% 6% 6%
% Hispanic 22% 27% 16% 27% 24%
% Asian 27% 29% 19% 27% 26%
All Other Groups 6% 5% 5% 5% 5%
Population 3,097,325 2,944,473 902,240 6,944,038 7,675,798
Number 34 36 31 101 NA

The Divergence Index, our preferred measure of segregation (for reasons covered in Part 3 of our series), compares the relative proportions of racial groups in one geography to another. The higher the index value, the higher the level of segregation, and vice versa. A value of zero indicates that the demographics of the locality match those of the larger geographic area, or perfect proportionality of group representation.

To show how segregation differs in cities with high or low levels of single-family zoning, we decompose the Divergence Index. This reveals each city’s level of internal segregation (“Within-City Segregation” in Table 2) and its divergence from the Bay Area’s racial demographics (“Divergence from Bay Area”).15

As a point of reference, we designated as “highly segregated” any census tract with a Divergence Index value greater than 0.215 and as “integrated” any census tract with a Divergence Index value below 0.1075.16 Logically, a tract with a value falling between those numbers was designated as “moderately” segregated. But actual scores varied enormously.

With the launch of our interactive segregation mapping tool, users are able to directly observe census tract Divergence Index values. If you open the mapping tool, simply hover your cursor over the tract you wish to examine, and the Divergence Index value will be visible. As the screenshot presented in Figure 2 below indicates, east Oakland has extremely high Divergence Index values, some of the highest in the entire region.

Figure 2: Divergence Index of Oakland neighborhood census tract


Using our single-family zoning analysis, we found that jurisdictions with the highest Divergence Index values are found in the communities with the highest proportion of single-family zoning, and vice versa. This comports with our expectation that segregation and single-family zoning are related, as the above-mentioned research suggested. The greater proportion of single-family zoning, the higher the observed level of racial residential segregation.

Nonetheless, single-family zoning has a complex relationship with racial residential segregation, as the case of east Oakland suggests.17 Cities with higher levels of single-family zoning are more racially homogenous within their boundaries, meaning that they are less diverse than cities with less single-family zoning. As a result, these high single-family zoning cities often have lower intra-city (or neighborhood-based) segregation than more diverse communities with lower levels of single-family zoning.18 That lower level of intra-jurisdictional segregation comes at the cost of diversity: cities with high levels of single-family zoning show demographics that are extremely different from the Bay Area as a whole. White people are overrepresented in these areas, and all other populations are underrepresented compared to their Bay Area proportions, as the chart below (Figure 3) illustrates.19

Figure 3: Race Composition of Bay Area Cities


At the other end of the spectrum, cities with low single-family zoning are more diverse and representative of the Bay Area as a whole. Though more diverse, these jurisdictions are more likely to suffer from internal segregation (intra-jurisdictional or neighborhood segregation). More diverse cities have greater capacity of intra-municipal segregation. Our data suggest that nonwhite populations are more likely to live in areas with a high percentage of multi-family zoning, which means that differential concentrations of single-family zoning may contribute to both intra- and inter-municipal segregation.

Finally, an obvious correlate of high levels of single-family zoning is that fewer people live in these areas. This structure allows a relatively small number of people to accumulate economic and social wealth that can perpetuate inequality through multiple generations, a subject that is a focus of this report.

Single-family zoning is perceived to be associated with certain neighborhood characteristics: more valuable homes (higher property values), better performing schools, fewer households earning below the federal poverty line, and so on. Broadly speaking, if true, it is easy to understand the motivation behind such zoning designations: local policymakers and the constituents they represent are sculpting their communities in ways that achieve these outcomes. However, if single-family zoning is designed to enforce a particular neighborhood character and increase property values, then such zoning will have an exclusionary effect, by making it harder for lower-income or higher-need residents to reside in those communities. This has the effect of exporting higher-poverty (and thereby higher-need) populations into denser residential zones, often in different jurisdictions. Both the better outcomes and the exclusionary effects are observed in our analysis.

Single-Family Zoning and Community Resources

Figure 4: Graph of ACS Data

In the San Francisco Bay Area, cities with high levels of single-family zoning have median incomes $34,000 higher and home values $100,000 higher than those of cities with low levels of single-family zoning. Similarly, a high percentage of residents in high single-family zoned cities own their homes. As homes are a major asset for many American families, this difference in homeownership may also indicate a significant wealth gap.

