Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 25


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 24
METHODS As highlighted in Table 2-1, the following methods can be applied to identify the locations and activity space of protected populations. Method 1. Local knowledge and public input Practitioners and stakeholders involved with the planning process usually are able to provide considerable insight into the population within a designated study area. Public input can be obtained as part of the community involvement process through interviews, surveys, focus groups, and feedback from public meetings. When to use. Using local knowledge and public input is among the most commonly applied methods for identifying protected populations. This method is well suited for identifying effects that are distributed spatially, as well as those that are distributed among transportation system users. It can be used in all environmental justice assessments performed for transportation system changes at the project, corridor, and systems level. Even if other methods are used to identify protected populations, local knowledge and public input should be used to verify results (see box titled "Using local knowledge and public input to validate census data," p. 25). This method is also effective as an initial screening technique to determine if other, more data-intensive methods for identifying protected populations are needed. Analysis. The following techniques should be considered for identifying protected populations through application of local knowledge and public input. Interviews - In-person one-on-one interviews can be conducted with individuals identified as community leaders (people who work with or represent other people, e.g., neighborhood activists, elected officials, clergy, and representatives of local interest groups). These individuals are likely to have knowledge of, and insight into, local issues that cannot be found elsewhere. Interviews should be conducted early in the process so that the information gained can be taken into consideration as soon as possible during project development. Surveys - Surveys may be carried out using samples that are broad-based (general population- based) or more narrowly focused (neighborhood-based). Some are scientific and produce statistically valid quantitative data; others are more informal and produce a mixture of qualitative and quantitative information. Surveys should be designed on the basis of the information that is needed. For example, if quantitative data are required for a particular district or area of the community, a formal survey should be designed to be statistically representative of individuals in the entire district. On the other hand, if information is needed from a group of people who have a particular interest in an issue, an informal survey should suffice. Focus groups - A focus group is a small group discussion run by a facilitator. The group is carefully selected, either randomly or nonrandomly (to secure representation of particular groups). A random group will ensure representation of all segments of a population, whereas the nonrandom group will be helpful in eliciting a particular viewpoint or position. A focus group generally has the following elements and objectives: a scripted agenda (including five or six 24

OCR for page 24
major questions); an emphasis on gathering many different points of view rather than on presenting information; identification of major points of agreement and disagreement; and approximately 8 to 12 participants. Feedback from public meetings - A public meeting (or open house or hearing) is a forum for receiving comments from the public. These meetings generally have ground rules regarding listening and speaking (e.g., a time limit for speakers). Participants also have an opportunity to submit written comments at the time of the meeting or afterward. An advertising strategy is always necessary to ensure good participation in the public meeting by as large a segment of the local population as possible. Using local knowledge and public input to validate census data The case of Louisiana Energy Services (LES) is a good example of why local knowledge and public input must be used to identify protected populations and issues that can arise if census data are not augmented by additional information sources. In 1994 the Nuclear Regulatory Commission (NRC) heard a complaint filed by the group Citizens Against Nuclear Trash (CANT) against proposed construction of a uranium enrichment plant. LES was planning to build the plant in Claiborn Parish near Homer, Louisiana. CANT represented the small communities of Center Springs and Forest Grove in the complaint. Among other things, the complaint alleged that the environmental impact statement (EIS) did not sufficiently address all adverse environmental, social, and economic effects, and that there was racial bias in the choice of the plant's location. LES planned to reroute a road and greatly increase the travel distance between the two small communities. These communities, populated by individuals in protected population groups, were not discussed in the Draft EIS and were not identified on a map that showed local communities. They were overlooked because the Draft EIS analysis relied solely on census data and map sources with a coarse level of resolution that did not show the two communities. The NRC agreed that the EIS did not adequately address impacts to the two communities and, ultimately, LES was required to resubmit sections of the EIS. The map was changed in the final EIS. In addition, LES had to revisit the entire siting process and show that racial bias did not play a part in the choice of the site. At this point, LES withdrew its application. Data needs, assumptions, and limitations. Information should be collected on locations and names of communities on maps, and important community centers should be identified, along with public spaces where protected populations may congregate. Assume that a representative cross-section of the communities in the study area are involved in the process. This process may be time consuming and does not produce estimated population sizes or demographic characteristics. Also, these data cannot be readily incorporated with quantitative assessment. Results and their presentation. Map census data indicating neighborhoods and important community centers identified in a survey or set of interviews. This will produce a tabulation of 25

OCR for page 24
protected population groups, their residential and activity spaces, and important modes of transportation. Assessment. Techniques for obtaining input on protected populations from individuals with local knowledge should be applied to assessment of most transportation policies, programs, and projects. Techniques such as interviews, surveys, focus groups, and feedback from public meetings are often the best way to identify where protected populations live, their activity spaces, and modes of transportation. Both qualitative and quantitative information can be collected, depending on need. Method 2. Threshold analysis using large-area census data Census data from 1990 and 2000 are reported at many levels of analysis and are available from numerous sources, including the U.S. Census Bureau, the Internet, and federal repository libraries. Large-area census data can be obtained for states, counties, census tracts, and census- designated places. Small-area reporting units available in census data include traffic analysis zones (TAZs), block groups, and blocks. Evaluation of small-area census data is discussed as a separate method. It should be noted, however, that the large-area evaluation methods discussed below can be used with small-area data when needed. When to use. Large-area census data are useful for evaluating the distributive effects of state and regional transportation plans and other systems-level planning efforts. These data also may be applied to initial assessment-of-corridor studies, depending upon the size of the corridor and the nature of the transportation system change that is being evaluated. Additionally, this type of data can be used when spatial demographic patterns must be evaluated or to estimate transportation user demographics at the systems level. In general, protected populations are identified from large-area census data using some form of threshold analysis. Analysis. Threshold analyses are carried out in five steps: Step 1 Define the study area. Consider the scope of the proposed transportation system change and select a study area that encompasses all affected areas and populations. For a transportation investment plan, the study area may even be statewide. A regional long-range transportation plan may require a multicounty study area. Step 2 Select analysis units to be used. The selected analysis unit (e.g., county, tract, TAZ) must balance the amount of data to be evaluated with the level of precision required to identify the distribution of effects. For example, county units of analysis may be appropriate to evaluate distributive effects of a statewide multimodal transportation investment plan. County-level data are not sufficient, however, to evaluate whether protected populations have equitable access to a regional transit system. In this case, census tracts or TAZs may be a more appropriate choice to address the distances that members of protected population groups must travel to reach regional transit nodes. Step 3 Acquire data and compute demographic statistics. Information on where to obtain large-area census data is presented in "Data needs, assumptions, and limitations." Depending on the data source, it may be necessary to compute the demographic statistics needed to identify 26

