by
Charles L. Purvis, Metropolitan Transportation Commission, Oakland, California
| This resource paper was prepared for the Transportation Research Board Conference "Decennial Census Data for Transportation Planning" held in Irvine, California, March 13-16, 1994. The final version of this paper appears in the formal conference proceedings: Decennial Census Data for Transportation Planning: Conference Proceedings 4, National Academy Press, Washington, D.C., 1995, pp. 55-67. The full proceedings are available from the Transportation Research Board at: http://www.nas.edu/trb/about/publicat.html. |
REVIEW OF LITERATURE OVERVIEW
GETTING THE DATA OUT: DISSEMINATING THE 1990 CENSUS
USE OF CENSUS DATA IN METROPOLITAN TRANSPORTATION PLANNING
The purpose of this resource paper is to describe the use and application of decennial census data for
transportation planning purposes in large metropolitan areas in the United States. In particular, use of the 1980
Census Urban Transportation Planning Package (UTPP) and the 1990 Census Transportation Planning Package (CTPP) will
be discussed.
Large metropolitan areas are defined as regions with populations of one million or greater. Though this
conference makes a distinction between large metropolitan and small-to-medium-sized metropolitan areas, the uses
and applications of census data can be quite similar. While transportation problems such as pollution and traffic
congestion are typically an order of magnitude more severe in the larger metropolitan areas, this may or may not
lead to more immediate and sophisticated uses of census data. The prime distinctions between large versus
small-to-medium-sized metropolitan areas are probably staffing levels and staff proficiency in managing large data
sets such as the 1990 Census.
The use of decennial census data in transportation planning has been covered extensively in the transportation
research literature. The reader should specifically review three Special Reports issued by the
Transportation Research Board covering census/transportation conferences held in 1970 in Washington, D.C.
(1); the 1973 Albuquerque, New Mexico conference (2); and the 1984
Orlando, Florida conference (3). Also useful is the collection of articles in
Transportation Research Record 981, published in 1984 (4). The reader also can
refer to the FHWA publication entitled Transportation Planners' Guide to Using the 1980 Census
(5) and Case Studies - Applying the Urban Transportation Planning Package (UTPP) in
Transportation Modeling (6). An ITE informational report entitled Use of Census
Data in Transportation Planning includes sections on how census data has been used in transportation analysis
(7). These reports provide a general overview of the use of 1970 and 1980 Census data in
transportation planning.
The 1973 Albuquerque and the 1984 Orlando conferences were an integral component of the U.S. Bureau of the
Census' formal and informal efforts to determine the census content for the 1980 and 1990 Decennial Censuses.
Details on the Census Bureau's content determination efforts are described in a series of Content Determination
Reports, including a report on place-of-work and journey-to-work issues (8). This 1994
Irvine, California conference will be an important element of the content determination process for the Year 2000
Decennial Census.
Complementary to the literature on the use of census data in transportation planning are several reference works
on census trend data, the most popular being Pisarski's Commuting in America report published by the Eno
Foundation (9); the 1986 FHWA report: Journey-to-Work Trends (10); and a new report by FHWA: Journey-to-Work Trends in the United States and its Major
Metropolitan Areas: 1960-1990 (11). Also of interest is a 1992 report by Pisarski
analyzing results of the 1990 Census (using Summary Tape File 3A data) and the 1990 Nationwide Personal
Transportation Study (NPTS)(12).
Processing and disseminating the 1990 Census data was (and still is) a mammoth operation. For the 1990 Census,
the Bureau of the Census collected data from 92 million households in the United States at a cost of approximately
$25 per housing unit, for a total cost of $2.6 billion (13,14).
Approximately one-in-six, or 15 million households, were given a census "long form" to fill out. Given the amount
of data and the complexity of the data processing operations, the Census Bureau has staged the release of new
census data on an almost continuous basis since 1990.
This staged release of census products has aided metropolitan transportation planners by effectively
distributing the workload over a period of years. Census data products are like a giant jigsaw puzzle with new
pieces added over time until the "picture" is finally complete. Had the opposite been true,with census data dumped
all at once on eager clients, the rush to get the "big picture" probably would have thwarted efforts to carefully
review results at a greater level of detail.
