Right Column

Chapter 3: Land and Site Constraints

Land is the most basic of housing inputs. Without adequate supplies of serviced and developable land, homebuilders can't build. Without new construction, home prices and rents will rise, and affordability will fall. Thus the most basic part of the housing equation, and the subject of this chapter, concerns land: does California have enough developable land to accommodate projected growth?

This question has five subparts. The first is a matter of definition; specifically, which types of lands should be considered developable? The second is a matter of location: Is there enough developable land available where the demand is? The third is a matter of reuse: How much of California's projected housing demand might be accommodated by reusing or redeveloping previously developed sites? The fourth is a matter of density: At what density can or should future housing be built? The higher the density, the less land will be needed; the lower the density, the more land will be needed. The fifth is a matter of cost: Is sufficient land available at a low enough price to insure that the land and housing markets are competitive? This chapter takes a careful look at all five of these issues.

Raw Land Capacity

Developers talk about land that is privately owned, lacks urban services, and has not been previously developed as "raw land." Raw land parcels located at the fringe of existing urban areas are commonly referred to as "greenfield" sites, to distinguish them from "infill sites," which are located within developed urban areas. In theory, California should have ample supplies of raw land. As of 1996, the 35 California counties for which detailed land supply data are available—including all of the state's urban counties (see Exhibit 12)—included approximately 3.5 million acres of urbanized land, 32 million acres of public or undevelopable land, and nearly 25 million acres of physically-developable raw land. ¹

To determine how much of this 25 million acres of physically developable land might realistically be considered developable, a geographical information system (GIS) and various digital map layers were used to identify and characterize potential development sites.² Detailed digital maps and data were obtained from many sources, including the U.S. Geological Survey, the California Farmland Mapping and Monitoring Project (FMMP), the State of California Teale Data Center, the National Wetlands Inventory, FEMA, the U.S. Census Bureau, the California Department of Fish and Game's Natural Heritage Program, and the Gap Analysis Project. (Appendix E includes a detailed listing of map and data sources.) The data were assembled for the entire state, "grided" into a common system of one-hectare cells, and then separated by county.³ Because no federal or state agency collects comprehensive data on sites within urban areas, the comparable potential for infill development could not be established.

Land Classification

The resulting maps and digital data sets encompass 60 percent of California's land area, cover 35 of the State's 58 counties (including all urbanized counties except San Francisco, see Map 5), and include information on nearly 57 million acres of land area as divided into one-hectare "sites." Each site was then classified into one or more categories of developability based on the following criteria, and as described in Exhibit 12:

• Already Developed Sites: These were sites identified by the California Farmland Mapping and Monitoring Project (FMMP) as being urbanized in 1996. They include sites developed in commercial, industrial, public, and residential uses. The threshold density used to distinguish urbanized from non-urbanized sites is one residential unit per two acres.
• Undevelopable Sites: This category and the next measure physical developability. This category includes already-developed sites (category #1, above), land under public ownership (such as federal and state-owned lands, public parklands, military bases, and some local government-owned sites), underwater lands, and lands with a slope of 15 percent or more. It does not include privately- or municipally-owned watershed lands.
• Potentially Developable Sites: This category is the converse of the previous one, category #2. It consists of all sites not classified as undevelopable. It includes all currently undeveloped and privately owned sites that are not underwater and have an average slope of 15 percent or less.
• Developable and Accessible Sites: This category includes all potentially developable sites (category #3, above) within 10 kilometers (6.2 miles) of a major roadway (interstate highways, four-lane freeways, and/or major federal or state highways) or within 10 kilometers of existing urban development. These parameters were used to eliminate sites judged too far from existing infrastructure to be economically feasible for development. California developers must typically pay the full costs of extending required public infrastructure (roads, and sewer and water service) to their projects. The more distant a site from existing hookups, the greater the infrastructure extension cost. Thus at some point, far-flung development simply becomes uneconomical; for this analysis that point is set at 10 kilometers.
• Developable & Accessible Sites Excluding Wetlands and Unique Farmlands: This category excludes all sites listed under the National Wetlands Inventory and all sites identified by FMMP as either "prime" or of "unique" importance. Wetland and farmland sites may be regarded as physically developable—wetlands can be drained—but, depending on the location, may be difficult to develop for political or regulatory reasons. Wetlands development typically requires the approval of the Army Corps of Engineers. With respect to farmland development, many California farming and suburban communities utilize large-lot and exclusive agricultural zoning to protect or otherwise discourage the development of agricultural lands.⁴
• Developable and accessible sites, excluding wetlands, prime and unique farmlands, and Q3 floodzones: This category also excludes sites identified as Q3 floodzones by the Federal Emergency Management Agency (1996).⁵
• Developable and accessible sites, excluding wetlands, prime and unique farmlands, Q3 floodzones, and areas identified as significant natural areas: This category also excludes sites identified as significant natural areas by the Natural Heritage Division of the California Department of Fish and Game.
• Developable and accessible sites, excluding wetlands, prime and unique farmlands, Q3 floodzones, and sites classified as highly suitable habitat for eight or more threatened and endangered amphibian, bird, mammal, or reptile species: This category excludes sites identified through the Gap Analysis Project (University of California-Santa Barbara 1998) as habitat for multiple numbers of threatened and endangered species. Appendix F explains the habitat rating system in greater detail.
• Developable and accessible sites, excluding wetlands, prime and unique farmlands, Q3 floodzones, and sites 1 mile or more beyond existing urban development: This category excludes otherwise developable sites (under category #6, above) which are located more than one-mile from existing urban development. It simulates the effect a comprehensively adopted series of 1-mile urban growth boundaries (UGBs) would have on the supply of raw land available for development.⁶ To date, about 20 California jurisdictions have adopted UGBs, and many more are considering doing so. The choice of a 1-mile UGB width is conservative and stringent, and is entirely arbitrary. Most of the UGBs adopted to date in California are far narrower than one mile, and thus include far less potentially developable land.
  