OCR for page 24
protected populations; the Census Bureau does not always precalculate these statistics. Some data sources, however, provide precalculated variables that can be used when appropriate. Guidance on techniques for calculating demographic statistics from standard census data files can be found in Appendix D. Step 4 Determine threshold levels. When using a data-driven technique, it is often necessary to define "protected populations" on the basis of threshold values for various demographic variables. The thresholds are used as comparison values to determine if protected populations exist in a study area. For example, if an area's percent minority population is not equal to or greater than the established threshold value, the level of environmental justice concern can be assumed to be lower than in areas where the statistic is greater than the threshold. Two basic approaches are used to define thresholds. The first approach commonly used by agencies is to establish a working group to evaluate and determine appropriate threshold levels. Recently in New York, the state environmental agency's environmental justice working group set statewide thresholds that were approved after being submitted to the public for comment. A second commonly used approach is to set thresholds to equal either state- or county-level averages, depending upon the size and geography of the study area (see box titled "Limitations of using comparison thresholds in environmental justice assessment," p. 28). Step 5 Identify protected populations. When using large-area census data, all evaluation units generally include members of protected populations. It is thus more appropriate to consider this technique as an approach for categorizing evaluation units based on the proportion of protected populations that they contain. Evaluation units with protected population levels greater than the established threshold values are considered to have substantial protected populations and higher potential for distributive effects than other evaluation units. An example is shown in Table 2-2. The results were obtained from a review of county-level 2000 census data for the Texas Department of Transportation, Houston District. Table 2-2. Relative level of environmental justice concern in the Houston district Percent Percent Relative level Area minority low income of concern Threshold value (State of Texas) 47.6 17.0 -- Brazoria County 34.6 12.5 Lower Fort Bend County 53.8 7.3 Higher Galveston County 36.9 16.8 Lower Montgomery County 18.6 11.9 Lower Harris County 57.9 14.9 Higher Waller County 50.1 20.0 Higher 27

OCR for page 24
Limitations of using comparison thresholds in environmental justice assessment Background. Thresholds often are a necessary component of data-driven evaluation techniques. They serve as a useful way to define "level of environmental justice concern" (a commonly used term in environmental justice assessments) based on demographic characteristics. It is important, however, to use threshold comparisons wisely, as part of a thorough evaluation process. Although they can serve to categorize and prioritize areas based on level of concern, the problem with thresholds is that they are arbitrary. Example. Consider as an example an environmental justice assessment of a proposed airport expansion. Environmental review indicates that most adverse noise and air quality effects on surrounding populations will occur within 1 mile of the airport boundary, so a 1-mile buffer of the airport is selected as the study area. Because the study area is completely contained within a single county, the county's average demographics are selected as the comparison threshold. A review of small-area census data (described later in this chapter) shows that the percent of study area population living below the poverty level is 14.5 percent, which is below the county average threshold of 16.5 percent. Similar results were obtained in a review of minority population data. Based on this information, it could be concluded that there is limited concern about adverse distributive effects to protected populations. Discussion. The above approach is common practice. The assessment process is relatively easy to perform, and the findings and conclusions are easily presented. It is often advisable, however, to perform a more thorough analysis in many situations. Figure 2-1 continues with the airport expansion example and presents the results of a thorough demographic evaluation performed at each quarter-mile increment from the airport boundary out to 8 miles. The figure supports the conclusion that the low-income percentage within the 1-mile area is less than the county average threshold. A more interesting observation is that the highest percent of people living below the poverty level is from 1.25 to 2.5 miles from the airport. Conclusion. Instead of establishing an arbitrary threshold for analysis, it may be more appropriate to perform a thorough review of the demographic data; identify areas where substantial minority populations and low-income populations live or work; and then evaluate effects to determine the beneficial and adverse distributive effects that would result in those areas. 100% 90% 80% Percent of Population 70% 60% 50% 40% 30% State (21.8%) 20% County (16.5%) 10% Persons Living Below Poverty Level 0% 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 Distance from Project Site (Miles) Figure 2-1. Percent of population below poverty level 28