A wealth gap exacerbated by restrictive zoning is implied by the income distribution between these cities, as single-family zoning correlates with middle-class income (twice the federal poverty level) and low levels of poverty. This combined with the correlation between home values and single-family zoning suggests that restrictive zoning is a powerful mechanism for maximizing wealth. The patterns of home values and income largely support the hypothesis that single-family zoning correlates with the hoarding of resources.

Single-Family Zoning and Educational Outcomes

Given the relationships between community resources and zoning, observers might assume that highly single-family zoned communities have better performing schools. Our findings comport with this assumption. We find that cities with high levels of single-family zoning show markedly better school performance, but do not show higher levels of adult educational attainment than low single-family zoning neighborhoods. This paradoxical finding may be explained by the so-called “creative class” thesis, that many highly educated people reside in denser, more urban neighborhoods.20

Schools in communities with lower levels of single-family zoning have twice as much of their population using free or reduced-price lunch (52 percent, compared to 26 percent in high single-family zoned cities). The difference in student resources is also reflected in the performance of schools. Students in high single-family zoned cities score nearly 15 percent higher in math and reading assessments in fourth grade. The difference in early childhood education persists throughout their education, as high schools in low single-family zoned cities fail to graduate their students at a rate twice that of high schools in high single-family zoned cities.

Figure 5: Education Graph

Single-Family Zoning and Opportunity

Because single family zoning often follows a long-term pattern of residence, the outcomes of children who grow up in these areas is also relevant to understanding its impact. In addition to looking at neighborhood characteristics, we compared the degree of restrictive residential zoning to opportunity outcomes based upon the Opportunity Atlas.21

The Opportunity Atlas contains information on childhood outcomes for nearly every American child born between 1978 and 1984. Focusing on the Bay Area, we are able to see a clear and consistent correlation between single-family zoning and positive outcomes for children across income distribution.22 Figure 6 below shows the outcomes of children in each Bay Area city, by race and the income percentile of their parents (i.e., whether their parents had an income within the twenty-fifth, fiftieth, or one hundredth percentile. Researchers have used this underlying data to draw many fascinating and important conclusions about economic mobility.23 To the best of our knowledge, however, no one has cross-referenced this data with zoning designations.

We found that children born between 1978 and 1984 in neighborhoods that are high in single-family zoned neighborhoods today had significantly better life outcomes. As adults in 2015, they earned higher incomes, and even children who were raised in the poorest 25 percent of households were 7 percent more likely to live in a high-income household as adults than those who grew up in low single-family zoned neighborhoods. Children in today’s high single-family zoned areas were more likely to move into other low-poverty neighborhoods, suggesting the effects of being raised in these areas persist into adulthood. This is also true for Black children raised in such neighborhoods, who were also more likely to move into low-poverty neighborhoods as adults.

Figure 6: Historical Opportunity

We compared these findings with a different set of opportunity maps developed in the context of low-income housing in California. The California Tax Credit Allocation Committee (TCAC) opportunity mapping project assesses the degree of opportunity provided by each tract in California using a mapping tool we helped develop.24 Using a slightly modified version of that index, we found that, in the Bay Area, “High Resource” and “Highest Resource” tracts (those scoring in the top fourtieth percentile on its opportunity index) are concentrated in single-family zoned neighborhoods.25

Because the TCAC Opportunity Map considers several of the above factors in assigning its score, the concentration of high resource tracts in single-family tracts shows the extent to which these aspects of community resources co-vary together. The opportunity index tells a similar story to the individual indicators listed above: opportunity for long-term success is concentrated in the neighborhoods where the fewest people are allowed to live, a tragic fact of life in the Bay Area.

Figure 7: Current High-Opportunity Areas

bar graph showing TCAC high opportunity areas
Single-Family Zoning and Environmental Conditions

The relationship between environmental conditions and restrictive zoning is less straightforward. High single-family-zoned cities have lower levels of diesel and traffic emissions, which accords with the fact that the communities with the highest levels of single-family zoned residential lands are suburbs or exurban areas. These emissions occur more in urban neighborhoods or inner-ring suburbs. Oakland, for example, has been in a high-profile lawsuit involving a coal terminal, with environmental groups raising concerns that coal dust will pollute the city’s neighborhoods.26

Interestingly, however, more restrictively zoned jurisdictions have higher levels of pollution in drinking water, as well as higher levels of ozone, pesticides, and PM2.5. This, too, accords with the assumption that infrastructure is more costly and perhaps less well developed in more outlying or far-flung areas. San Francisco, for example, has among the best drinking water in the Bay Area, and among the most sophisticated water-treatment facilities.27 Also, rural or exurban areas may have higher levels of fertilizer run-off or other pesticides than urban neighborhoods.