One of the findings from the 1984 Orlando conference was the desire to have "staged" releases of census
journey-to-work data. Many metropolitan transportation planners had to wait until 1983 or 1984 data to get basic
data on 1980 Census county-to-county commute patterns. The 1984 conference said "Get us county-to-county data as
soon as possible; get us the zone-to-zone, or tract-to-tract data after that." In response to these concerns and
other data user comments, the Census Bureau, the U.S. Department of Transportation and some volunteer
transportation professionals devised a "split" package scheme for disseminating 1990 Census journey-to-work results
- the Census Transportation Planning Package/Statewide Element (CTPP/SE), containing place-to-place and
county-to-county commuter flow data as well as place-of-residence and place-of-work tables; and the Census
Transportation Planning Package/Urban Element (CTPP/UE), containing zone-to-zone or tract-to-tract data (and
zone-of-residence and zone-of-work tables). In addition to the CTPP/SE and the CTPP/UE packages, the Bureau of the
Census developed a new product, the Summary Tape File S-5, which included 1990 Census county-to-county commuter
flows (without stratification by means of transportation.)
Other standard census products were an important component of metropolitan planning organizations' census
analysis plans. These products included the 100 percent count data in the "redistricting" tape and Summary Tape
File 1A, as well as the sample data in the Summary Tape File 3A and the Public Use Microdata Sample (PUMS).
By law, the Bureau of the Census must provide total population counts by state to the President of the United
States for purposes of apportionment of the House of Representatives by December 31 of each census year. In January
1991, the Census Bureau released place, county and state total population counts as part of its' Thank You
America count program. This was followed in March 1991 with the release of the Public Law 94-171 tape. The
PL94-171 "redistricting" tape provided block-level population characteristics by race and ethnicity and for persons
of voting age (18+). Within months, the rest of the 100 percent count items included in the 1990 Census were
released in the Summary Tape File 1A (STF1A) data sets.
The most significant release of census data in 1992 was the first "long form," or sample data, included in the
much awaited Summary Tape File 3A (STF3A). The STF3A tape file included small area (block-group) data on all sample
"long form" data: means of transportation to work, commute vehicle occupancy, average commute time, intracounty
versus intercounty commuting, household vehicle availability, household income, and number of employed residents.
The release of STF3A was a benchmark for census analysts, a cause for celebration as well as a call for
consternation. Carpool shares went down with respect to 1980 Census values. Drive-alone shares went up. Transit and
walk shares declined. The share of workers working at home increased dramatically. Metropolitan transportation
planners were turned into "spin doctors" overnight trying to explain the 1980 to 1990 trends only a matter of hours
after receiving the data themselves. The savvy transportation planner quickly assembled trend data and came up with
logical answers for the inevitable question: what do the numbers mean? It was the Census Bureau's job to
disseminate the data files to the local clients, the metropolitan planning organizations. It was the MPO's duty to
analyze the data in terms of trends, highlights and missed and met expectations, and to articulate the reasons why
these trends were occurring. Census data could then be readily digested by the public, the policymakers and the
media.
In December 1992, the Census Bureau released Summary Tape File S-5. This popular data file included all
county-to-county worker flow data for the entire United States. No data on means of transportation was provided,
but the basic county-to-county commute "puzzle" was filled in with STF S-5.
The CTPP/SE packages soon followed in spring of 1993. By the fall of 1993 and early 1994, the CTPP/UE packages
were streaming into metropolitan planning organizations.
The major disadvantage of a March 1994 conference on the decennial census and transportation planning is the all
too brief time that metropolitan and state transportation planners have had to analyze the Census Transportation
Planning Package/Urban Element. Certain metropolitan areas may have received their CTPP/Urban Element packages as
early as October 1993. Other major metropolitan areas still may not have their Urban Element package. Most likely
less than half of the approximately 300 urban element packages are available now. On the other hand, all states and
metropolitan areas have had nearly a year to review results from the Census Transportation Planning
Package/Statewide Element.
Despite the prematurity of this March 1994 conference, the immediate concern is to consider the Census Bureau's
tight deadlines for determining content for the Year 2000 Census. This process, scheduled from 1993 to 1996, will
culminate in a national content test in 1996, with the final Year 2000 Census questions to be transmitted to
Congress in 1997. Usefulness of data tabulations in the CTPP/SE and the CTPP/UE, as well as specifications for year
2000 journey-to-work tabulations, may wait until CTPP data users have had sufficient time to fully explore and
analyze these new 1990 data sets. Recommendations for Year 2000 Census content cannot wait.
The following sections discuss various uses and applications of census data in metropolitan transportation
planning, including: trend analysis; travel demand model estimation, calibration, and validation; demographic and
land use allocation model estimation, calibration, and validation; census data and estimating small area employment
data; census data and household travel surveys; transit market analysis; miscellaneous transportation planning
applications; and non-transportation planning applications of the journey-to-work data.