OCR for page 24
Data needs, assumptions, and limitations. The single best sources of data are Summary File 3 of the 2000 Census for geographically distributed effects and the Census Transportation Planning Package (CTPP) for effects experienced by system users. Still, useful demographic data are also available in other census summary files. For comparison with previous census data, use Summary Tape File 3 from the 1990 Census. Online resources for these data are included in the resources at the end of this chapter. These data can be combined with census TIGER (Topologically Integrated Geographic Encoding and Referencing) files in GIS if maps are needed to analyze spatial distributions. This method contains several important assumptions that serve to limit its effectiveness. The most important limitations include the following: Populations are not distributed uniformly. Most techniques that use census data assume that populations and population characteristics are uniformly distributed within census units. This assumption can be especially problematic when working with large- area census data. Census geography is hierarchical. So, for example, counties are composed of tracts; tracts are composed of block groups; and block groups are composed of blocks. In general, population characteristics will be most uniformly distributed in blocks and variability will increase with the size of the census unit. It has long been known that statistical studies will yield different results depending upon the level of census data that are analyzed. This is in part because the data violate the assumption of uniform distribution to differing degrees (Fotheringham and Wong 1991; Amrhein 1995). An example of the variability in results is provided in the discussion of "GIS-based techniques to estimate demographic characteristics," where block and block group-level data provide different results for the same study area. The level of demographic resolution should match the scale of effects. Resolution, or scale, is another limitation of large-area census data and is related to the uniform distribution assumption. Data for large-area census units are totals of the small-area census units contained within them. Large-area census units thus have less resolution than small-area census units. Therefore, it is important that large-area census data not be used in instances where a high degree of demographic resolution is needed. An example would be assessing the effects of noise on nearby residences, which is a highly localized effect. In such instances, it is more appropriate to use small-area census data, even if the study area is very large. In other instances, large-area census data are appropriate. An example would be an analysis of transportation policy changes on regional air quality. The Census may undercount protected populations. Survey and enumeration techniques used in census data collection rely heavily on address lists. It has long been argued that low-income, minority, and other protected populations are consistently undercounted and that the census is not an entirely accurate representation of these populations. Elaborate statistical techniques are available to correct for population undercounting, but these techniques rely on assumptions that are often difficult to validate. It is probably more appropriate when working with census data to understand that protected populations may be undercounted and to apply conservative threshold levels and use local knowledge as a means of verification. 29

OCR for page 24
GIS-based techniques to estimate demographic characteristics Study area boundaries often do not coincide with the boundaries of census reporting units. When performing environmental justice assessments, it is important to understand the relationship between study area geography and census unit geography. It is also important to understand that use of different census reporting units, such as block groups or blocks, may yield different results. Three common GIS-based methods for estimating study area population characteristics are known as polygon containment, polygon intersection, and areal interpolation. Each involves a different approach for identifying the census reporting units that overlay a study area. Areal interpolation is the most accurate but is more data intensive than the other methods. The polygon-intersection technique tends to over- predict population, whereas the polygon-containment technique tends to under-predict population. Census blocks offer the most detailed census geography and will yield the most precise estimates, but only limited demographic information is reported at the block level. Figure 2-2 shows the census block and census block group geographies used to derive population estimates for a 1-mile study area surrounding a proposed bridge location. The shaded census units show the area from which the 1-mile area estimates are determined. Note that the same 1-mile buffer is shown in each diagram. The population estimates obtained from the various methods and census units vary by as much as 12,092 persons, or 4.7 percent. Total population: 10,992 Total population: 11,187 Percent minority: 95.8 Percent minority: 96.2 Total population: 13,278 Total population: 18,942 Percent minority: 95.1 Percent minority: 92.9 Total population: 9,648 Total population: 6,850 Percent minority: 97.2 Percent minority: 97.6 Figure 2-2. Census reporting units and study area geography 30

OCR for page 24
Results and their presentation. Results of large-scale census data analysis are often presented in narrative format. Using the data presented in Table 2-2, results could be expressed in the following manner: In the Houston District, Fort Bend, Harris, and Waller Counties have a higher proportion of individuals in protected population groups than other counties in the District. Tables can be used to convey large amounts of information in technical reports, but the amount of information should be kept to a minimum if tables are to be used in public forums. Maps can be used to convey spatial patterns, but maps are often more appropriate for presenting small-scale census data. Assessment. Large-scale census data are suitable for identifying protected populations if the patterns of effects are uniform over large areas. Much useful information is available for large- scale census units, such as tracts and counties. This information can be used to assess both the spatial distribution of protected populations and the demographics of transportation system users. Large-scale census data are best suited for assessing state and regional policies and programs, and for transportation system changes that would generate system-wide effects. Such census data are generally not suited for project-level analysis and should be used cautiously for corridor-level assessment. Method 3. Spatial interpolation using small-area census data Small-area census data at the block, block group, and TAZ level offer the most detailed nationally available demographic information useful for identifying protected populations. Database, spreadsheet, and GIS software are often necessary for this type of analysis because even relatively small study areas commonly encompass a large number of small-area census units. When to use. Small-area census data should be used in situations where the scale of effects to be analyzed requires a high degree of demographic resolution, such as when project effects are limited to relatively small, localized areas. Small-area census data can also be used if results of studies using large-area census data are questioned, making it necessary to obtain the "best available" or most accurate census data. The small-area census data and the techniques described below are most applicable to project- and corridor-level analysis. Blocks and block groups can be used to assess spatial demographic patterns, whereas traffic analysis zones should be used in situations where transportation user demographic characteristics are required. Analysis. Spatial interpolation using small-area census data is conducted in five steps: Step 1 Define the study area. Because small-area census data are best suited for identifying protected populations in situations where effects are localized, it is often possible to define a study area based on detailed geographic patterns of effects. For example, contours or receptors developed from noise and air quality analyses could be used to define areas of effects. Viewsheds may be selected as the area of effects to address visual quality impacts. A more simplified approach is to select a buffer distance that encompasses the geographic extent of 31