The important point is that environmental conditions are determined and shaped by factors that often go beyond the influence of single-family zoning, and that larger, more urban cities are more likely to have both more developed infrastructure and higher particulate pollution than outer-ring suburbs or exurbs. 

Figure 8: Environmental Health and Hazards

Identifying Cities for Reform

As presented in Part 5 of our racial residential segregation series, zoning reform is one piece of the necessary policy strategy to reach equity goals. Restrictive zoning is a powerful mechanism for hoarding resources, and that will not shift without actively reshaping zoning regulations at the municipal level. In the short term, zoning reform alone cannot accomplish the goal of creating more affordable housing or providing more resources to lower-performing schools. In the long term, however, it will be exceedingly difficult to accomplish any equity objective without it.

Given the degree of restrictive zoning in the Bay Area but variability across the region, one intuitive policy response is to target reform efforts to jurisdictions with the highest levels of single-family zoning. While this has a mathematical appeal, it may be overly simplistic in relation to ultimate equity objectives. For that reason, we have created a simple, multi-factor approach to identify the jurisdictions where restrictive zoning reform may have the most impact, using three indicators:

  1. Percentage of residential land zoned for single-family housing
  2. Opportunity designation using the TCAC opportunity maps
  3. Proximity to regional economic centers (central business districts)

These three indicators reflect important facets of zoning reform proposals. Many areas within the Bay Area may have high single-family zoning, but may still not be ideal locations for new residents. Areas that force people into long commute times, for example, have shown to be detrimental to physical and mental health. To measure this, we use proximity to central business districts as a proxy for access to nearby gainful employment.28

Furthermore, the broad aim of reforming single-family zoning laws is not to punish current residents, but to expand opportunity for would-be residents in those cities. To that end, we also select for cities that have high concentrations of high opportunity areas, as designated in the previously mentioned TCAC index. This ensures that new residents would have a good chance at living healthy, successful lives and passing that success to their children. Finally, we select for high single-family zoned cities. Based upon this analysis, we find the following jurisdictions as the best candidates for zoning reform:

Table 3: Cities Most in Need of Reform

  Single-Family Zoning High Opportunity Miles from CBD
Monte Sereno 100% 100% 9.1
Los Altos Hills 100% 100% 14.4
Orinda 100% 100% 6.9
Piedmont 100% 100% 2.1
Saratoga 99% 100% 8.1
Lafayette 98% 100% 9.8
Los Altos 98% 100% 11.9
Belvedere 94% 100% 7.5
Mill Valley 92% 100% 11.6
Cupertino 91% 100% 7.4
Pacifica 91% 100% 11.6
Moraga 91% 100% 8.3
Los Gatos 91% 100% 7.9

Single-family zoning dominates the Bay Area’s residential neighborhoods, squeezing out much-needed denser housing options. Using a thoroughly crafted dataset, we found that 85 percent of residential land in the Bay Area is restricted to single-family zoning. No jurisdiction larger than 10,000 people in the Bay Area (not even San Francisco or Oakland) has less than 40 percent of its residential land dedicated exclusively to single-family homes.

We find that single-family zoning is strongly correlated with a wide range of predictable neighborhood characteristics, including better schools, higher property values, higher incomes, and lower poverty. In addition, opportunity research indicates that these values aren’t merely static. Instead, the higher levels of resources in these neighborhoods enable their residents to attain better outcomes throughout their lives.

None of this information condemns the single-family home nor the residents of these neighborhoods. The policy solution we advocate for is not to deprive single-family neighborhoods of resources, but to eliminate the barriers that prevent the rest of the Bay Area’s residents from accessing them. The elimination of single-family zoning will help to allow a greater supply of housing in these neighborhoods so that the opportunity they provide will become more broadly and equitably distributed.

Appendix: Categorization Methodology

Data Sources and Tools

General plan land use and zoning map shapefiles were provided by the Association of Bay Area Governments and directly from municipal planning departments, or downloaded from ESRI’s ArcGIS HUB. General plan land use designations were the primary source for categorizing residential zones; when municipal zoning maps differed from general plan land use maps, zoning took precedence. Land use and zoning maps were designated and modified using ArcGIS and QGIS, land use category data was processed in R, and satellite imagery was accessed through Google Earth.  