  

Exhibit 13 lists the total amount of land in each of the above categories for 35 counties (detailed colored maps indicating the land classification categories for each county are included at the end of this report).

Among the 35 counties listed in Exhibit 13, the effect of excluding wetlands and prime and unique farmlands (i.e., moving from Category #4 to Category #5) would be to reduce the supply of developable land from 17.3 to 12.8 million acres. Excluding Q3 floodzones (Category #6) would further reduce developable land supplies to 11.6 million acres. If special natural areas identified by the California Department of Fish and Game (Category #7a) were prohibited from being developed, the supply of developable land would fall to 9.9 million acres. Excluding sites with an Endangered Species Index of 40 or more (Category #7b) would reduce developable land supplies to 8.2 million acres.

The effects of particular constraints vary widely by county, as illustrated in Exhibit 14. More than half of San Benito, San Bernardino, and San Luis Obispo counties, for example, are either too inaccessible or too steep to be developed. Wetlands and vernal pools account for 20 percent or more of developable sites in Sutter, Solano, San Joaquin, Sacramento, and Contra Costa counties; rendering such lands undevelopable would reduce land supplies accordingly. Coupling prohibitions on the development of wetlands and prime and unique farmlands would reduce the supply of developable land by 72 percent in Yolo County, 66 percent in San Joaquin County, 64 percent in Sutter County, 52 percent in Solano County, and by more than one-third in Kings, Madera, Merced, Sacramento, and Stanislaus counties. Prohibiting development in Q3 floodzones would further and substantially reduce the potential for new housing construction throughout the Central Valley, especially in Kern, Kings, Madera, Merced, and Sacramento counties. Of the counties listed in Exhibit 14, further prohibiting development on sites that are habitat to significant numbers of threatened and endangered species would render undevelopable more than two-thirds of Amador, El Dorado, Madera, Marin, Mariposa, Monterey, Napa, Nevada, San Joaquin, Santa Barbara, Sutter, Ventura, and Yolo counties. Allowing for all of these prohibitions, more than 7.9 million acres of raw land would still be available for development in the 35 counties listed in Exhibits 13 and 14.

From Land Supply to Housing Capacity

Land is needed to accommodate projected commercial, industrial, and public infrastructure development, as well as for housing. How much land should be reserved for each use? Since the State does not publish long-term employment projections—so the amount of land needed for job growth can not be estimated—and since the mix of urban uses varies widely by jurisdiction, this issue was side stepped altogether by estimating gross housing densities for 1996 for each county. Presented as the first data column in Exhibit 15, a county's gross housing density is simply its total urbanized land area as of 1996 (including all urban non-residential uses), divided by its total number of housing units. Calculated this way, gross housing densities account for the total amount of non-residential land uses —including commercial, industrial, and public land uses—associated with residential land uses. Assuming future residential/non-residential land use mixes do not change, gross housing densities can be used to estimate the total amount of urban land associated with projected residential growth.

Among the 35 counties included in Exhibit 15, the average gross housing density in 1996 was 2.4 housing units per acre. ⁷ The counties with the highest gross housing densities are Los Angeles (4.2 housing units per urbanized acre), Mariposa (4 units per acre), Alameda (3.8 units per acre), San Mateo (3.7 units per acre), and Orange (3.6 units per acre). The counties with the lowest gross urban housing densities are Kings (1.3 units per acre), Madera (1.7 units per acre), Amador (2.1 units per acre), and Merced (2.2 units per acre).

These densities are descriptive, not prescriptive. They measure current residential densities as they currently exist on the ground, not the densities those policy- makers, planners, or local citizens groups might wish to promote in the future.⁸

Next, to estimate each county's ability to accommodate future housing development, we multiplied the amount of developable land available in each county by its gross housing density. Assuming that future development occurs at 1996 gross housing densities, and that there are no additional constraints to development beyond slope and access, the 35 counties listed in Exhibit 15 could accommodate 32 million additional housing units. Prohibiting the development of wetlands and prime and unique farmlands (category #5) would reduce this total to 24 million units. Prohibiting development in floodzones (category #6) would further reduce it to 21 million units.

Further prohibiting development in special natural areas and in areas identified as endangered and threatened species habitat (categories #7a and #7b) would reduce the housing capacity of the 35 counties shown in Exhibit 15 to 18.6 million and 15.6 million housing units, respectively. This is far more than enough capacity to accommodate projected housing demand through 2020 and beyond.

Further prohibiting housing development beyond one-mile of existing urban areas—as the imposition of a series of one-mile urban growth boundaries would do—would reduce the housing capacity of the counties listed in Exhibit 15 to just 5.2 million units (as evaluated at 1996 gross residential densities). Unlike the previous capacity reductions, which are based on the physical capabilities of the landscape to support development, such a reduction in housing capacity is entirely arbitrary. At its core, the decision where and how to locate a UGB is based on local concerns over growth, not on land capacity. The looser a UGB, the greater its capacity; the tighter a UGB, the less its capacity. More than anything else, this brief and admittedly simple analysis illustrates the power of UGBs to significantly affect the ability of the state and its component municipalities to accommodate projected growth. Whatever their pros and cons as planning tools, any designation of UGBs can have substantial and unforeseen impacts.