OCR for page 24
effects. GIS can be a useful tool for defining study areas. Contours generated from modeling programs can be brought into GIS for overlay with census data, and buffering tools can be used to develop study areas from site locations or project corridors. Step 2 Compute statistics for protected populations. If patterns of effects are distributed spatially, block groups and blocks should be selected as the census units for analysis. A combination of block groups and blocks is recommended because, while blocks offer the highest level of resolution, all potentially necessary data are not reported at the block level. If patterns of effects are distributed among transportation system users, traffic analysis zones should be used where possible. Data for traffic analysis zones are available in the 1990 and 2000 CTPP. Table 2- 3 lists the protected population demographic data that are available at the various census reporting levels. Statistics can be calculated using various spreadsheet, database, or GIS applications. Formulas for computing useful demographic statistics can also be found in Appendix D. Table 2-3. Availability of protected population data by 2000 census reporting unit Census Total English reporting unit population Minority Income Age Gender Disability speaking Block Traffic analysis zone Block groups, tracts, and larger Step 3 Overlay demographic data with area of effects. The purpose of this step is to identify the census-reporting units that fall within the area of effects. The three overlay methods for selecting census-reporting units are known as polygon intersection, polygon containment, and areal interpolation. Each approach can provide a different estimate because areas of effects do not commonly coincide with census unit boundaries. For more information, see the discussion on "GIS-based techniques to estimate demographic characteristics." Step 4 Estimate demographic characteristics of the study area population. The demographic characteristics of the study area population can be tabulated once the census units within the area of effects have been identified. Values can be reported for the individual census units within the study area, which is useful for assessing the population distribution and characteristics. It can also be useful to generate an estimate of the population characteristics for the area as a whole. An areal interpolation technique is best suited for developing this estimate. The most common and easily performed method is called area-weighted interpolation, which assumes that populations within the census units are uniformly distributed (Goodchild and Lam 1980). Many interpolation routines do not rely on the uniform distribution assumption. One of the more promising techniques, especially useful in sparsely populated areas, uses an overlay of 32

OCR for page 24
road networks with census units under the assumption that most residences are near a roadway (Mrozinski and Cromley 1999). Another method uses ancillary data as an input mask to identify areas within census units where there are no residences (Bloom, Pedlar, and Wragg 1996). In this way, nonresidential areas such as parks and areas zoned for commercial or industrial use are assigned zero population. Step 5 Compare to thresholds for analysis. The final step is to compare demographic estimates to analysis thresholds in order to identify the presence of protected populations. This step is the same as that for the analysis using large-area census data. Data needs, assumptions, and limitations. The following data sources will be necessary, depending on the census-reporting units selected for analysis: Summary File 1, to compute block-level statistics; Summary File 3, to compute statistics for block groups, tracts, and larger reporting units; and The Census Transportation Planning Package to compute statistics for TAZs. Information about acquiring and using these data sources is provided in Appendix D. Analysis of small-area census information can be quite data intensive and requires some combination of spreadsheet, database, and GIS software. GIS software is needed to perform the interpolation methods discussed above. The assumptions and limitations discussed for large-area census data (see Method 2) generally apply to the evaluation of small-area census data. In addition, the complexity of many analytical processes and the amount of data required often limit the use of small-area census evaluation methods to situations where other less intensive techniques have not provided adequate results or results have been contested. Results and their presentation. Results can be presented as maps, graphs, and tables (as shown in Figure 2-3) depending on the purpose and the intended audience. Maps are the best means of presenting geographic patterns and are often essential for conveying the proximity of protected populations to sources of beneficial and adverse effects. Maps should be relatively simple, showing only the census data theme, such as percent minority by census block, and enough other features to orient the reader. Graphs, on the other hand, can be used very effectively to provide a comparison of study area demographics to comparison areas or threshold levels. Tables should be used primarily in technical reports; only tables with five to seven data values or fewer should be used for communication with the public. Tables do, however, offer a very useful way of organizing and summarizing the results of small-area census data evaluations. Assessment. Compared to other techniques, analysis of small-area census data is more complex and more data intensive. However, this method offers the finest demographic resolution available with census data and should be selected over other methods in situations where the effects from a proposed transportation system change will be localized to specific areas. 33

OCR for page 24
90.0 81.6 78.2 80.0 70.0 Percent Minority Population 62.3 58.1 60.0 47.6 50.0 40.0 30.0 20.0 10.0 0.0 1/2 Mile 1 Mile 1.5 Miles County State Area Distance Total population Total minority Percent minority 1/2 mile 124 72 58.1 1 mile 1,249 1,019 81.6 1-1/2 miles 4,717 3,690 78.2 2 miles 9,195 7,497 81.5 County 313,645 195,467 62.3 State 20,851,820 9,918,507 47.6 Figure 2-3. Techniques for presenting study area demographics Method 4. Field survey A field survey, also known as a dashboard or windshield survey, involves obtaining local knowledge by actually traveling about the area and taking notes. When to use. A field survey generally is recommended as part of project-level environmental justice assessments; they are less practical for corridor and system-wide assessments. Field surveys are especially important in situations where project effects such as noise and air quality will be highly localized and census data do not provide a fine enough level of resolution. A field survey is also a good technique for verifying the accuracy of small-area census data. Field surveys can be comprehensive in small, manageable study areas. If projects cover large areas or field surveys are conducted at the corridor level, it may be more appropriate to identify specific locations for a field study. Field surveys can be used to collect the following information that is not obtainable from a review of census data alone: Mapping the location of residences in a study area. 34