Categorization Method

Land parcels were categorized using general plan land use designations, based on definitions published in municipal general plans. If the general plan was not specific enough to parse into analysis categories or did not match municipal zoning designations, municipal zoning definitions were used. If available general plan land use map shapefiles were out of date or did not match the most recently published general plan or zoning maps, shapefiles were manually reshaped or redesignated to match current land use maps.

The built environment does not always reflect current zoning; there can be differences in how zoning laws are regulated, developments grandfathered in, or variances allowed that could lead to the built environment on-the-ground looking differently than the zoning code. For example, an area zoned for multi-family housing may have a large number of single-family homes, suggesting that there might be some discrepancy between zoning designation and what is built. Identifying these issues was beyond the scope of this research. This analysis followed the municipal zoning codes to categorize, thus focusing on the codes rather than “as built.” However, some types of zoning categories prove difficult to separate and easily categorize into a basic analytical framework. These mixed-use or planned development categories do not necessarily distinguish between non-residential and residential. When land use and zoning definitions did not delineate between single- and multi-family, or there was some discrepancy, researchers referred to the current built environment to determine the type of residential land use by visually examining satellite imagery. In categorizing ambiguous zones, we referred to the built environment only when there was no other way to distinguish between categories; we prioritized defined zoning, defined land use, published specific or area-specific plans, and dominant land use based on manual sorting (in that order).

Several common land use designations required this type of manual sorting: 

Mixed Use: Mixed use land use and zoning is typically used to enable a mix of residential densities with commercial and public land areas. Mixed use areas defined by their dominant use (i.e. “Mixed Use Commercial”) were categorized accordingly. General mixed use areas that did not designate the type were categorized based on the current built environment, by examining satellite imagery and sorting land areas accordingly. If residential areas could not be parsed out, mixed-use categories were defined by their dominant residential category.

Specific Plans: Specific plan areas are typically open land areas where municipal planners anticipate or intend a high amount of development around a specific vision or purpose, and therefore have a separate planning process and land use map. If a shapefile for the specific plan areas was not available, the residential zoning categories were manually parsed to match the land use in published specific plan documents. If the development depicted in the specific plan map was not complete based on satellite imagery, categorization followed the current built environment and only incorporated built environment that currently exists.

Planned Development: Planned development is typically a designation that allows for more flexible or dynamic zoning and land use, where planners anticipate or intend future development. If the intended use for each parcel was not defined in municipal zoning code, then planned development land area was categorized manually.    

Low Number of Parcels (<500): For each parcel of land, satellite imagery was examined, and land use was categorized manually. If a shapefile parcel included multiple residential densities, it was parsed to delineate the two.

High Number of Parcels (>500): When the number of planned development parcels was too high to manually parse, researchers applied an automated heuristic based on building footprint shapefiles: OSM building footprint (2017) and Microsoft building footprint (2020). Researchers used Santa Clara as a baseline to develop an average building footprint measure for single-family and multi-family residential land use. First, they manually categorized planned development parcels using satellite imagery, and then calculated the average area of buildings for each parcel in each category. The reference average building footprint was 2,500 sq ft for single-family residential and 8,000 sq ft for multi-family residential. Researchers then applied that reference average building footprint to planned development parcels in other cities’ planned development areas by calculating each parcel’s average building square footage in the Microsoft building footprint shapefile. Those were sorted in ascending order and parcels with building footprint averages between 2,500 and 8,000 were manually examined to determine the specific cut-off between the two categories. Additionally, researchers randomly selected 15 parcels in each category to manually check the efficacy of the building footprint heuristic. While the average square foot of each category differs for each city, the baseline reference category of average building footprints reduces the number of parcels that require the time-intensive manual categorization using satellite imagery.

Use Permits: Some zoning descriptions included caveats that alternative types of housing can be built with a special request to the city (a “special use permit”). Because the acceptance criteria, permitting review processes and relative acceptance rates were not necessarily published for each city, we did not have a standardized way of assessing how special use permits would enable different types of housing development. In zones that had special use permits, we defined the category based on its primary use, disregarding special use permits.

Level of Uncertainty

As discussed above, analyzing zoning and land use regulations is not an exact science; applying a useful analytical framework to different structures of regulation requires some manual sorting based on observation and therefore introduces some uncertainty. Readers should expect a small amount of error in the report maps, and this may increase over time if zoning maps are changed. For the vast majority of municipal zoning, these ambiguous zones make a small portion of the land area.