The most common application of census data is for trend analysis. How have things changed and why have they
changed? How have growth rates changed over the decades? What are the emergent trends? Trend analyses afford an
excellent opportunity for detailed cross-sectional and cross-temporal review of the socio-demographic conditions
within and between metropolitan areas.
In contrast to trend analysis are the "area profile" analyses, in which all census data for a geographic area
are included in a series of printed tables. These "area profiles" are an extremely popular way of disseminating
census data, especially STF1A and STF3A data. Census analysts as part of the state data center (SDC) and regional
data center (RDC) programs use commonly available software packages such as SAS or other database software for
preparing these tabulations. Federal Highway Administration staff, working with MPO staffs, are currently preparing
program code to create "area profile" reports using data from the CTPP parts A, B, 1 and 2.
An important element of trend analysis is understanding the changes in census content over the decades. Common
questions, such as "What is the average commute trip duration for residents in your region in 1970" or "What was
the drive alone share in 1960 and 1970" or "How many ferry commuters resided in your region in 1980," can only be
answered with "the data does not exist because census takers did not ask the same question in earlier censuses." A
useful addition to any trend analysis report is a brief recap of census content changes over the analysis period.
Examples of trend analysis reports includes publications by the metropolitan planning organizations in Chicago,
the San Francisco Bay Area, Philadelphia, San Diego and Seattle. These reports are the best source for
understanding "within" region changes in commute patterns and socio-economic characteristics. In contrast, the
Journey-to-Work Trends report published by the FHWA provides the best information on "between" region
trend comparisons.
The Chicago Area Transportation Study (CATS) publishes a monthly, two-color, six-page newsletter,
Transportation Facts, which includes census trend information and other results from its household travel
surveys. CATS also recently published a report containing profiles for all Illinois counties on
transportation-related data from the STF3A and CTPP/SE (15).
The San Francisco Bay Area's Metropolitan Transportation Commission (MTC) has produced a series of working
papers describing county, place and "superdistrict" results based on the STF1A (16), STF3A
(17), STF S-5 (18,19), the CTPP/SE
(20) and the CTPP/UE (21). Trend analyses include
county-to-county commuters from 1960 to 1990; change in total population since 1860; change in households since
1940; and change in household vehicle availability since 1960. In addition, the MTC has released an "electronic
publication" (computer file on floppy diskette) which includes place-to-place workers, by detailed means of
transportation, comparing 1980 UTPP and 1990 CTPP/SE commuter flows (22). In order to
maximize the use and understanding of census data, MTC provides copies of census working papers to Bay Area public
and private libraries, as well as to interested public, professionals, and policymakers.
The Delaware Valley Regional Planning Commission (DVRPC) in the Philadelphia region has published a report
documenting county-to-county commuter flows, by means of transportation, comparing the 1970, 1980 and 1990
journey-to-work (23). The report includes useful "desire line" maps showing changes in
commuting patterns - within the Pennsylvania suburbs, within the New Jersey suburbs, commuting to Philadelphia,
"reverse" commuting from Philadelphia and inter-regional commuting.
The San Diego Association of Governments (SANDAG) produces a multi-color bi-monthly newsletter, SANDAG
INFO, which contains a number of multi-color graphics as well as tabular results.
The Puget Sound Regional Council (PSRC) in the Seattle region publishes a monthly data newsletter entitled
Puget Sound Trends. The PSRC, as the regional data center for the Seattle region, also provides "area
profile" reports in hard copy and computer format, and maps showing census tracts, census blocks and zip codes.
The aforementioned reports and products are just a sampling of the ways in which census data is processed and
disseminated by metropolitan planning organizations in the United States. These tabular and graphic reports are
excellent means to provide information to the clients and partners of the MPO. (The parallel "poster session" or
"applications show and tell" session scheduled for Monday evening hopefully will reveal other innovative and
creative ways of analyzing, displaying and disseminating census data.)
Travel Demand Model Estimation, Calibration and Validation
One of the most common uses of census journey-to-work data is in the field of travel demand forecasting. The
census not only serves the purpose of providing base-year benchmark socio-demographic information for use as input
into standard travel demand model simulations, but the journey-to-work commuter flow matrices can be adapted by the
transportation planner into an "observed" work trip table for aggregate validation of work trip distribution and
mode choice models.
The following working definitions are provided for the terms estimation, calibration and
validation. Also discussed are the terms aggregate and disaggregate. These are offered
as "working definitions" rather than as accepted fact, given their various and conflicting usage in the profession.