OCR for page 24
population projections for jurisdictions can be no greater than the estimates produced by the state's Demographic Research Unit. The housing inventory is used to estimate population by analysis unit. This is done by combining housing unit information with base-year census population data and correlations of population by dwelling type, average household size, and housing unit vacancy rates. These factors are combined to generate a base-year household population estimate. Populations for group quarters (available from local agencies) and homeless individuals (available in the census) are added to household population estimates to generate the base-year total population estimate for each analysis zone. For purposes of environmental justice assessment, it is important to identify the protected population characteristics of the base-year population. The employment inventory is established through surveys of the study area. Employment is considered because it is a significant factor in drawing new residents to an area and in retaining current residents. School facilities and enrollment information is collected in order to (a) better plan for educational infrastructure needs and (b) refine age-cohort information. Step 1c. Evaluate holding capacities. Determine the maximum number of jobs and housing units that can be accommodated by each analysis zone. This is the zone's holding capacity. For housing, compute densities for each type of residential land use from general plan and zoning information. The average housing density can then be applied to the total acreage of each land use type in the analysis zone. These holding capacities are not fixed; they vary as land use changes and must be updated on a regular basis, at least every few years. For jobs, develop an employment yield matrix. The SACOG approach is to develop estimates of number of employees per acre for five employment types: retail, office, medical, manufacturing, and other. Each employment type can be correlated with land use maps and acreages computed for each analysis zone. Multiplying acreage by number of employees per acre for each employment type yields the employment holding capacity for each analysis zone. Estimates for education-related employment can be gathered from surveys of local educational institutions, and the number of education employees in each analysis zone assigned by address. Step 1d. Determine phasing. Phasing is the process of developing growth curves, using current population and employment as the starting point for the base year and holding capacity as the end point for the horizon year. Individual growth profiles should be developed for each unique type of development pattern that exists in the study area. SACOG uses the following four development patterns: Limited room for growth applied to areas already at or near their holding capacities; Growth occurring currently steady growth beginning in the base year and continuing through a future build-out date determined from general plan information; Current development static; growth occurring later areas expected to begin growing at a later date and to continue growing through a build-out date based on general plan information; and Redevelopment when land use changes alter the housing and employment composition of an analysis zone. In this case, you should base projections on the expected year of redevelopment and the extent of redevelopment. 44

OCR for page 24
Step 1e. Compute jobs-housing ratio. The jobs-housing ratio is calculated by dividing an analysis zone's total number of jobs by its total housing units. The ratio is a measure of the mix of employment and housing in a given area. Step 1f. Derive population. Compute the number of households by household type for each analysis zone and for each year and then multiply by the estimated number of persons per household by each household type to calculate the total household population. Add the population living in group quarters. Adjust the population per household and vacancy rates based on the growth profile of the area. Adjustments are related to changes in housing type from single family to multifamily and changes in age of development. Step 1g. Conduct jurisdictional review. As described earlier, plan-based population projection is an iterative process that relies heavily on the knowledge and understanding of planners and decision-makers as to how proposed and expected land use changes will affect population over time. Jurisdictional review is therefore an opportunity for experts to validate results and, if necessary, provide the further information necessary to alter projections. Stage 2 Estimate populations of protected groups The commonly used small-area projection techniques do not consider protected population characteristics. This stage of the population projection method is a technique for estimating future protected population characteristics based on the results of Step 1f above and trend information. Step 2a. Collect and prepare data. Use census data for the base year to generate predictor variables for the response variables percent low-income population and percent minority population. The following variables utilized in Stage 1 should be considered as candidate predictor variables for both percent low-income and percent minority: housing costs (housing unit values and rental costs), housing unit density, housing unit vacancy rates, single family to multi-family housing ratio, average household size, jobs-housing ratio, unemployment rate, and housing density to housing capacity ratio. For minority population, also consider low-income as a predictor variable and vice versa. Housing cost and unemployment information are both potential predictors of protected populations. These variables must be added to the Stage 1 data collection efforts and future-year estimation efforts in order to be used as predictor variables. Step 2b. Conduct exploratory study. The result of a multivariate analysis is a best-fit curve allowing estimates of the response variable to be derived from known values of predictor variables. A benefit of this approach is that prediction error estimates can be reported. Note that prediction error rates can be high for exploratory observational studies such as that described here. The purpose of the exploratory study is to reduce the set of candidate predictor variables listed in Step 2a to the set to be used in the regression model. Highly intercorrelated predictor variables should be eliminated, as should other variables found to have low correlation with the response variable. A variable reduction procedure should be used to develop correlation matrices and identify candidate predictors that should be retained in the final model. Although it is beyond the scope of this guidebook to provide an in-depth discussion of variable reduction, there are numerous books on applied regression analysis that can be read for further information, such as Neter, Kutner, Nachtsheim, and Wasserman (1996). 45

OCR for page 24
Step 2c. Refine and select model. The result of Step 2b is a small subset of candidate regression models with a limited number of explanatory variables that provide good predictive ability. Step 2c results in selection of the final model, based on review of residual plots and analyses to identify lack of fit, outliers, and influential observations. Step 2d. Validate model. Model validity is determined based on the ability to make generalized inferences. Validation should be performed in two stages. First, evaluate the ability of the model to predict known values for percent minority and percent low-income from base year analysis zones. Ascertain if the model provides a suitable fit across all ranges of land use types. Next, evaluate the predictions made for the projection years. Alternative step generate trends. It may be necessary to compute protected population projections from generalized trends if an acceptable regression model cannot be developed. This alternative step is technically less complex than the regression model steps, but it is otherwise less desirable because it does not provide the ability to report prediction errors. Estimates for changes in percent minority and percent low-income are categorical, more representative of trends than of actual population numbers. This may not be a severe drawback, however, because effects being evaluated over a long time horizon tend to be nonspecific in the case of policies and programs. In addition, for projects it is valuable insight to know, for example, that there is no present day concern about distributive effects but that the project area is expected to experience an increase in low-income population during the ensuing 15 years. The approach is similar to Step 1d (Determine phasing). The trends are developed based on general plan information and can be used to estimate percent low-income and percent minority population for each year in the projection. The following trends are provided as examples, but the actual trends must be developed to reflect the characteristics of the study area. Stable community the area is near housing capacity; and jobs-housing ratio and housing costs are stable. In this situation it could be expected that percent minority and low income would be stable through time. Growing community new housing is being developed and new jobs are being created in the area. Percent low-income would be expected to remain stable or even fall. Growth in the low-income population would be related to the availability of affordable housing. Changes in percent minority could be based on the racial composition of the expanding workforce. Declining community upper and middle-income residents are moving out of the area, housing vacancies are increasing, housing costs are stable or declining, jobs-housing ratio is declining. In this situation, percent low income could be expected to increase because affordable housing is important to low-income individuals. Changes in percent minority could be tied to the correlation between race and income in the study area. Redeveloping community land use changes alter the housing and employment composition of an analysis zone. These changes in turn affect the relative desirability of housing and could affect percent minority and percent low income. In this case you should base projections on the expected year of redevelopment and the nature of redevelopment. Step 2e. Compute protected population statistics. Using the regression model approach, percent minority and percent low income for each analysis zone can be calculated directly from the model. Reporting the confidence intervals allows for further assessment of the certainty as to 46

