Subcommittee on Census Data
For Transportation Planning
Issues to Consider in Determining
Content of 2000 CTPP
Prepared by Phil Salopek, Elaine Murakami and Tom Mank
July 10, 1998 (revised 7/27/98)
Disclosure Avoidance and Average Cell Size
The Census Bureau Disclosure Review Board (DRB) will have a role in evaluating
the tabulation requests for the CTPP. This was not the case in 1990. The
most important issues for us are:
For example, if a TAZ has 400 residents and 200 workers, a table with 40
cells, which counted workers, would have an average cell size of 5. The
DRB has NOT determined a minimum average cell size for table definition.
If you review the tables from 1990, you will find that some CTPP tables
had over 200 cells. Unless the counts of residents and workers typically
exceed something like 800 or 1000, we will not be able to request a table
with 200 cells. We believe that the tables can be defined with fewer cells,
without losing too much ability to analyze the data for specific geographic
areas. One variable which we know is likely to cause problems is "means
of transportation" because in most areas, modes such as subway, bicycle,
taxi, ferry, are likely to be "0" (see item 12 below).
1) the population (residents and workers) in each TAZ, and
2) the average cell size in a table.
Phil Salopek and his staff will be making a presentation to the DRB
on the CTPP in the near future. They are currently tabulating TAZ resident
and worker populations from the 1990 CTPP for about 25 metropolitan areas.
1). Delete Parts D, E, and F from the Statewide Element.
These parts contained more detailed tabulations than in Parts A, B, and
C. The size cutoff for counties and places was 75,000+ persons in D-F,
not 2,500 as in A-C. The D-F data were tabulated by place of residence,
place of work, and residence by work, respectively.
2). Delete Part 4 from the Urban Element. This was a single
table consisting of a cross-tab of Household Size by Units in Structure
by Vehicles Available by Household Income. It was shown for the CTPP region
as a whole, the Study Area(s), Urbanized Area(s), MSA(s)/CMSA(s), and PMSA(s).
3). Delete Part 7 from the Urban Element. This was a set
of 31 tables containing data on the characteristics of people working in
census tracts in selected places (those with sufficient tract-coding rates).
Part 7 was always tract-based, even in areas that used TAZs in the rest
of the CTPP. Other geographic levels shown were US, State, county, place,
and metropolitan areas.
After 1990 CTPP, Census Bureau did develop special tract-to-tract tabulations
which were very popular. We may want to consider this type of tabulation
as part of the CTPP.
4). Delete Part 6 from the Urban Element.
This set of eight tables was produced only for MPOs whose regions contained
a million persons or more. Part 6 contained origin by destination data
for super-districts, which were aggregates of 100,000 or so persons built
up from the smallest units (TAZ, tract, or BG). Other geographic units
included the CTPP region, study area, MSA/CMSA, and PMSAs.
5). Delete Part 8 from the Urban Element.
These tables were produced only for MPOs whose regions contained a million
persons or more. Part 8 contained origin by destination data for TAZs,
tracts, or BGs and was requested by the "modelers." There was basically
a table of means of transportation by household income and one crossing
mode by vehicles available. Other summary levels included CTPP region,
study area, and central business district.
Regarding Items 1 -5. Since we are strongly considering the deletions
of these several parts, we believe that a custom tabulation package that
will allow 3- and 4-way tabulations will better serve our needs. The geographic
unit would be larger than a TAZ, and perhaps even larger than a tract.
Historically PUMS has permitted these types of tabulations using a sample
of the long form responses (1% or 5% samples), it may be possible to develop
a custom tabulation package for the complete sample of long form responses.
6). In general, decrease the number of cells
in the largest tables. Disclosure avoidance procedures may dictate
an average cell size.
7). Increase, in general, the number of tables
with data for Race/Hispanic Origin groups. Also, reach an agreement
on the Race/Hispanic categories to be used, and decide if any multiple
race breakdowns are wanted.
8). Decrease, in general, the number of tables
with data by sex.
9). Decrease the number of income and earnings
categories used. Especially when crossed by other variables with
large numbers of categories.
10). Decrease the number of departure time categories
11). Alter the disability tables, based on the
changes in these questions on the 2000 form. In addition, look at decreasing
the number of age categories used in the cross of disability.
12). It may be worthwhile to
divide the means of transportation into two mode choice groups:
a) summarized list for selected
areas having little public transportation (drive alone/car pool/public
b) a detailed list for those
with multi-faceted public transportation choices (alone/carpool /bus/trolley/taxi/railroad,
Summarizing small response categories may aid in avoiding situations
where data would be suppressed. Consolidating means of transportation into
fewer categories would diminish "0" counts, thus increasing the average
cell size (which may turn out to be a suppression criterion). FTA could
provide a list of areas where multiple modes are possible. MPOs would have
to ask to be put on the list if not already there.
Summarized List (12 Detailed
Drove alone Drove alone
Carpool 2 Carpool
Carpool 3 Carpool
Carpool 4+ Carpool
Transit (all) Carpool
5 or 6
Taxi Carpool 7+
Motorcycle Bus/trolley bus
Other means Railroad
Worked at home Ferryboat
Worked at home
13). Are the data needed for standard transportation
models used by all the states included? We know some states (e.g.,
Florida) have been looking at making some changes. Here are the tables
1-13 Household size (5) by Number of Workers in Household (6)
1-14 Number of Workers in Household (6) by Household Income (26)
1-17 Household Size (5) by Vehicles Available (9)
1-18 Number of Workers in Household (6) by Vehicles Available (9)
1-20 Vehicles Available (9) by Household Income (26)
14). The "Hours worked last week" item is no
longer on the census questionnaire for 2000. Should the tables
which utilized this data (Part 2/B table 9, Part 1/A tables 28 and 32)
be deleted, or do we want similar tables using the "usual hours worked
per week, last year" item?
15). Continue to include the unweighted sample
counts of persons and housing units? These items are useful in evaluating
data quality (e.g., calculating standard errors) and it is felt we should
continue to provide them even if they are not widely used.
16). Potential new tables:
household earnings by means of transportation
time of arrival by occupation
time of arrival by industry
means of transportation by occupation
means of transportation by industry
vehicles available per worker
household income per worker by means of transportation
household income per worker by means of transportation
17). Ordering or sequencing of the tables within
each part. The logic used in 1990 is not apparent and should be changed
and/or explained and documented so things are easier to find. For example,
put all the control total tables up front, both pop and housing, instead
of putting all the housing tables, including the controls, together in
the back of Part A/1.
Send you comments to either Ed Christopher or Tom Mank
Back to Home Page