Estimation is the process of determining model coefficients and constants using statistical software
packages. Logit models, cross-classification models and regression models are estimated.
Calibration is the process of adjusting model coefficients and constants using manual (or mechanical)
procedures. The friction factors and k-factors in gravity models are calibrated. The modal constants in regression
and logit models are also calibrated (adjusted) to match "observed" choices. Oftentimes the terms
calibration and estimation are used interchangeably, generally leading to confusion in
communication between a group of two or more transportation planning professionals. The term validation
refers to the process of comparing model simulated choices to "observed" choices. Validation is typically a stage
in the model development process whereas calibration is the actual activity to achieve a "validated"
model. "Observed" choice databases are independent estimates of socio-demographic or travel behavior
characteristics. "Observed" databases include, for example, census data, traffic counts, transit on-board surveys
and household travel surveys.
Aggregate refers to survey or census records tabulated or analyzed at any level greater than the
original level of data collection, e.g., 1990 Census block level data is aggregate data as well as place or
county-level data. Most 1990 Census products, including the STF1A, STF3A and the CTPP/SE and CTPP/UE, are aggregate
data. Disaggregate refers to survey or census records maintained at the original level of data collection,
e.g., the household level or the person level. Household travel surveys collected and maintained by MPOs and state
DOTs are disaggregate data sets. The census Public Use Microdata Sample (PUMS) is a disaggregate data set of
individual census household and person records, even though the geographic identification is suppressed at the fine
level of geography (less than 100,000 population groupings).
This last point about the CTPP/UE being an aggregate data set and the PUMS being a disaggregate data set may be
confusing, given the very small geographic areas associated with the CTPP/UE in contrast to the very large
geographic areas associated with the PUMS. This is a critical distinction given that disaggregate choice models
cannot be estimated using the CTPP/UE since the analyst does not have information on each household's or
worker's characteristics and choices. Disaggregate choice models can, on the other hand, be estimated from PUMS
data given that the analyst does have full information on each household's and worker's characteristics and choices
(though not any detailed geographic characteristics).
Can models be estimated using the CTPP/UE data sets? Yes, aggregate gravity models can be calibrated
using zone-to-zone "observed" trip tables. Yes, aggregate mode choice models ("diversion curve" models)
can be calibrated using the same "observed" trip tables. Should travel demand models be estimated using
the CTPP/UE data sets? Aggregate models should be avoided when the analyst can develop
disaggregate models instead. (The reader should refer to transportation planning textbooks to hear the
arguments for and against disaggregate and aggregate demand models.) On the other hand, since all gravity models
are aggregate models, it is quite appropriate to use the CTPP/UE as a fallback data set to calibrate an aggregate,
home-based work person trip distribution model.
Demographic and auto ownership models, other than land use models, can be estimated and/or validated using census data. Examples of demographic models include:
Pearson (24) describes the estimation of aggregate household by household size and
households by vehicle available models using the 1980 UTPP. Purvis (25) discusses the
estimation of disaggregate household by workers in household and households by vehicles available models using the
1990 Census Public Use Microdata Sample. These two papers demonstrate the viability of using census data in
estimating disaggregate and aggregate demographic and auto ownership models for use in regional travel demand
forecasting systems.
Part 1 of the CTPP/UE contains numerous zone-of-residence cross-tabulations that will be invaluable for
aggregate validation of demographic and auto ownership models. For example, Table 1-13 includes a cross-tabulation
of workers in households (six categories) by persons in households (five categories) by zone or tract of residence
(26). If the transportation planner carries a household size segmentation through his or
her travel model set, this Table 1-13 provides excellent "observed" data on workers in households by household size
for validation at a zone, superzone, district, superdistrict, county and regional scale. (In fact, the CTPP/UE is
the only source of small area census data that includes the distribution of households by workers in households.
The STF3A file only has total employed residents by small area of residence, not differentiating between
workers-in-households versus workers-in-group quarters units.)
A commonly used market segmentation in travel demand model systems is households by household size and vehicles
available. Table 1-17 is the only small area census source for data on distribution of households by household size
by vehicles available. The analyst may use this table for the estimation of aggregate models for splitting
households by household size and/or vehicle availability level; or he/she may use this cross-tabulation for the
aggregate validation of these demographic models.
Trip generation models cannot be estimated using census data due to the total lack of information on trip
frequency per household or per worker. On the other hand, the census workers-at-work data can be adjusted and
factored to create "observed" home-based work person trip tables by means of transportation. Work trip generation
and trip distribution models can then be calibrated to match, or closely approximate, the observed work trip travel
patterns.