OCR for page 24
whether the minority and income composition of the population will change as predicted. Using the trend-based approach, percent minority and percent low income can be adjusted for each projection year based on knowledge of general demographic trends (i.e., general trend of increasing minority population throughout the study area), knowledge of housing unit turnover rates (to understand phasing), and findings from other areas that have experienced similar land use changes. Calculate number of low-income persons and number of minority persons by multiplying by the total population estimate. Step 2f. Conduct jurisdictional review. As with Stage 1, jurisdictional review is necessary to validate findings and to refine the assessment if the findings are deemed to be invalid. Data needs, assumptions, and limitations. Many of the data needs, assumptions, and limitations have been described above. In general, the technique for projecting changes in minority and low-income population groups is data intensive and time consuming. Trend-based or regression-model techniques can be used to derive protected population projections from base- year and future-year data. General population projection techniques such as trending and extrapolation assume that current population patterns will continue through time and are not well suited to small-area assessment. Land use and plan-based techniques work to overcome this limitation by incorporating information on how the study area is planned to change through time. Common projection methods used in practice estimate future year changes in total population but do not predict changes to subpopulations such as protected groups. The method described above is exploratory at best, and considerable effort must be given to fit the method to the characteristics of the study area it is being applied to. Results and their presentation. Results of this method are estimates of future year populations, including percentages and derived counts for minority populations and low-income populations. Present results as you would results for current year estimates. Assessment. Population projections are a difficult, complex, and time-consuming effort, but in certain instances they are a useful means for assessing the environmental justice of transportation policies, programs, and projects with long time horizons. Population projection techniques useful for transportation environmental justice purposes must provide estimates for small areas such as tracts and TAZs, and they should include predictions for protected population groups. The planning-based method presented in this guidebook must be carefully calibrated to the study area in which it is being applied. As with any small-area projection approach, it is important to refine projections so that they are current with planned land use changes, population trends, and economic fluctuations. Method 9. Environmental justice index The environmental justice index (EJI) is a method of scoring the relative level of environmental justice concern for census-reporting units based on population density, minority population, and low-income population factors. Because the EJI uses multiple factors, it is a good method for combining information to show the distribution of protected populations on a single map. 47

OCR for page 24
When to use. The EJI is particularly effective for showing relative concentrations of minority or low-income populations. It also is suitable as a screening technique to identify areas where detailed assessment and outreach should be conducted. Finally, the EJI can be used as a demographic variable in combination with indicators of the level or presence/absence of environmental effects in statistical evaluations to determine if effects are disproportionately high and adverse (U.S. DOT 2000). Analysis. There are three steps in constructing an environmental justice index: Step 1 Gather census data and compute demographic variables. Census data should be collected for the study area, and the level of analysis (tracts or block groups) must be selected. Demographic variables should be computed for population density, percent minority population, and percent low-income population. These variables need to be computed for the selected unit of analysis and at the state level for comparison purposes. Step 2 Calculate EJI factors. The standard EJI is represented by the following formula: EJI = DVPOP x DVMAV x DVECO where DVPOP = degree of vulnerability based on population density DVMAV = degree of vulnerability based on presence of minority population DVECO = degree of vulnerability based on presence of low-income populations These factors are computed as follows. DVPOP Population per square mile Score 0 0 > 0 and < 200 1 > 200 and < 1,000 2 > 1,000 and < 5,000 3 > 5,000 4 DVMAV and DVECO Percent minority or percent low income Score < State average 1 > State average and < 1.33 times the state average 2 > 1.33 times and < 1.66 times the state average 3 > 1.66 times and < 2.0 times the state average 4 > 2.0 times state average 5 In the standard formulation, the EJI thus ranges from 0 to 100. High EJI values indicate that a large population density is present and that a large proportion of that population is minority 48

OCR for page 24
and/or low-income. It is also possible to modify the EJI to include additional protected population factors for disability, age, and other characteristics. Step 3 Review maps and identify presence of protected populations. Once the EJI factors and the EJI are computed for each census unit, GIS can be used to produce maps of the study area. The maps will depict areas of high population density and large proportions of minority or low-income population. In many cases, specific neighborhoods can be identified and labeled on the maps. Data needs, assumptions, and limitations. The EJI is best suited for relatively large study areas and in situations where evaluation of small-area census data is desired. Block-group-level data are the most commonly used evaluation unit. Census data from 1990 or 2000 are required for the evaluation units (block groups, tracts, or TAZ), as are state data. For mapping in GIS, the 1990 or 2000 TIGER files for the study area also are needed. Appendix D contains instructions as to how these data can be obtained. The census data can be formatted and variables calculated using spreadsheet software such as Microsoft Excel or database software such as Microsoft Access. The EJI has the limitations endemic to any mathematical index. Indexes are useful for depicting combinations of variables as a single value, which makes them valuable for screening assessments and for mapping. The underlying factors must be used, however, if more detailed analysis is required. For example, an area with an EJI score of 40 to 45 indicates that there is potential for effects to protected populations. The EJI score by itself cannot be used, though, to determine if the area is densely populated (high DVPOP), has a high proportion of minority individuals (high DVMAV), or has a high proportion of low-income individuals (high DVECO). Furthermore, the EJI does not provide meaningful results in areas with relatively uniform population density and protected population characteristics. Results and their presentation. Figure 2-7 is a map of the EJI for block groups in a 1-mile study area surrounding a project corridor. The map clearly depicts areas of minority population and low-income population concentrations, based on EJI score. These areas are often well- defined neighborhoods and can be labeled for illustrative purposes. Figure 2-7. Example presentation of EJI results 49