In developing these estimates, the effects of local general plans, zoning ordinances, and subdivision codes were not considered. This is not to suggest that these documents are unimportant. At their core, however, general plans and zoning designations are normative policy statements, not analytical documents. They set forth the land use patterns and densities desired by the community. They may consider development or "build-out" capacity, but are not permanently limited by it. Moreover, because both general plans and zoning ordinances are frequently amended, their build-out capacities can also change. Some regions, i.e., San Diego, have identified a significant shortfall in planned residential capacity relative to projected housing needs.

Balancing Housing Capacity With Demand

How do these various estimates of housing capacity compare with demand projections presented in Chapter 2? To find out, the housing capacity estimates presented in Exhibit 15were divided by the 1997-2010 and 1997-2020 household growth projections previously presented in Exhibit 7. The results of this operation are presented in Exhibit 16 and in Maps 6, 7, 8, and 9. Four sets of capacity-to-demand ratios were calculated: The first is the percentage of capacity-to-demand for all developable and accessible sites (category #4 lands). The second is the percentage of capacity-to-demand, excluding mapped wetlands, prime and unique farmlands, and Q3 floodzones (category #6 lands). The third also excludes lands suitable to large numbers of endangered species (category #7a). The fourth includes only those category #6 sites within one-mile of existing urban development (category #8 lands). Percentages greater than one hundred indicate that there is sufficient developable land within the county to physically accommodate projected household demand at current gross densities. Ratios less than one indicate that projected household demand will exceed available capacity.

Some amount of raw land over and above the level required to physically accommodate projected growth is essential to preserving a competitive housing market and allow for residential mobility. Where such reserves of land are not adequate, land owners are able to charge more than the fair-market price, causing rapid land and housing price inflation. Orange County's lack of land reserves, for example, is one reason why its land and housing prices are among the state's highest. All else being equal, more reserve land is needed in fast-growing areas than in slow-growing ones. The bolded entries in Exhibit 16indicate the counties and scenarios in which available land reserves would be less than fifty percent. ⁹

• If the only requirement were to be that developable sites be within 10 kilometers of an existing highway or urban area, then every county listed in Exhibit 16 (and every county in California except San Francisco) would have sufficient raw land to accommodate projected household growth through 2010. Los Angeles and Orange counties would start running down their land reserves by 2020. Under this most-liberal definition of developability, the 35 counties included in Exhibit 16 would have eleven times the land capacity needed to accommodate projected household growth through 2010, and 6.6 times the land capacity needed to accommodate projected household growth through 2020.
• Adding additional environmental constraints—such as prohibitions on wetlands, floodzones and prime and unique farmland development—would slightly reduce the state's ability to accommodate projected household growth through 2010. Under these circumstances, the 35 counties included in Exhibit 16 would still have seven times the land capacity required to accommodate projected household growth through 2010, and 4.3 times the capacity needed for 2020. As above, only Orange and Los Angeles counties would run up against land capacity limits.
• Prohibiting development on sites suitable to multiple endangered species would further reduce—although again, not onerously so—the amount of land available to accommodate projected housing development. Allowing for this moderate level of species protection, the 35 counties included in Exhibit 15 would still have 5.3 times the land capacity required to accommodate projected household growth through 2010, and 3.2 times the capacity needed for 2020. Land supplies would begin to fall below recommended reserve levels in Los Angeles and Orange counties by 2010, and in Santa Clara County by 2020. Without an increase in densities, both Los Angeles and Orange counties would run short of developable land sometime between 2010 and 2020.
• The imposition of a series of one-mile UGBs would have a far more drastic effect on the state's housing capacity than any other potential constraint. Were such boundaries adopted, the 35 counties listed in Exhibit 16 would have less than twice the land capacity required to accommodate projected household growth through 2010, and just barely enough capacity to accommodate projected growth through 2020. Four counties would have run out of land for development by 2010—Fresno, Orange, Stanislaus, and Yolo. Five other counties—Kern, Los Angeles, Madera, San Joaquin, and Santa Clara—would not have sufficient land reserves. By 2020, eight counties would have run out of land altogether, and another seven would have less than the recommended reserve level. The constraining effects of UGBs would be most strongly felt in the Central Valley, where many cities are already adjacent to prime agricultural areas.
  
  

In summary, California and its metropolitan regions should have more than enough raw land to physically accommodate projected housing, commercial, and economic development through the year 2020 and beyond—even allowing for aggressive wetlands, farmland, hillside, and floodplain protection. Among individual counties, according to this analysis, Los Angeles and Orange will not have sufficient raw land to accommodate projected 1997-2020 housing demand. The key constraint on the State's development capacity is political, not environmental. As this analysis reveals, the careless adoption of even moderately-limiting urban growth boundaries by high-growth counties in the Bay Area and Central Valley would constrain land supplies below the levels required to meet future housing production needs.

As above, we note that the various estimates of development capacity are not comparable to those implicit in local general plans and/or zoning ordinances. Depending on how and when they were developed and amended, local planning documents may stipulate more, less, or different types of housing development capacity than these estimates indicate.

Density Trends

Thus far, this analysis has assumed that future housing development will occur at current (1996) average densities. This may not be the case. When land is in short supply, landowners and developers have an incentive to use it more intensely. This causes densities to increase. The opposite is also possible: greater development pressures can lead to falling development densities, usually as a result of neighborhood opposition and concern over traffic and environmental impacts.