The 1980 and 1990 Census asked persons in the "long form" to indicate "At what location did this person work
[most of] last week" and "How did this person usually get to work last week." If the person was an employed
resident but was absent from work the last week of March 1980 or March 1990 due to sickness, vacation, labor
dispute, etc., then that worker would not have provided information on his/her usual means of commuting or usual
place of work. This is referred to as "weekly absenteeism." Any information on "within week" variation in commute
behavior, such as daily absenteeism or commuting one day a week via transit or carpooling, or commuting from
home-to-work in one mode (say, casual carpool) and commuting from work-to-home in yet another mode (say, public
transit), would not be accounted for in census journey-to-work data. No census information is available on
"moonlighting" - increasing the number of jobs held by an employed resident.
The census is not an origin-destination survey. The census does not ask "From whose home did this person usually
leave for work LAST WEEK?" This is the "travelling salesman" phenomenon in which the person could be away from
his/her real home on business and view a hotel or motel as a "home" during the census period. This is a cause for
amusing and illogical commuter flows, e.g., persons reporting walk commutes from San Francisco to Los Angeles, or
subway commuters "living" in Honolulu and "working" in New York City. Typically a metropolitan area will a have a
small fraction of workers making absurdly distant commutes. The recommendation is to laugh them off and put them
aside - there will always be unusual outliers in census (and survey) data sets that cannot be treated seriously in
transportation planning analysis.
Metropolitan transportation planners have developed several techniques for factoring journey-to-work commuter
matrices into observed home-based work trips. Mann describes procedures used for the Washington, D.C., metropolitan
area to convert the 1980 UTPP commuter matrices to observed work trip tables (27). These
Washington, D.C. procedures were implemented in the Puget Sound region as described by Deardorf and Schneider
(28). Kollo and Purvis describe the use of the San Francisco Bay Area 1981 household
travel survey in computing work trip rates per commuter to convert journey-to-work matrices to observed home-based
work trips (29,30). Walker discusses the Philadelphia region
procedures for conversion of 1980 UTPP commuter matrices in reference (31). The above
referenced 1980 UTPP adjustment procedures for the Washington, D.C., Seattle, San Francisco and Philadelphia
regions are based on a traditional definition of home-based work "person" trips which includes mechanized modes
(drive alone, carpool, transit passenger) but excludes non-mechanized modes (walk, bicycle, other). The resulting
home-based work trip rates range from 1.57 person trips per commuter in the Bay Area, to 1.78 person trips per
commuter in Philadelphia, and range from factors of 1.54 to convert drive alone commuters and 2.15 to convert
carpool commuters into observed home-based work carpool trips for the Washington metropolitan area.
Probably the most legitimate technique to convert the 1990 CTPP/UE commuter matrices into observed home-based
work trips is using work-trips-per-worker trip rates collected as part of regional household travel surveys.
Several metropolitan areas in the United States conducted household travel surveys in the 1989 to 1991 time period,
including, Los Angeles, San Francisco, Sacramento, Chicago, Boston, Minneapolis, Atlanta and San Antonio. Perhaps
even data from the Nationwide Personal Transportation Survey (NPTS) could be used for estimating work-trip
frequency per-worker trip rates for metropolitan areas without current travel survey information.
Multi-day household travel surveys would be an ideal source of information for adjusting and factoring census
journey-to-work commuter flows. The Bay Area MTC, for example, collected multiple weekday travel diaries from
nearly 1,500 households in the spring and fall of 1990. This type of data set could be used for analyzing daily
versus weekly absenteeism patterns, work trip mode switching during the week, and the different travel modes used
in the trips from home-to-work as well as from work-to-home.
The calibration and aggregate validation of home-based work-trip attraction models may be more problematic given
potential differences in independent estimates of total employment as compared to the CTPP/UE workers at
zone-of-work. The CTPP workers at zone-of-work, derived from Parts 2 and 3, excludes the "weekly absentees" and
"moonlighting." Weekly absenteeism (only by area-of-residence) can be estimated from the STF3A or the CTPP/UE Part
1 tables. Moonlighting rates can be estimated from local sources, such as household travel surveys, or from
national sources, such as the Current Population Survey (CPS) conducted by the Bureau of Labor Statistics and the
Bureau of the Census.
Other errors in the census workers-at-work data will include standard sampling error, geocoding errors,
allocation errors, and the use of "default" or "workers at-large" zones for communities or counties with incomplete
address coverage in the Census TIGER files. Ideally, the "default" or "workers-at-large" zone should be no more
than one to two percent of the region's commuters.