OCR for page 24
Assessment. Identifying protected populations through use of the EJI relies upon census data and is best suited to block, block-group, and TAZ levels of analysis. The EJI can be used effectively to identify areas where further environmental justice assessment should be targeted based on demographic characteristics. Results of EJI evaluation are readily presented in simple maps that are easy to interpret. The EJI can also be used in statistical evaluations to identify disproportionate adverse and beneficial effects. Examples of this use of the EJI are given in other sections of the guidebook. As with any census-based evaluation method, local knowledge and public participation efforts should be used to verify that protected populations have been accurately identified within the area of concern. Method 10. Activity space analysis using personal interviews This simple approach can be an effective tool to provide a general idea of the areas that comprise the individual and communal activity spaces of the populations of interest. Such information supplements (it does not replace) insights derived from the foregoing methods that are largely based on the place of residence. When to use. This method is most appropriate for projects that are likely to be low impact and whose effects are expected to occur within a reasonably well-defined communal activity space. In such cases, a sample of residents is likely to provide useful insights on local activity spaces of the residents represented by the sample. Analysis. Space-time activity survey data should be able to help answer the following questions: What are the important daily activity centers (i.e., relatively frequent destinations) and what routes do individuals take to and from them? How frequently do members of the community typically access these centers? Are there numerous centers for different subgroups or do a few dominant? How do the important activity spaces relate to the impact area of the proposed project? Data needs, assumptions, and limitations. The key to any survey-based analysis, of course, is to draw a representative sample that is large enough to provide a clear picture of the population. In this case, the sample should be of members of protected populations in the general area of the community likely to be affected by a proposed transportation project. You should have a general idea of where protected populations move about before carrying out a survey to learn more about this activity space. The most important assumption is that the activity space defined by the survey sample will enable you to assess, with a fair degree of accuracy, the nature and magnitude of probable impacts. A limitation of survey-based analyses of activity spaces is that they afford only a general indication of each respondent's daily activity space. Thus, in the aggregate, a survey can be expected to provide a general sense of the activity space of a population of interest. Results and their presentation. The survey will enable you to derive a distribution of origin- destination pairs that depict the relative concentration of activity on the part of protected populations. This distribution can be presented graphically using GIS methods. Cells or zones 50

OCR for page 24
that have a relatively high concentration of activity should be given special attention when analyzing the several types of effects addressed in the chapters to follow. Assessment. It is widely understood that surveys are only as good as the sampling strategy used to select respondents and the questionnaire developed to address the issues. A statistically significant sample size and a well-conceived questionnaire can provide considerable insight into the areas of the community frequently visited by minority and low-income populations. Method 11. Activity space analysis using an abbreviated diary Measuring habitual travel behavior--determining what places a given population consistently travels to and the routes that they use--can be an effective way of measuring communal activity space. Because time and space constraints limit the number of activities that a person can perform on any given day, it is logical to assume that people try to maximize their utility in their choice of daily activities. A variety of research indicates a great degree of stability in weekday daily activities. Schonfelder and Axhausen (2003) found that two to four locations could account for up to 70 percent of all destinations visited in a 6-week period. Their study also indicated that eight locations represent fully 90 percent of the total trips made. Kitamura and van der Horn (1987) also characterized daily participation in different activities--work, leisure, shopping--as very stable. From an environmental justice perspective, the questions of interest are: What locations do protected populations access on a relatively frequent basis? How and by what mode do they access these locations? How much travel variability is there within the population of interest? Habitual behavior studies typically involve a simplified version of a travel diary that asks the time and place of locations visited throughout a specified period. When to use. This option is most appropriate for use when the understanding of a community's activity space is less certain than is needed for Method 10 but where potential impacts of the project or the data collection needs are not great enough to warrant a more time-consuming and expensive process. Because of the sensitive and personal nature of the data collected, anonymity is a prerequisite to gain the confidence of subjects. This approach requires both sufficient sample size and adequate administration of the equipment. Analysis. While the overall approach to designing the database and extracting the necessary data will vary, space-time activity data for a given study area should at the very least enable you to answer the following questions: What are the principal daily activity centers? What are the most common routes to and from these activity centers? How frequently do members of the community typically access these centers? 51