  • 1Michael Manville, Paavo Monkkonen & Michael Lens, “It’s Time to End Single-Family Zoning,” Journal of the American Planning Association, 86, no. 1 (2020): 106-112, accessed October 1, 2020, DOI: https://www.tandfonline.com/doi/full/10.1080/01944363.2019.1651216. We note that this entire issue of JAMA is dedicated to debating the problem of single-family zoning.
  • 2Stephen Menendian, Samir Gambhir, and Arthur Gailes, Racial Segregation in the San Francisco Bay Area, Part 5, (Berkeley, CA: Othering & Belonging Institute, 2020), accessed October 2, 2020, https://belonging.berkeley.edu/racial-segregation-san-francisco-bay-area-part-5. See text associated with endnotes # 4-8.
  • 3Benjamin Schneider, “CityLab University: Zoning Codes,” CityLab, August 6, 2019, accessed October 1, 2020, https://www.bloomberg.com/news/articles/2019-08-06/how-to-understand-municipal-zoning-codes.
  • 4For a review of the earliest known efforts, mostly surveys devised and conducted in the 1970s and 80s, see Raven Saks. 2006. “Job Creation and Housing Construction: Constraints on Metropolitan Area Employment Growth”, Federal Reserve Board of Governors Working Paper 2005-49 p. 36-37 (describing 6 broad-scale survey efforts: 1) the Wharton Urban Decentralization Project (conducted in the late 1980s), 2) The Regional Council of Governments (conducted from 1975-78), 3) The International City Management Association (conducted in 1984), 4) Fiscal Austerity and Urban Innovation (conducted in 1983-84), 5) the National Register of Historic Places, and 6) American Institute of Planners Study (from 1976)). available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=873866
  • 5Joseph Gyourko, Albert Saiz and Anita A. Summers, “A New Measure of the Local Regulatory Environment for Housing Markets: The Wharton Residential Land Use Regulatory Index,” accessed October 1, 2020, http://realestate.wharton.upenn.edu/working-papers/a-new-measure-of-the-local-regulatory-environment-for-housing-markets-the-wharton-residential-land-use-regulatory-index/.
  • 62017 Census of Governments, United States Census Bureau, accessed October 1, 2020, https://www.census.gov/data/tables/2017/econ/gus/2017-governments.html.
  • 7Lydia Lo, Megan Gallagher, Rolf Pendall, Ananya Hariharan, and Christopher Davis, National Longitudinal Land Use Survey, (Washington, DC: The Urban Institute, 2019), accessed October 1, 2020, https://datacatalog.urban.org/dataset/national-longitudinal-land-use-survey-nllus. (Data developed at the Urban Institute, and made available under the ODC-BY 1.0 Attribution License.)
  • 8Rolf Pendall, Robert Puentes, and Jonathan Martin. 2006. From Traditional to Reformed: A Review of the Land Use Regulations in the Nation’s 50 Largest Metropolitan Areas. Washington DC: The Brookings Institution. (Note: The original survey instrument is no longer available on the Brookings Institution website, but we found it on the internet archive: https://web.archive.org/web/20060902030712/http://www.brookings.edu/metro/pubs/20060810_Survey.pdf)
  • 9Sarah Mawhorter and Carolina Reid, Local Housing Policies Across California, (Berkeley, CA: The Terner Center, 2018), accessed October 1, 2020, http://californialanduse.org/download/Terner_California_Residential_Land_Use_Survey_Report.pdf.
  • 10Moira O’Neill, Giulia Gualco-Nelson, and Eric Biber, “Comparing Perceptions and Practice: Why Better Land Use Data is Critical to Ground Truth Legal Reform,” The Terner Center for Housing Innovation, June 28, 2019, accessed October 1, 2020, https://ternercenter.berkeley.edu/blog/comparing-perceptions-land-use-data.
  • 11Ibid.
  • 12Stephen Menendian, Samir Gambhir, and Arthur Gailes, Racial Segregation in the San Francisco Bay Area, Part 5, (Berkeley, CA: Othering & Belonging Institute, 2020), accessed October 2, 2020, https://belonging.berkeley.edu/racial-segregation-san-francisco-bay-area-part-5. (Jonathan Rothwell’s study used the Terner Center survey, as did Jenny Schuetz and Cecile Murray’s. Jessica Trounstine study, as well as Michael Lens and Paavo Monkkonen’s, relied upon the WRLUI.)
  • 13Emma Nechamkin and Garaham MacDonald, Predicting Zoned Density Using Property Records, (Washington DC: Urban Institute, 2019), https://www.urban.org/sites/default/files/publication/99629/predicting_zoned_density_using_property_records_1.pdf. (This paper proposes a model for inferring zoning from property assessment records as another possible shortcut to the more labor-intensive approach we employ).
