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 119
Transportation Market Analyses Using ACS Data 119 the distribution of elderly people by county,77 an analysis which is commonly done within envi- ronmental justice analyses. A list of specific examples of using census data to do market analysis also is provided at the end of this section. 7.2 Benefits and Limitations of ACS for Transportation Market Analysis This section summarizes the perceived benefits and limitations of using ACS data for trans- portation market analysis. Transportation planners contacted about the implications of ACS felt that the availability of ACS data on a continuous basis would allow for more timely analysis of route and service planning, as well as environmental justice issues. In contrast, the variability of ACS data at small area geography may severely limit some of the applications. In many cases (e.g., environmental justice analysis, corridor analysis, etc.), planners seemed more interested in obtaining a firm quantitative assessment, even if it used point-in-time data rather than qualitative "moving average" data. Concerns were raised that in small areas where moving averages are used, some variables needed for many types of transit analyses--especially environmental justice analyses--would be hard to interpret. 7.3 Transportation Market Analysis: Environmental Justice Case Study The following case study illustrates how a data user might do transportation market analysis using ACS data. The case study provides a step-by-step description of how one might obtain the data, do the computations, and present the results. For this purpose, assume that you are a transportation analyst working in an MPO. In this analysis, your manager has asked you to perform an environmental justice analysis by comput- ing an index of dissimilarity (ID) for your area. Section 3 of this guidebook has detailed instruc- tions on downloading ACS data, and Section 4 and the previous case studies describe many of the generic ACS analyses that will be used in these analyses. You were asked to compute the ID, which often is used in planning to measure the evenness of sub-population groups across specific geographic areas,78 at different geographic levels for a target population consisting of non-Hispanic whites ("white" hereafter), non-Hispanic blacks ("black" hereafter) and Hispanic residents of Broward County, Florida. You have been asked to do the computations using both the 2000 ACS and census data. The goal of this analysis strategy is to ascertain the differential impact of moving from census to ACS data on this given measure of importance to public policy; transportation; and municipal, community, and regional planning personnel. In this case study exercise, begin by treating both the ACS estimates and the census counts as point estimates, without regard to the ACS sampling error. This will always be an option for analysts and, 77 Missouri Department of Transportation Socio-Economic Indicator Resource, "The Relationship of Environ- mental Justice Populations to Key Socio-Economic Indicators in the St. Louis Area District," 2003. See http://oseda.missouri.edu/modot/planning/stlouis_analysis.shtml. 78 NCHRP Report 532: Effective Methods for Environmental Justice Assessment. Transportation Research Board, National Research Council, Washington, D.C., 2004.
OCR for page 120
120 A Guidebook for Using American Community Survey Data for Transportation Planning for complex analyses, may be the only logical way to make use of ACS estimates. However, the Cen- sus Bureau recommends including the sampling uncertainty in any calculations using ACS esti- mates, so a discussion of including this uncertainty in the calculations follows the initial discussion. The ID measures the evenness with which two mutually exclusive groups are distributed across the geographic units that comprise a larger geographic entity. An example is the distribu- tion of blacks and whites across the census tracts that define a metropolitan region, county, or state. The key considerations are: 1) the total population of two particular groups at a higher- level geography (i.e., whites and blacks in Broward County) and 2) the proportion of each group's population within a particular areal unit. In this process, "areal unit" is used to describe smaller geographic areas within the larger study area. This can be a census tract, block group, or TAZ. The ID has a minimum value of zero and a maximum value of 100. It can be defined as ID = 100 × 1/2 · (| ai / A - bi /B |) where ai = the population of Group A within the ith areal unit, e.g., census tract; A = the total population of Group A within the larger geographic entity for which the ID is being calculated, e.g., county; bi = the population of Group B within the ith areal unit, e.g., census tract; and B = the total population of Group B within the larger geographic entity for which the ID is being calculated, e.g., county. The calculation of ID values involves specifying