OCR for page 24
Data needs, assumptions, and limitations. The habitual travel survey should have the following qualities: 1) An adequate sample size and representation. As noted earlier, a carefully selected sample of sufficient size to mirror the population being represented is critically important. Recruiting the sample and adequately briefing participants is a substantial undertaking. 2) Clear and brief design. The user should be able to complete the diary in a relatively short amount of time. 3) Prompts that do not lead. Questions should strike a balance between stimulating participants' thoughts and pointing to a certain response. Sample questions could include the following: Where are the locations to which you travel daily or almost daily (i.e., work, school, day care, and any others you can think of)? How do you typically travel to these locations? Are these locations commonly visited by your neighbors? What other locations do you regularly (at least once a week) travel to (i.e., stores, restaurants, church, public park, and any others you can think of)? How do you typically travel to these locations? Are these locations also visited by your neighbors? What other locations do you travel to less than once a week and yet consider an important part of your life outside the home? How do you typically travel to these locations? 4) An adequate time frame. Participants will often vary their activities over a period of days (e.g., one may grocery shop every Thursday). It is thus helpful to give them a few days to complete their diaries to gain an accurate picture of these activities. Results and their presentation. Habitual travel behavior surveys can provide a fairly accurate picture of where protected populations travel within a community. They allow you to distinguish between frequently versus occasionally or rarely visited locations. Once you have an understanding of the general areas where respondents live and the destinations they travel to, it becomes possible to gain considerable insight into activity spaces. As with Method 10, GIS can be used to develop maps and other graphic depictions of relatively common activity spaces and the routes connecting people's homes to them. Assessment. This method relies on a diary format to define representative destinations and routes to and from them. As with Method 10, it is critical that the selected sample is sufficiently large and representative of the protected populations whose activity space you are interested in defining. It also is essential that the respondents understand exactly what information you are seeking. Because the data needed are quite simple, the questions should be simple and very clear. Method 12. Space-time activity analysis using GIS A more advanced method for exploring how the activity space of protected populations is configured within a given community is a form of daily trip diary. Miller (2001, p. 11) identifies four traditional forms for collecting spatial activity data using diaries: 52

OCR for page 24
1) Recall methods require subjects to recall and report activities during some previous time period. 2) Stylized recall methods require subjects to report "normal" activities that occur during some typical time period. 3) Diary methods require subjects to record activities in a diary, either in a free-format manner or at predetermined time periods. "Beeper studies" complement this approach by prompting subjects via a pager at selected time intervals to record their current activities. 4) Prospective methods, typically game-based, are employed in conjunction with other methods to investigate the effect of potential changes in the activity environment. Miller has sought ways to combine these traditional forms with GIS to create a robust, sophisticated model for analyzing activity space. Although he acknowledges some inherent flaws, the method combines space-time activity data with new GIS technology to analyze people's movements. When to use. With a sufficient sample size and appropriate sample selection design, this method can provide a fairly accurate picture of the activity space of the sample frame (i.e., the protected population within a particular area of a community or across a community). As this is a relatively advanced method, it should typically be used under two conditions: (1) when the nature of the project is large enough to warrant the time and expense and (2) when the data collected will help meet the data needs for multiple objectives. Robust data such as this may be necessary to reveal how people--especially low-income and/or minority populations--move and interact in their community differently than the broader population. Analysis. Typical travel diary analysis utilizes a point-based approach to mark the characteristics of points A and B. A path-based approach, however, is likely to better represent the type of disaggregate activity data required for a full environmental justice analysis. Although a more complex data-gathering task, path-based assignment meets the dually important objectives of identifying communal activity spaces and providing a reasonable estimate of the paths linking homes with these spaces. Path-based approaches can encompass a variety of methods for gathering data on personal activity space. A modern approach is to provide the respondent with a GPS receiver combined with recording devices [e.g., personal digital assistants (PDAs) or cellular transmitters tied to location-based services (LBS)]. The data stored by a recording device typically is entered into a GIS file to first calculate the shortest path between two points and then assign the trip along that path. Shaw and Wang's study (2000, pp. 167-168) separated the data for each trip into four components: Spatial--trip ends and path; Temporal--date, start and end times; Individual--person(s) making the trip; and Event attributes--trip purpose, travel mode. 53

OCR for page 24
There are numerous potential benefits of this approach, including the ability to support space- time queries. A complicating factor is that the respondent must enter the trip purpose into the recording device as there is no way to identify this information from the trip geography. Data needs, assumptions, and limitations. A critically important component of this analysis is sample selection. It is essential that you select, recruit, and fully brief a representative sample of the low-income population and minority population that potentially would be affected by the proposed project. Special attention may be required to prepare illiterate or non-English speaking participants. There are several steps involved in applying this method. Using a path-based approach, the GIS database should include addresses (or dummy equivalents) and an interactive address-matching process to identify common locations and expedite the coding process. A GPS receiver should be tied into the database, and a data-recording device is needed to develop a record of respondents' activity space. Using this basic approach, Shaw and Wang (2000) successfully matched 85 percent of more than 6,000 trip ends to the GIS streets database. An alternative approach is to use a cellular data transmitter to forward data obtained using a GPS receiver to a central data management facility. With this approach, it is possible to gather path- based data that give a very clear indication of respondents' travel patterns in time and space. As with the preceding methods, the assumption must be made that participants' activity patterns are representative of the population of interest. Another assumption is that participants will not alter their activity patterns as a result of taking part in the study. A limitation of this type of analysis is that it is static in nature. The travel patterns identified using this approach may not reflect the activity spaces that would develop if, over time, new employment centers, shopping centers, and other facilities were to emerge. Thus, it may not accurately predict the future state of things if the proposed transportation project were to move forward. Results and their presentation. While the database design may vary, space-time activity data should at the very least be able to answer three questions for a given study area: What are the vital individual daily activity centers? Where are frequently visited communal centers? How and when do members of the community access these centers? Because this method utilizes such sophisticated data collection and processing needs, it has considerable potential for providing you with a rich data repository able to satisfy a variety of analysis needs across a range of environmental justice-related issues. Assessment. Due to the sensitive and personal nature of the data collected, assurance of absolute anonymity is a prerequisite in order to gain the confidence of subjects. To ensure this, sample size must be sufficient and administration of the equipment adequate. If the data are collected heeding the points raised in this discussion, a very good sense of the activity space of the 54