In fact, as Exhibit 17 shows, population and development densities in many California counties have risen significantly during the last 25 years. 1996 population densities were calculated (as previously) by dividing each county's population, as estimated by the California Department of Finance, by its urbanized land area, as reported by the California Farmland Mapping and Monitoring Project. 1972 densities were calculated by dividing each county's 1972 population by its urbanized land areas as estimated using U.S. Geological Survey digital maps.¹⁰

Between 1972 and 1996, gross population densities increased in 17 of the 19 California counties shown in Exhibit 17; only Kings and Monterey experienced declines in density. Gross population densities increased by 10 percent or more in Amador, Contra Costa, Mariposa, Orange, San Benito, Santa Barbara, Santa Clara, Santa Cruz, Solano, and Sonoma counties, and by 5 to 10 percent in Alameda, Marin, Placer, San Diego, Ventura, and Yolo counties.

The trend toward higher-density development is even more apparent if marginal densities are considered instead of average densities. Average densities measure the density of all urban development, regardless of its age. Marginal densities measure the density of new development. They are calculated by dividing the change in population during a given period by the change in urbanized area. Development densities in many California counties have risen significantly.

Except for Santa Cruz, the counties with the highest 1972-96 marginal densities were either in the Los Angeles or San Francisco Bay Areas. They were Santa Clara (54.8 additional residents per additional acre of urban development), Sonoma (17.8), Orange (17.1), and Alameda (16.4). New development in these counties typically occurred at two or three times the densities of existing development. Marginal densities were also much higher than existing densities in several rural counties, including Amador, Mariposa, San Benito, and Yolo. In Kings, Monterey, and Napa counties, on the other hand, 1972-96 marginal densities were actually less than existing densities.

Without additional research, any identification of the factors underlying these trends would have to be regarded as speculative. Densities are rising in the state's major urban counties such as Santa Clara and Orange, probably because high land costs and the complexity of the entitlements process require developers to use land more intensely. The reasons why development densities are rising in rural counties are probably more varied. As urban growth moves eastward into parts of Placer and Amador counties, for example, new residential development is increasingly taking the form of higher-density, suburban-style subdivisions. Higher densities in Santa Cruz and Marin counties—and to a certain extent, Sonoma and Santa Clara counties, as well—are probably the cumulative result of public policies designed to focus urban growth within existing urban corridors.

Whatever the reasons, the fact that development densities have risen suggests that it should be possible to accommodate some of California's anticipated population and household growth in housing forms which consume less land than has traditionally been the case. To investigate how such a shift might affect total land consumption, the amount of land required to accommodate projected household growth was re-estimated, replacing the average 1996 densities shown in Exhibits 15 and 16 with the higher marginal densities suggested by Exhibit 17:

• For core urban counties such as Alameda, Los Angeles, Orange, and Santa Clara, a marginal density of 5.3 households per acre (the equivalent to 15.8 persons per acre) was applied.
• For predominantly suburban counties (e.g., Riverside, San Bernardino, Solano), a marginal household density of 4.0 households per acre (the equivalent of 12.1 persons per acre) was applied.
• For rural counties, a marginal density of 3.2 households per acre, the equivalent of a population density of 9.7 persons per acre, was applied.
  
  

All three of these "alternative" densities are consistent with single-family housing forms. The results of these simulations are presented in Exhibit 18. Assuming future housing were developed at the higher marginal density levels indicated above, the total land area required to accommodate household growth in the 35 counties listed in Exhibit 18 would be 650,000 acres. This represents a savings of 48 percent over the amount of land required to accommodate projected household growth at 1996 densities. The biggest percentage-wise land savings would be in Kings (146 percent), Madera (140 percent), Contra Costa (107 percent), and Merced (82 percent) counties. The smallest percentage land savings would occur in Nevada (7.6 percent), and Santa Cruz (16.9 percent) counties.

It is, of course, much easier to postulate higher densities than to achieve them. Nonetheless, this analysis suggests that the combination of appropriate local development controls and market pressures has the potential to substantially reduce the amount of land required to accommodate projected household growth and to reduce the development footprint on the California landscape. It should be possible, moreover, to achieve these benefits without requiring substantial changes in the current form of housing development.

A Caveat

The finding that California has more than enough developable land to accommodate future housing needs should not be interpreted as an endorsement of current densities or development patterns. There are many reasons why undeveloped land should be conserved, foremost among them the preservation of irreplaceable natural habitat. Other reasons for encouraging higher densities and less land consumption include promoting the development of more varied and interesting neighborhoods; encouraging pedestrian and bicycle travel and transit use as a way of relieving transportation congestion; and creating greenbelts and urban buffers to separate communities and provide an increased sense of community.

The worthwhile nature of these goals notwithstanding, specific determinations regarding how land is best used are most appropriately made at the local level through comprehensive and careful long-term planning. This analysis demonstrates that, frequent complaints to the contrary, California has more than enough raw land to accommodate projected population and housing growth and still maintain current levels of agricultural production and environmental quality. To the extent that California is running out of land, the shortage is a political one.

Infill Development Capacity

Not all new housing construction need occur on raw land. A significant amount of new housing can, and indeed, many would make the argument, should, occur as "infill" in existing urban cores. Infill projects take three forms: (i) vacant parcel development, consisting of new projects built on previously undeveloped parcels;¹¹ (ii) redevelopment or reuse, consisting of new projects built on previously developed sites; and, (iii) the rehabilitation and upgrading of existing residential buildings.