both the overall geographic unit of analysis as well as the areal units that form the basis for the index. Depending on the goals and needs of a project, this step is essential to achieving efficient and robust results. In this study, the ID will measure the overall race/ethnic residential composition at the county level for Broward County, Florida. The next decision of importance is the identification of the areal unit--the lowest level of geog- raphy where race/ethnic residential composition will be measured and then compared to the tar- get area (i.e., Broward County). Traditional analyses over the last half-century have focused on utilizing census tract-level data as the preferred areal unit of analysis. The use of tracts was prompted by convenience more than theoretical requirements. The average U.S. census tract in 2000 included approximately 4,200 residents and was neither initially developed nor intended to represent an average residential neighborhood. Unfortunately, very few generalizable data sources included units of analyses smaller than a census tract. This, combined with the relatively low level of computational resources needed to analyze a small number of areal units, resulted in the cen- sus tract becoming the standard reference areal unit in most geographic area analyses. As com- puter power has increased, many analysts prefer and can support analysis at the level of census block groups (about 1,300 residents) and census blocks (about 34 residents). Ideally, the smallest areal unit will "uncover" the truest picture of residential segregation in any given geographic unit. One can present the ID computed values at different geographic levels and using both ACS and census, as shown in Table 7.1. The lowest levels of race/ethnic segregation for all groups studied was at the PUMA level. This is to be expected because the PUMA represents the most aggregated data (aside from county level), with only 13 areal comparison units in the county. At the PUMA level, the following conclusions can be made: · The first column of values shows that, on average, roughly 35 percent (ID value at 35.1) of all blacks would have to move from their current PUMA residence in order to be proportionally
OCR for page 121
Transportation Market Analyses Using ACS Data 121 Table 7.1. Index of dissimilarity values for Broward County, Florida, by area, data source, and race/ethnicity. PUMA Tract TAZ Block Group Block W/B W/H B/H W/B W/H B/H W/B W/H B/H W/B W/H B/H W/B W/H B/H Data Source Census 2000 35.1 17.5 32.9 62.7 31.5 53.5 66.0 35.0 56.6 64.5 33.2 54.8 70.4 41.5 60.2 ACS 1999-2001 35.3 18.0 33.0 63.0 32.5 55.5 66.2 36.9 58.7 -- -- -- -- -- -- Net Difference 0.2 0.5 0.1 0.3 1.1 2.0 0.2 1.9 2.1 -- -- -- -- -- -- % Difference 0.6 3.0 0.4 0.5 3.3 3.7 0.3 5.1 3.6 -- -- -- -- -- -- N 13 279 777 689 20,136 *Source: Census 2000 & American Community Survey, 1999-2001 (weighted) Broward County, Florida. *Note: W/B = White/Black, W/H = White/Hispanic, B/H = Black/Hispanic. distributed within all PUMAs in the county in the same way as whites. The converse relation- ship holds as well. · The percentage of segregation between whites and Hispanics is roughly half the white/black value at 18. black/Hispanic segregation resides at approximately the same level as black/white difference, with ID value equaling 33. At this level of analysis, the percentage differences between the census and ACS are relatively small, with the highest value being within white/Hispanic residents at 3 percent. Moving to higher numbers of smaller population area units creates a resultant increase in ID values. This is to be expected as larger areal units can, and often do, mask greater race/ethnic neigh- borhood segregation. For the 279 tracts in the county, the following conclusions can be made: · Both higher ID values and greater differences between the census and ACS are noted, particu- larly for the black/Hispanic comparison. ID values have approximately doubled for all groups. · White/black ID value equals 63, thereby indicating a much higher residential segregation than previously detailed. Minority group data comparisons show about 3.7 percent difference between ACS and Census 2000. At the TAZ level (777 TAZs in the county), the following conclusions can be made: · There is only a slight increase in overall ID values and percentage differences by data as compared to tract data. · The highest percentage difference between ACS and Census 2000 detailed is among the white/Hispanic comparison, at 5.1 percent. 7.3.1 Available Data Selection of Geography For this case study, ID values were calculated for several subcounty areal units available in Census 2000 and ACS data. These include · PUMA--a geographic area with a minimum population of 100,000 residents. From within these PUMA areas, the census extracts a 5 percent sample of census Long Form data for pub- lic use. In Census 2000, there were 13 PUMAs within Broward County; · Census tract; · TAZ; · Census block group; and · Census block.