The CTPP/UE data is not equivalent to total employment. Ideally, the CTPP/UE workers at work should be
90 to 95 percent of the regional agency's independent estimates of total employment, i.e., total jobs in the
region. The recent study by the Delaware Valley Regional Planning Commission (23) used a
2.2 percent weekly absenteeism rate (derived from the 1990 Census) and the national multiple jobholding rate of 6.2
percent of employed residents holding multiple jobs (derived from the Current Population Survey). DVRPC used the
national moonlighting rates by industry sector, ranging from 4.7 percent for construction workers to 9.3 percent
for those working in governments. (DVRPC also used other factors to bring the CTPP more in line with independent
estimates derived from Dun and Bradstreet, and municipal tax records.)
Trip distribution models can be calibrated using the adjusted and factored "observed" home-based work
person trip tables and network levels-of-service files. This means calibrating the standard "friction factors" used
in aggregate gravity models using either highway travel times or some combined impedance data. Socio-economic
adjustment factors, or "k-factors" in transportation planning jargon, could also be used to adjust
county-to-county, or district-to-district model-simulated home-based work person trip flows to match or approximate
the "observed" trip patterns. The Seattle (28) and Philadelphia (31) reports provide more in-depth coverage on the use of the 1980 UTPP in work-trip distribution
model calibration.
Work-trip mode choice models cannot be estimated from census data. On the other hand, existing work-trip
mode choice models (estimated from disaggregate household travel survey data) can be calibrated and validated to
match or approximate CTPP-derived "observed" home-based work trips by means of transportation. The modal constants
in the model utility functions can be adjusted (calibrated) upwards or downwards to change the base-year model
simulation. These modal constants are typically calibrated on a county-to-county or district-to-district basis.
Travel assignment models can use census journey-to-work travel time data as an element of the traffic
assignment process. Walker describes the use of travel time data from the 1980 UTPP in analyzing New Jersey
counties in the DVRPC region (32). Walker's research is very germane in light of current
federal regulations on the Clean Air Act Amendments which relate to the use of "actual" or "observed" data in
calibrating travel models for use in developing mobile source emissions budgets. The census journey-to-work data
set can be an excellent source of data for the calibration and adjustment of speeds and travel times from traffic
assignments.
To summarize this section, metropolitan transportation planners have demonstrated the utility of census data in
the estimation, calibration and validation of regional travel demand model systems. One essential use of census
data is for benchmark, base-year socio-economic small area data used as input into travel model simulations.
Analysts have used census data in statistically estimating and validating demographic and auto ownership models,
work-trip generation and work-trip attraction models, work-trip distribution models, work-trip mode choice models,
and for validating the highway speed simulations in traffic assignments.
The 1990 Census journey-to-work data included in the CTPP is not a substitute for a comprehensive
household travel survey. While the census contains invaluable socio-demographic data that are imperative for use in
travel demand model systems, it does not have any information on work or non-work trip frequency, on non-work trip
distribution, or on non-work mode choice patterns. Transportation planners must not approach the CTPP data as the
sole source of data to develop and maintain adequate travel demand models. This may sound obvious to the majority
of metropolitan transportation planners in the United States, but sometimes the obvious needs to be said. The CTPP
is a useful, independent, secondary data set to augment the disaggregate household data sets that a successful
metropolitan planning organization needs to collect for the development of state-of-the-art or
state-of-the-practice travel demand model systems.
Land Use Allocation Model Estimation, Calibration, and Validation
Land use allocation models are used in metropolitan planning organizations in the United States and elsewhere
for distributing regional forecasts of employment and workers to districts (zones) within the metropolitan area.
Examples of these models are the DRAM/EMPAL system of models used in several metropolitan areas in the United
States; the POLIS model, used in the San Francisco Bay Area; and the MEPLAN model system, applied in various
Canadian, European and African metropolitan environments (33,34).
The written record on the use of U.S. Census journey-to-work data for calibrating and validating urban location
models is weak, though efforts are afoot to incorporate 1990 CTPP/UE commuter flow data as it becomes available.
Twenty years after Lee's Requiem for Large-Scale Models appeared in the Journal of the American
Institute of Planners, the American urban model building scene was somewhat reinvigorated by a federal clean air
act lawsuit in San Francisco and new federal air quality conformity regulations that "encourage" the use of formal
land use allocation models in regions with serious, severe, or extreme air quality non-attainment status, though
these models are "not specifically required" (35). As such, metropolitan planning
organizations are actively reassessing their land use model systems in order to meet the rules and requirements of
the 1990 federal Clean Air Act Amendments and the Intermodal Surface Transportation Efficiency Act of 1991.