Outside California, the terms infill and redevelopment are often used interchangeably. Inside California, the term redevelopment refers to the specific actions of local redevelopment agencies, or RDAs.¹² RDAs are public entities whose responsibility it is to eliminate blight and promote economic development within designated redevelopment areas. To promote private development, RDAs may build or subsidize supporting infrastructure (such as roads, sidewalks, utilities, and parking), as well as provide grants and loans to private and non-profit housing developers. RDAs may also use the power of eminent domain to purchase and assemble land for redevelopment. RDAs raise funds through the issuance of tax-anticipation-bonds. Bondholders are repaid from assigned property tax increments.

State law requires RDAs to set aside a minimum of 20 percent of their annual tax-increment financing (TIF) revenues for affordable housing. RDAs may use set-aside funds to build and operate affordable housing, to assist for-profit enterprises and non-profit agencies in developing and operating affordable housing, or to subsidize individual low-income households. Set-aside funds may be used anywhere within the city, unlike TIF revenues in general, which may only be used in designated redevelopment areas.

Infill and Reuse

When planners talk about infill, they typically envision the development of an unused or previously cleared site. In fact, more infill housing development probably occurs through site reuse than through the development of vacant sites. We say probably because no one knows for sure. California cities keep close track of how much housing construction occurs, but not where it occurs, nor the types of sites upon which it occurs.

Nor do California cities commonly keep track of vacant sites. Only a few California cities possess a comprehensive, up-to-date, and usable parcel database. A 1994 survey of developable sites undertaken for the Silicon Valley Manufacturing Group, for example, identified 23,888 acres of potential infill sites within Santa Clara County—more than enough to accommodate projected housing demand. A closer inspection of individual sites, however, revealed that most were at the urban edge and should probably not have been classified as potential infill sites.

Even when vacant infill sites are available, there may be powerful disincentives to developing them as housing. For planning or historical reasons, many vacant sites are zoned for commercial or industrial uses. Once a site has been zoned to a "higher" use, rezoning it for housing can be difficult. Many owners of sites previously associated with toxic uses prefer not to develop them in residential use for fear of downstream legal exposure. California cities routinely assess residential impact fees and require significant subdivision improvements, even in infill areas. Because vacant infill sites are typically smaller than greenfield sites, this has the effect of making them more expensive to develop on a dollar-per-square-foot basis.

Information regarding land re-use and private redevelopment activity is even more difficult to obtain. A tabulation of successive land inventories undertaken by the Association of Bay Area Governments found that outside of San Francisco, roughly 22,300 acres of previously developed land in the Bay Area changed land use between 1985 and 1995 (Exhibit 19). Eighty-five percent of the recorded changes involved a shift to residential use. Indeed, residential recycling dominated all other forms of land reuse in the Bay Area between 1985 and 1995. Residential reuse was especially concentrated in Santa Clara, Alameda, and Contra Costa counties.

Even when it does occur, the factors behind residential redevelopment are far from systematic. In Santa Clara County, for example, most of the sites recycled as housing between 1985 and 1995 were near freeways. In adjacent Alameda County, however, sites near freeways (and BART) were actually less likely to be reused as housing. In nearby Contra Costa County, the sites most likely to have been recycled as housing were near major commercial uses. In summary, the few empirical studies which have been done indicate that land reuse and recycling remains a highly localized and idiosyncratic process.

A Statistical Model of Residential Infill Activity

Market conditions, regulatory policy, and subsidy availability influence the construction of infill housing far more than land availability. A strong demand for in-town housing will lead the owners of vacant parcels to sell or develop them, the owners of some non-residential properties to convert them to residential use, and the owners of some existing residential properties to upgrade them. Conversely, regulatory policies which make it too difficult or too risky to develop infill housing will stifle production, regardless of market conditions. The production of affordable housing anywhereùnot just in infill areas—is most closely tied to the availability of subsidy dollars.

Even when all these factors are in place, infill housing development is still idiosyncratic. In an attempt to decode some of the idiosyncrasies, regression analysis was used to compare infill housing production between 1980 and 1990 with various economic and housing market conditions. The analysis was undertaken at the census tract level for major urban counties. Census tract characteristics were identified for 1980 and 1990 and, where possible, compared.¹³ Regrettably, housing construction data at the census tract level is not available after 1990.

Two sets of regression models were estimated. The first was a model of the change in multi-family units in "infill" census tracts, which were identified as those having a 1980 housing unit density greater than 1,000 housing units per square mile. The second was a model of the change in the number of single-family homes in all census tracts, regardless of density. These two sets of dependent variables were compared with seven independent, or explanatory, variables:

• The initial (1980) number of single-family or multi-family units in the tract [InitUnit]: This variable was included as a scaling variable, to reflect the greater potential for multi-family construction in multi-family tracts, and for single-family construction in single-family tracts.
• Median household income in 1980[MedInc80]: We hypothesized that the willingness-to-pay for housing, and thus demand, would be greater in higher-income tracts than in lower-income tracts.
• Percent of households below the poverty line in 1980 [BlwPov]: Similarly, we wanted to test the hypothesis that, all else being equal, proportionately less housing construction would occur in lower-income tracts.
• Median gross rent in 1980[MedRnt80]: We hypothesized that builders would be more inclined to build in tracts where tenants were willing to pay higher rents.
• Percent multi-family in 1980 [PctMulti]: We hypothesized that multi-family construction would be more likely to occur in tracts in which the housing stock was predominantly multi-family.
• Percent of housing units 30 years old or older, in 1980 [PctOld]: We hypothesized that a greater level of new construction would be needed to replace housing units in tracts with large proportions of old or obsolete housing.
• Residential density in1980 [ResDen80]: We hypothesized that it might be more difficult to develop housing in higher-density tracts than in lower-density ones.
  