OCR for page 122
122 A Guidebook for Using American Community Survey Data for Transportation Planning Table 7.2. Areal unit characteristics, Census 2000 Broward County, Florida. Areal Unit Average Population Average Area (Miles2) N County 1,623,018 1,320.0 1 PUMA 124,848 101.1 13 Tract 5,817 37.5 279 TAZ 2,089 13.5 777 Block Group 2,356 15.2 689 Block 81 0.5 20,136 *Source: Census 2000. The Census 2000 data could be used at all of the listed geographies, but the ACS estimates do not contain information at the block group or block levels. Therefore, data comparisons for this study will concentrate on the levels that are possible in both datasets: PUMA, TAZ, and tract.79 However, data at the block group and block levels will be presented in order to assess the impact of how much detail may be lost due to the lack of small area geography in the ACS. Table 7.2 summarizes the characteristics of the areal units for this case study. Identification of Required Data Tables The analysis relied on custom Census 2000 and 1999-2001 ACS data tables provided by the Census Bureau in the Statistical Analysis System (SAS). ACS data with similar levels of geographic detail will be available for full ACS implementa- tion. Given the race/ethnic groups under study, the following tables in both the ACS and census custom tables were identified as relevant for this case study: · Total Hispanic population, · Total non-Hispanic white population, and · Total non-Hispanic black population. If accessing the census data using American FactFinder, the appropriate tables would be as follows: · ACS--P003: Hispanic or Latino by race; and · Census (Summary File 1)--P4: Hispanic or Latino, and not Hispanic or Latino by race. Within each table, all of the included geographic variables and their associated documenta- tion should be retained. This includes Sumlev, State, County, Place2, TAZ, Tract, and Puma5. 7.3.2 Analysis Steps The ID measures the evenness with which two mutually exclusive groups are distributed across the geographic units that comprise a larger geographic entity. As will be detailed later in this case study, the values for each areal unit are transformed into absolute values. They are then summed for the specific areal units under study (e.g., census tracts or block groups). The result- ant statistic (the ID) theoretically indicates the extent to which the sub-population groups of interest are evenly distributed and, if not, identifies the proportion of either group that would 79 It should be noted that the ACS data used in this analysis are custom tables prepared from 1999-2001 average data, where the sampling rate has been increased so that the three-year data would possess the level of accuracy obtained from five-year data under normal sampling rates. Thus, when analysts compute the ID at the tract or TAZ level, they should be aware that multiple-year data are needed.