  • 14Suisun City (0%) and Benicia (11%) are the two exceptions.
  • 15Tracts are assigned to the city in which they have at least 50% of their landmass.
  • 16Within-city segregation and divergence from the Bay Area are both calculated using the Divergence Index, developed by Elizabeth Roberto (2016). Within-city is the weighted average of the Divergence Index for each census tract that has at least 75% of its land area within a city’s boundaries. Divergence from the Bay Area is calculated as the divergence score for each city (using the city’s demographics) compared to the 9-county Bay Area. The index includes white, Black, Hispanic/Latinx, Asian, and all other groups, totaling the full population of each tract or city.
  • 17Within-city segregation and divergence from the Bay Area are both calculated using the Divergence Index, developed by Elizabeth Roberto (2016). Within-city is the weighted average of the Divergence Index for each census tract that has at least 75% of its land area within a city’s boundaries. Divergence from the Bay Area is calculated as the divergence score for each city (using the city’s demographics) compared to the 9-county Bay Area. The index includes white, Black, Hispanic/Latinx, Asian, and all other groups, totaling the full population of each tract or city.
  • 18All racial and population statistics for the zoning and RHNA analysis are drawn from the 2014-2018 5-year American Community Survey unless otherwise noted.
  • 19Figure 3 is compiled using a LOESS regression on actual results.
  • 20Richard L. Florida, The Rise of the Creative Class: And How It's Transforming Work, Leisure, Community and Everyday Life (New York, NY: Basic Books, 2002).
  • 21“The Opportunity Atlas,” Opportunity Insights, accessed October 2, 2020, https://www.opportunityatlas.org/.
  • 22The Opportunity Atlas provides data by Census tract. To aggregate this data to the city, we calculate the average value of all tracts with at least 50% of their landmass within the city, weighted by the population of children within each race and parent income percentile demographic.
  • 23Raj Chetty, Nathaniel Hendren, Maggie R. Jones, Sonya R. Porter, “Race and Economic Opportunity in the United States: An Intergenerational Perspective,” Opportunity Insights, March 2018, accessed October 2, 2020, https://opportunityinsights.org/paper/race/.
  • 24Fiona Ma, “California Tax Credit Allocation Committee: TCAC/HCD Opportuity Area Maps,” California State Treasurer, accessed October 2, 2020, https://www.treasurer.ca.gov/ctcac/opportunity.asp.
  • 25Because the TCAC index is scored differently in rural and urban areas, it is difficult to accumulate TCAC tract and block group scores in many jurisdictions. To compensate for this issue, we re-ranked each tract within the Bay Area against each other, developing an internally consistent index.
  • 26Kimberley Veklerov, “Oakland sued, again, over terminal project to ship coal,” San Francisco Chronicle, December 4, 2018, accessed October 2, 2020, https://www.sfchronicle.com/bayarea/article/Oakland-sued-again-over-terminal-project-to-13443361.php.
  • 27San Francisco Public Utilities Commission, City of San Francisco 2019 Annual Water Report, (San Francisco: San Francisco Public Utilities Commission, 2019), accessed October 2, 2020, https://sfwater.org/modules/showdocument.aspx?documentid=15360.
  • 28Kyle Fee and Daniel Hartley, "The Relationship between City Center Density and Urban Growth or Decline," Federal Reserve Bank of Cleveland, Working Paper, no. 12-13, (2012), accessed October 2, 2020, https://www.clevelandfed.org/en/newsroom-and-events/publications/working-papers/2012-working-papers/wp-1213-the-relationship-between-city-center-density-and-urban-growth-or-decline.aspx. (The Central Business District (CBD) locations used for this index are from Fee and Hartley’s 2012 analysis. These data points reflect the historic CBD locations originally identified in the 1982 Census of Retail Trade and commonly used as a reference point for business centers. The original 1982 Census identified the most business-dense census tracts within the largest city of each MSA; Fee and Hartley produced single latitude and longitudes to represent those areas by taking the centroid of each census tract cluster, as well as updating to include CBD’s that have emerged since 1982. All of the CBD located in the Bay Area are from the original 1982 Census designations.)