Future work on building and applying urban location models is challenged by the increasing share of multiworker
households and their household location patterns; the increasing share of "footloose" industries and their
commercial and industrial location patterns; the increasing share of workers working at home and the whole issue of
telecommuting; the confounding issues of local zoning controls and NIMBYism (not in my backyard) in determining the
location of housing and jobs; and the increasing importance of community attributes (housing prices, crime,
schools, shopping) in determining a household's location choices. Given these challenges, can we accurately
simulate the metropolitan system? The CTPP data can function as a validation data set for urban location models,
but it cannot be substituted for a theoretically complete, consistent and practical system of urban location
models.
Census Data and Estimating Small Area Employment Data
As previously stated, the CTPP workers-at-work data is not equivalent to total employment or jobs. The CTPP
workers-at-work "universe" excludes workers absent from work during the "census reference week" and does not
account for second, or "moonlighting" jobs held by employed persons. However, after taking these two
characteristics into account, the CTPP can be a fairly good data source for small area employment data.
Many metropolitan planning organizations utilize employment record data from state employment security
departments or employment development departments. These are employment security files that states must submit to
the federal Department of Labor and include data on employment and unemployment statistics. There are problems,
however, with state employment security files: They are oftentimes difficult to acquire and require careful
negotiations with state agencies that may not be too cooperative in sharing this information; and they only include
covered wage and salary jobs, and typically exclude family and self-employed workers.
Other metropolitan planning organizations may conduct employer censuses as part of trip reduction programs or
ridesharing databases. These programs will probably exclude small employers of less than, say, 50 or 100 employees.
The best situation is to have two independent sources of employment: the CTPP adjusted for weekly absenteeism
and moonlighting, and employment security records adjusted for family and self-employed workers. Unfortunately the
numbers may be pitted against each other with in some cases the CTPP having the "right" number of jobs and in other
cases the employment records having the "right" number of jobs - or neither estimate is correct! The job of the
employment data analyst is to creatively adjust and reconcile the two competing estimates of small area employment.
Census Data and Household Travel Surveys
Small area census data is critical for use in the weighting and expansion of household travel surveys. Weighting
and expansion of survey data is needed to adjust for non-response biases and geographic biases that will occur as
part of any household travel surveying effort. For surveys conducted in the 1989 to 1991 time period, 1990 Census
data can be used directly in weighting and expanding household surveys. For surveys conducted mid-decade the
analyst must carefully adjust the census to account for changes in the number of households and household
composition. The analyst may even choose to reweight household survey conducted in the mid-1980s by interpolating
1980 and 1990 Census data values.
Survey analysts for the 1990 San Francisco Bay Area and the 1991 Los Angeles household travel surveys used
similar, complex weighting schemes. The Bay Area analysts used the 1990 Census STF3A data to weight the survey by
superdistrict of residence (34) by household size (1, 2, 3, 4, 5+) by vehicle availability (0, 1, 2, 3+) by tenure
(owner, renter) (36). The Los Angeles analysts also used the 1990 Census STF3A data,
expanding the survey by regional statistical area (49) by household size (1, 2, 3, 4, 5+) by vehicle availability
(0, 1, 2+) by structure type (single family, multi-family) (37). Further "validation" of
the sample expansion scheme is done by comparing the expanded survey to other census variables such as workers per
household, tenure, structure type, sex, age, ethnicity, etc. A Chicago study also used 1990 Census data in
weighting and expanding regional household travel surveys with an increased emphasis on correct expansion for low
response neighborhoods within larger weighting districts (38).
This use of census data in transit market analysis is discussed in the resource paper by Cervero. The role of
the metropolitan planning organization, it should be noted, is to provide the CTPP to the transit operator partners
within a region; host training sessions on use of census data, particularly the CTPP, in transit service analysis;
and generally help the transit operator meet their analysis requirements. Of special note are the Title VI Federal
Transit Administration requirements related to low-income, auto-free, and minority populations within the transit
operator service area.
Miscellaneous Transportation Planning Applications
Miscellaneous transportation planning applications of the census, including the 1980 UTPP and the 1990 CTPP,
(excluding transit planning and travel demand forecasting use), include:
Other transportation applications will crop up as the data is disseminated to potential data users, and applied
in ways we just cannot imagine.