OCR for page 123
Transportation Market Analyses Using ACS Data 123 have to move from their current areal unit to another in order for both groups to be proportion- ally distributed within the larger geographic unit of interest. The following basic steps undertaken to calculate an ID for Broward County are: 1. Following the index formula of ID = 100 × 1/2 (| ai /A bi /B |), calculate the necessary param- eters (columns) for formula completion. This is done for each group pair (white/black, white/Hispanic, and black/Hispanic). a) For the two groups to be compared, calculate the proportion of each group within each areal unit i you have chosen to utilize (tracts, TAZs, etc.). Every row in the data file is a particular level of i, or areal unit. Divide each a or b value within the ith row by the corre- sponding total population (A or b) for the geographic unit you are studying (the county in this case). b) Subtract a/A from b/B within the ith row. c) Take the absolute value of Step 1b. d) Sum the column values from Step 1c for the ith rows that make up areal units you are uti- lizing (i.e., all tracts, TAZs, etc.) e) Multiply summed values from Step 1d by 0.5. 2. The summed result multiplied by 100 is the ID value for Broward County. Table 7.3 is a worksheet that shows the ACS data used in the Broward County ID calculations involving the comparison of non-Hispanic whites (Groups a and A) and non-Hispanic blacks (groups b and B). Column and row letters/numbers are shown in the table to help guide the reader through the calculation steps. Within this table, data rows consist of population totals and counts by the 13 PUMAs. The columns represent summary variables such as geography (B = PUMA5), and race/ethnicity (C = white, D = black). Note that · Row 1 represents the county population totals for both whites and blacks. Table 7.3. ID calculations utilizing five percent of PUMA for non-Hispanic whites and non-Hispanic blacks in Broward County, Florida. A B C D E F G H 0.5*sum SUMLEV PUMA5 WHITE BLACK wh/WH blk/BLK |a/A-b/B| (|a/A-b/B|) 1 50 929090 337925 PUMA ID 2 795 03601 94565 12415 0.10 0.04 0.07 35.3 3 795 03602 74905 13555 0.08 0.04 0.04 4 795 03603 72855 21185 0.08 0.06 0.02 5 795 03604 74620 18940 0.08 0.06 0.02 6 795 03605 75560 44945 0.08 0.13 0.05 7 795 03606 85775 52435 0.09 0.16 0.06 8 795 03607 31270 59930 0.03 0.18 0.14 9 795 03608 89740 25955 0.10 0.08 0.02 10 795 03609 73680 11155 0.08 0.03 0.05 11 795 03610 90910 4095 0.10 0.01 0.09 12 795 03611 72195 13735 0.08 0.04 0.04 13 795 03612 23150 40585 0.02 0.12 0.10 14 795 03613 69865 18990 0.08 0.06 0.02 *Data source: American Community Survey: 1999-2001.
OCR for page 124
124 A Guidebook for Using American Community Survey Data for Transportation Planning · Rows 2 through 14 show the population counts for each group within the 13 Broward County PUMAs (This example assumed the calculation of ID at the PUMA level). · Column A shows the summary level of the data. SUMLEV = 50 is the row with county data, SUMLEV = 795 identifies the rows with PUMA-level data. · Column B represents the unique identifier of each PUMA in Broward County, Florida. · Beginning with the data in Rows 2 through 14, Columns E, F, and G show the construction of the PUMA-level components of the ID. c) Column E shows the outcome of dividing the white population in each PUMA by the total white population in Broward County (for example, E2 is the result of dividing C2 by C1). d) Column F repeats this exact same calculation for the black population (for example, F2 = D2/D1). e) Column G calculates the absolute value of the difference between Columns E and F (|E2- F2| - with rounding). f) Lastly, Column H, Row 2 sums the values in column G and then multiplies that value by 0.5. This is the ID value for this areal unit (PUMA) for the selected geography (Broward County, Florida). The process outlined above was repeated for the white/Hispanic and black/Hispanic groups in order to develop an overall ID for the region. The results are shown in Table 7.1, discussed previously. In this table · The columns show the different areal units (PUMA, tract, TAZ, block group, and block). · The subcolumns show the ID for each comparison group pair (W/B is white/black, W/H is white/Hispanic, and B/H is black/Hispanic). · The first two rows (Data Source) show the ID for each comparison group pair calculated first using the Census 2000 data, then using the ACS data. Given data availability, there is no block group or block data available through the ACS, so those cells are blank. · The next two rows (Net Difference and % Difference) are calculated differences in magnitude between the ID calculated using Census 2000 data versus that calculated using ACS data for the same geographic unit and comparison group pairing. · The final row provides the number of areal units used in the calculations for each geographic unit. There are substantial differences between the results of the ID calculations at the PUMA level and those at the tract level, indicating that the PUMA level of geography is probably too large to do meaningful environmental justice analyses. However, there are only moderate differences between the tract-level results and the TAZ-level results, and the TAZ-level results are quite similar to the block group and block-level analyses. This would indicate that, at least for Broward County, analyses at the tract or TAZ level are probably adequately capturing the over- all level of dissimilarity in the county. Since the block-level analyses do not vary by a large amount from the analyses for which ACS data will be available, such analyses using ACS data (and supplemented with decennial census Short Form data) would appear to be valuable and useful. Accounting for the Uncertainty of ACS Estimates Analysts can reflect the ACS sampling in the ID calculations by applying the equations that have been presented above. The following dis- cussion describes these calculations for the PUMA level of geography. The actual reported ACS estimates would include lower and upper bounds such as the rep- resentative figures shown in the columns on the left of Tables 7.4 and 7.5. The 90 percent con- fidence level margin of error of the estimates can be calculated by finding the differences between these estimates and the lower and upper bounds. The standard errors of the estimates can then be calculated by dividing the margins of error by 1.65, as has been previously shown. The margin of error and standard error calculations are shown in the middle columns of the two tables.