  

To account for local political and regulatory variations, different models were estimated for different counties. Each model was run twice: first, with all variables entered; second, using step-wise procedures to enter only those variables determined to be statistically significant. Overall, the model results were disappointing (see Appendix G for a detailed listing of the model results). The models "fit the data" poorly, and the coefficient signs and magnitudes were inconsistent:

• Multifamily Models: Depending on the county, the various independent variables "explain" between 11 percent and 54 percent of 1980-90 multi-family unit change in census tracts with densities above 1,000 units per square mile. The best-fitting model is for Contra Costa, the worst is for San Diego. The only independent variable which consistently entered more than half the county models was the percentage of the housing stock more than 30 years old [PctOld]. Except in Orange County, the coefficient of this variable was negative, indicating that, all else being equal, tracts with greater proportions of older units attracted less, not more, multifamily construction. As expected, the multifamily share variable [PctMulti] entered the model positivelyùsuggesting that multifamily construction is more likely to occur in predominantly multifamily tractsùbut only for Los Angeles, Orange, and San Mateo counties. The residential density variable [ResDen] also entered the model in three cases, but with a negative coefficient. In Contra Costa, Los Angeles, and Santa Clara counties, at least, multi-family construction levels were found to be higher in lower-density tracts. Median household income [MedInc80]entered only the Santa Clara model, and it did so negatively, an unexpected result. As expected, median rent [MedRent80]entered the San Mateo and Santa Clara models positively, but it entered the San Francisco model with a negative sign. The proportion of white residents had no effect on multifamily unit change, while the proportion of the population below the poverty line [BlwPov] had a negative effect, as expected, but only in Los Angeles.
• Single-family Models: The single-family models were estimated across all census tracts in a county, not just those identified as "infill" tracts. As above, the model fits varied widely: the various independent variables explained between 9 and 51 percent of the 1980-90 change in single-family homes, measured at the census tract level. Two independent variables entered the various models in a consistent fashion: density [ResDen]and median rent [MedRnt80]. The higher the density, the lower the number of additional single-family homes. Given the preference of homebuilders for low-density, fringe areas, this result is not surprising. Nor is the result that homebuilders were drawn to census tracts with consistently higher apartment rents. Initial unit mix affected production only in Contra Costa and San Francisco counties, where single-family builders were less likely to build in predominantly multi-family tracts. The importance and directionality of other effects varied by county. Homebuilders were drawn to higher-income tracts in Alameda, Los Angeles, and Solano counties, but lower-income tracts in San Francisco. Single-family production levels were higher in Los Angeles County tracts with larger poverty populations, but lower in Orange County tracts with large poverty populations. In San Mateo County, the share of the tract population that was white lowered single-family construction levels; although it increased them in Fresno and Riverside counties.
  
  

More than anything else, these results indicate exactly how hard it is to predict when and why infill housing development will occur. What was true during the 1980s is equally true today: infill development remains an uneven and idiosyncratic process, whether measured at the city, tract, or project level.

RDA Housing Activity

Most of the infill housing built or rehabilitated in California since 1980 has occurred with assistance of local redevelopment agencies (RDAs). As noted previously, RDAs are required by law to use at least 20 percent of their annual tax increment revenues to subsidize the construction and rehabilitation of affordable housing. RDAs may also subsidize low-income renters through vouchers. RDAs may undertake land assembly, provide infrastructure, write-down land prices, pay for construction, and provide grants and financing (including pre-development) to non-profits. RDA-funded projects, including housing, are developed on both vacant and reuse sites. (A 1998 study by the Public Policy Institute of California found that 19 percent of 134 sample projects undertaken by California RDAs were located on sites that had previously been fully or predominantly vacant.) To better leverage their housing funds—and because they typically view themselves as being in the economic development business, not the housing business—most RDAs work indirectly, by providing "gap" grants and financing to housing developers. Depending on the deal and location, RDAs are usually the first or second largest source of gap funds for affordable housing.

The California Department of Housing and Community Development collects annual statistics on the number of affordable housing units built and rehabilitated with RDA funds. A summary of those statistics, organized by county and spanning the 1988-96 period is presented as Exhibit 20.¹⁴ (Appendix H reports this information by year and agency). From 1988 to 1996, as many as 360 redevelopment agencies were active in 38 of California's 58 counties.

From 1988 through 1996, California RDAs assisted or financed the construction of 42,118 new housing units and the rehabilitation of 32,1360 existing housing units. The Los Angeles Redevelopment Agency assisted the most new units during this period (4,574), followed by San Francisco (4,387), the City of Santa Clara (4,290), San Jose (3,596), Contra Costa County (1,983), and the City of Paramount (1,289). The Los Angeles Redevelopment Agency also led in terms of rehabilitated units (5,973 units rehabilitated between 1988 and 1996), followed by the cities of Santa Clara (1,752), Fontana (1,383), Cathedral City (1,329), and San Jose (1,236). While some RDAs, notably those in Los Angeles and San Jose, try to balance new construction with rehabilitation, the majority appear to favor one approach over the other.

As is the case with assisted housing production in general, the contribution of RDA-assisted housing production to total housing construction varies widely by county. In San Francisco, for example, nearly half of all housing units constructed between 1988 and 1996 were either built or sponsored by the city's redevelopment agency. Other counties in which RDA-sponsored housing accounted for a substantial share of permits activity were Santa Clara (20.8 percent) and Contra Costa (7.5 percent). And although it led the state in volume, RDA-sponsored housing production in Los Angeles County accounted for only six percent of 1988-96 permits. Statewide, RDA-funded housing production accounted for 4.2 percent of California residential building permits between 1988 and 1996. ¹⁵

RDA housing activity has varied over time as well as place (see Exhibit 21). Since 1988, California RDAs have funded or sponsored the rehabilitation of 2,500 to 3,000 housing units each year. RDA-sponsored new construction has varied more widely, ranging from a high of 6,000 units in 1988 to a low of 2,000 units in 1991, and then back up to 7,000 units in 1994. While the California recession played some role in depressing production during the early 1990s, exactly why there should be so much year-to-year variation in RDA-sponsored new housing production is not clear.

Year-to-year variations in RDA-funded housing production are even more evident at the local level, where construction is affected by many factors in addition to fund availability. Beyond meeting the 20 percent set-aside requirement, local RDAs have almost unlimited discretion regarding how and where they wish to use their housing funds within the jurisdiction. In addition to the volume of TIF revenues, yearly RDA housing activity depends on site availability; on the availability of complementary gap funding sources; on the interests and priorities of local public officials; on the willingness of neighborhoods to accommodate projects; and on the presence of capable non-profit housing developers and staff. In sum, RDA housing activity, like all manner of infill, is highly idiosyncratic.

These findings have two implications for the future. They indicate, first, that local redevelopment agencies as currently structured play a significant, albeit moderately-sized role in meeting projected housing needs through both new construction and rehabilitation—especially in older areas. RDA housing activities tend to be highly uneven. Currently, they occur only when and where there is strong local interest. Second, and more fundamentally, these findings suggest that if infill development is to play a bigger role in accommodating projected housing demand, the function of local RDAs may need to be expanded. Thus, changes to state redevelopment law may be among actions necessary if infill is to become a more regular and reliable source of housing production throughout California.

Summary

Regrettably, little more is known about infill development's potential for meeting California's future housing needs at the end of this analysis than was known at the beginning. Increasing national and local policy interest in infill housing notwithstanding, it remains a difficult activity to measure and explain. While infill activity during the 1980s was more likely to have occurred in transitional multi-family neighborhoods than in single-family ones, whether this was due to market forces, local zoning ordinances, or community acceptance is difficult to discern. Infill housing's attraction to growing, moderate-density neighborhoods suggests—but does not demonstrate conclusively—that it is site availability, economic feasibility, the ease of development, and most of all, consumer preferences, which most affect infill production levels. What role State laws and production programs can have in promoting infill housing is a topic which is again considered in Chapter 7.

Chapter Summary

• Enough Land for Housing: A detailed analysis of land supply in 35 urban counties reveals that California has more than enough raw land to accommodate projected housing growth (at current densities) through the year 2020 and beyond, even allowing for wetland, farmland, hillside, and floodplain protection. Based on the criteria here, two of the State's most populous counties, Los Angeles and Orange, will not have sufficient raw land to accommodate projected 1997-2020 housing demand.
• Land Supplies Complicated in Counties: The land supply picture becomes more complicated at the level of individual counties. Allowing for wetland, hillside, floodplain, and farmland protection, supplies of developable land are most plentiful (relative to projected housing demand) in the south Central Valley, the Central Coast, and the Inland Empire. In the Bay Area, developable land supplies are most plentiful (relative to projected housing demand) in the four North Bay counties.
  
  Elsewhere, land is not quite not so plentiful. Among high-growth counties, Alameda, Contra Costa, Riverside, Sacramento, San Diego, San Joaquin, San Mateo and Santa Clara, each have between two and five times the amount of developable land required to accommodate projected household growth through the year 2020.
• Endangered Species' Impact: Prohibiting development on sites with large concentrations of endangered species would further constrict developable land supplies in Alameda, Contra Costa, San Diego, San Mateo, Santa Clara, and Ventura counties to a point where each would have less than three times the amount of developable land required to meet projected 1997-2020 household growth. Santa Clara County, in particular, would lack sufficient land reserves to accommodate projected growth.
• Careful Analysis Needed: The adoption of even moderately-limiting urban growth boundaries by high-growth counties in the Bay Area and Central Valley would further constrain land supplies below the levels required to meet future housing production needs. This finding points to the need for carefully analyzing and ensuring the development capacity inside any potential UGBs before they are adopted.
• Development Densities on the Rise: At the same time, development densities are on the rise in many parts of California—as a result both of market forces and regulations. Should this trend continue, even modestly, it will substantially increase the ability of California counties to accommodate projected housing growth.
  
  Beyond these conclusions, this analysis says nothing of the ability of individual California jurisdictions to accommodate or not accommodate projected population and household growth within the conditions and restrictions of their respective general plans and zoning ordinances.
• Infill Housing Production is Complex: Infill housing production, whether on vacant parcels or through site reuse, remains an idiosyncratic and hard-to-explain process. The importance of land and site availability in encouraging infill housing development is unclear. Market forces matter more in some places than others, as do the type and quality of the existing housing stock. If infill housing is to meet a larger share of California's housing needs in the future than it has in the past, simply promoting infill will not be enough. The reasons why infill flourishes in some places but not others, and the constraints which limit infill housing will need to be studied and addressed on a city-by-city, project-by-project basis.
• RDA Activities Uneven at Local Level: Most of the infill housing built (or rehabilitated) in California since 1980 has occurred with assistance of local redevelopment agencies. Statewide, RDA-funded housing construction and rehabilitation accounted for less than four percent of total 1988-96 housing production. At the local level, RDA housing activities tend to be highly uneven. Currently, they occur only when and where there is strong local interest. These findings suggest that if infill development is to play a bigger role in accommodating projected housing demand, the function of local RDAs may need to be expanded.
  