OCR for page 125
Transportation Market Analyses Using ACS Data 125 Table 7.4. Estimates of non-Hispanic white population of Broward County. PUMA Estimate LB UB ME SE PUMA a/A SE(a/A) County Total 929090 921936 936244 7154 4336 3601 94565 89487 99643 5078 3078 3601 0.102 0.0033 3602 74905 72321 77489 2584 1566 3602 0.081 0.0017 3603 72855 69737 75973 3118 1890 3603 0.078 0.0021 3604 74620 72075 77165 2545 1542 3604 0.080 0.0017 3605 75560 73271 77849 2289 1387 3605 0.081 0.0015 3606 85775 82327 89223 3448 2090 3606 0.092 0.0023 3607 31270 29853 32687 1417 859 3607 0.034 0.0009 3608 89740 85459 94021 4281 2595 3608 0.097 0.0028 3609 73680 70409 76951 3271 1982 3609 0.079 0.0022 3610 90910 90210 91610 700 424 3610 0.098 0.0006 3611 72195 69560 74830 2635 1597 3611 0.078 0.0018 3612 23150 22458 23842 692 419 3612 0.025 0.0005 3613 69865 67070 72660 2795 1694 3613 0.075 0.0019 The next step in the previous point estimate analysis was to develop the (ai/A) and (bi/B) fac- tors. The formula for calculating the standard error of this type of ratio was provided previously in Section 4. ^ ^2 X SE( ) = ^ Y 1 ^ Y SE X ^ Y ( ) ^ 2 + X SE(Y 2 ^ ) 2 The results of applying this calculation for the estimates of (ai/A) and (bi/B) are shown in the rightmost column of Tables 7.4 and 7.5. The next step in the ID calculation is to calculate the differences between (ai/A) and (bi/B), and then to sum these differences. Table 7.6 shows the results of performing these steps and calculating the corresponding standard errors. Table 7.5. Estimates of non-Hispanic black population of Broward County. PUMA Estimate LB UB ME SE PUMA b/B SE(b/B) County Total 337925 335086 340764 2839 1721 3601 12415 12039 12791 376 228 3601 0.037 0.0007 3602 13555 13010 14100 545 330 3602 0.040 0.0010 3603 21185 20225 22145 960 582 3603 0.063 0.0018 3604 18940 18037 19843 903 547 3604 0.056 0.0016 3605 44945 42949 46941 1996 1210 3605 0.133 0.0036 3606 52435 52031 52839 404 245 3606 0.155 0.0011 3607 59930 57743 62117 2187 1325 3607 0.177 0.0040 3608 25955 25179 26731 776 470 3608 0.077 0.0014 3609 11155 10709 11601 446 270 3609 0.033 0.0008 3610 4095 3875 4315 220 133 3610 0.012 0.0004 3611 13735 13261 14209 474 287 3611 0.041 0.0009 3612 40585 38848 42322 1737 1053 3612 0.120 0.0032 3613 18990 18342 19638 648 393 3613 0.056 0.0012