  

References

Association of Bay Area Governments. Digital land use inventory. Oakland. 1985, 1995.

California Farmland Mapping and Monitoring Project. Digital map layers. Sacramento: California Resources Agency. 1984, 1996.

California Department of Fish and Game. Special Natural Areas digital maps. Sacramento. 1996.

California Department of Housing and Community Development. Annual summaries of redevelopment activity. Sacramento. 1987-1998.

California GAP Analysis Project. 1996 Gap digital data and maps. Santa Barbara: University of California, Santa Barbara.

Dardia, Michael. 1998. Subsidizing Redevelopment in California. San Francisco: Public Policy Institute of California.

Federal Emergency Management Agency. 1996. Digital Q3 floodzone map layers. Washington, D.C.

Silicon Valley Manufacturers Association. 1995. Housing Solutions for Silicon Valley.

United States Bureau of the Census. 1995 TIGER files (California). Washington, D.C. 1996.

United State Bureau of the Census. 1980 Census of Population and Housing. Washington, D.C.

United State Bureau of the Census. 1990 Census of Population and Housing. Washington, D.C.

United States Geological Survey. 1:24,000 digital (digital) topographic maps. Menlo Park. 1996.

United States Department of the Interior. National Wetland Inventory digital maps. Washington, D.C. 1994.

Endnotes

• These estimates were developed from the 1996 digital maps provided by the California Farmland Mapping and Monitoring Project, a division of the California Resource Agency.
• The purpose of the analysis was to determine where future urban development might occur, not where it should occur. Under California planning law, all determinations regarding when, where, and what types of urban development should occur are made by local elected officials.
• We did not collect or analyze parcel data, which includes ownership and land use information. A hectare measures 100 by 100 meters and equals 2.47 acres.
• Exclusive agricultural zoning, which is rare in California, stipulates that land may not be developed for uses other than agricultural. Large-lot zoning permits non-agricultural uses, but limits their intensity well below typical suburban residential levels. This makes development economically infeasible. In addition, many California farms are protected from development under 10 and 20-year, renewable "Williamson Act" contracts.
  
  Not all farmlands are of the same value. Lands classified as "prime" or "unique" by the California Farmland Mapping and Monitoring Project (FMMP) are typically the most fertile and/or used to grow higher-value crops. Lands classified as "of state importance" or "of local importance" are less fertile, or else are planted with lower-value crops. "Grazing" lands have the lowest fertility. Disagreements frequently arise regarding which types of farmlands most deserve protection. Many general plans, for example, favor the protection of grazing land not because of its fertility or crop-value, but because it commonly functions as a greenbelt or as urban open-space.
• As identified and mapped by FEMA, Q3 floodzones indicate those areas within the 100-year, or 1% probability floodzone.
• California municipalities wishing to limit outward expansion have a number of regulatory tools at their disposal, not just UGBs. They can refuse to annex unincorporated lands. Working through county Local Agency Formation Commissions (LAFCOs), they can adopt strict "sphere-of-influence," or urban service boundaries (USBs). USBs differ from UGBs in that they don't categorically prohibit urban development, they just refuse to provide required urban services, effectively prohibiting intense urban development. USB and UGBs adopted via legislative action can be extended via the city council or county board of supervisors at any time. Depending on how they are worded, UGBs and USBs enacted by initiative may not be so easily extendable. Hence, the recent attraction of initiative-based UGBs.
• Gross urban densities are generally lower than net residential densities. This is because they are based on the total amount of urbanized land, not just residential land. Assuming that housing accounts for 75 percent of urbanized area, a gross housing density of 2.4 units per acre would be equivalent to a net residential density of 3.2 units per acre.
• Densities also change in response to market pressures. Residential densities rise when land is in short supply, when the market favors more dense housing forms (e.g., apartments), or when low-density development is precluded.
• Little empirical research has been done estimating land reserve requirements. Land economists and land use planners typically use a range of 30-50%.
• The U.S. Geological Survey uses satellites to collect land cover data, including information on urbanization. Data is currently collected using LandsatT. Prior to 1976, land cover data was collected using the Corona satellite. USGS land cover data is distributed in multiple forms, including GIRAS, NSDTS, and Arc/Info export format.
• The difference between raw land development and vacant land infill development isn't always clear. Vacant land infill, for our purposes, refers to the development of serviced sites located within the existing urban areas (as identified using the California Farmland Mapping and Monitoring Project's definition of urban). Raw land or greenfield development refers to the development of unserved sites at or beyond the urban footprint's edge.
• Redevelopment agencies are not the only local government agencies involved in housing production. Local public housing authorities and departments of planning, housing, community development, and economic development also sponsor and fund housing construction.
• The Census Bureau's system of numbering and identifying census tracts changed between 1980 and 1990. Large tracts, particularly those in growing suburban areas, were split into smaller sub-tracts. Even in places that did not gain new tracts, sub-tract boundaries and identification codes frequently changed. Because of these inconsistencies, it was not always possible to match 1980 census tracts with their 1990 counterparts. Where possible, sub-tracts were aggregated to their tract totals. Tracts whose sub-tract numbers had changed without reference were eliminated from the analysis. Even when clean matches were achieved, in some cases the data still seemed questionable.
• A complete evaluation of the role of redevelopment agencies in meeting California's past and future housing needs is far beyond the scope and purpose of this report.
• Assisted housing production by other entities for the study period is not available. It is estimated such production would also represent a relatively small share of total housing production over the period.
  
  

Copyright ©2000. All rights reserved.