Non-Transportation Planning Applications of the Journey-to-Work Data
This section discusses the non-transportation planning applications of the 1980 UTPP and the 1990 CTPP data.
Other innovative and clever applications of this data will appear as potential users and clients are made aware of
the availability and content of the 1990 CTPP.
The article by Hammel (39) provides a good introduction to the non-transportation
planning applications of the 1980 UTPP.
Census journey-to-work data provides detailed information on commuter flows and daytime population which can be
critical in disaster preparedness and disaster response planning. Census journey-to-work data was useful in
disaster response planning efforts after the October 17, 1989 Loma Prieta Earthquake in Northern California and the
January 17, 1994 Northridge Earthquake in Southern California. The reader should refer to Fulton (40) for a description of estimating daytime population using data from the census journey-to-work.
City planning applications of the Census Transportation Planning Package are numerous, including using the CTPP
data in support of revision of general plan circulation, bicycle, housing, land use, seismic safety, and public
safety elements; in understanding labor force characteristics of city resident workers; in understanding the
characteristics of workers working within the community; and in local employment development programs. The
information may be of interest to local policymakers who want to know "Who is commuting to my city and where do
they live?", "Who commutes through our city?" and "Where do our city residents work?"
The journey-to-work data can be used by residential real estate developers to understand the commuteshed for
residents of particular neighborhoods or communities. By knowing the current commuteshed of an area a developer can
then market a product to workers working within that commuteshed. For example, a developer may use the information
to determine the newspaper in which to advertise.
The journey-to-work data can be used by commercial real estate market analysts to determine optimal sites for
locating or relocating a firm, based on minimizing employee's commute times, or based on the characteristics of the
labor force currently working within, say, 30 minutes of a particular site. Another example is U.S. military base
planners who use journey-to-work data to understand commutesheds around existing or proposed military bases and the
STF3A data on housing prices within that commuteshed to determine site suitability.
The journey-to-work data can be used by radio stations to ascertain how many commuters are in private vehicles
during any hour of the day.
The journey-to-work data can be used in Federal Transit Administration-sponsored reverse commuting demonstration
programs to understand the current magnitude of inner city resident workers commuting to jobs in the suburbs. The
American Public Transit Association (APTA) has been actively involved in reverse commuting demonstration programs,
publishing a report entitled Access to Opportunity (41) and sponsoring a session
on this topic at the October, 1993 APTA annual meeting in New Orleans. The Urban Institute in Washington, D.C. and
other organizations have also been involved in reverse commuting demonstration programs in the country, including
Philadelphia, Baltimore, Milwaukee, Chicago, St. Louis, and Nashville (42-44).
This paper provides a discussion on the staged release of 1990 Census data and the use of census data in large
metropolitan planning organizations in the United States. The various transportation and non-transportation uses
and applications are discussed. One conclusion is that the decennial census is a major source of primary, small
area socio-demographic information that is critical for metropolitan transportation planning activities.
The census cannot provide the necessary disaggregate travel behavior information needed by metropolitan
transportation planners. The census is not a substitute for a well-conducted household travel survey, but the
census does provide critical data needed to adjust household travel surveys and to adjust independent estimates of
small area employment. Census journey-to-work data is appropriate for use as an independent, secondary data source
for the calibration and validation of regional work-trip generation, distribution and mode choice models.
Where do we want to be in 10 years, at the next conference on decennial census data and transportation planning?
Can we anticipate the inevitable changes in technology and society, and can we anticipate our data needs in the
year 2004? Will the oil wells run dry and will we all be commuting over a virtual reality network? Will there be
new "means of transportation" that should be included in the Year 2000 Census? Will we have traffic and travel
behavior monitoring systems in place that will render the census obsolete? It may be too obvious that we cannot
answer these questions in three days, let alone the next three years, but a conscious attempt by metropolitan
transportation planners is needed to anticipate the travel demands of society after the year 2000. How can the Year
2000 Decennial Census be improved to anticipate these demands?
1. Use of Census Data in Urban Transportation Planning. Special Report 121,
TRB, National Research Council, Washington, D.C., 1971.
2. Census Data and Urban Transportation Planning. Special Report 145, TRB, National
Research Council, Washington, D.C., 1974.
3. Proceedings of the National Conference on Decennial Census Data for Transportation
Planning, Special Report 206, TRB, National Research Council, Washington, D.C., 1985.
4. Census Data and Urban Transportation Planning in the 1980s, Transportation Research
Record 981, TRB, National Research Council, Washington, D.C., 1984.
5. Arthur B. Sosslau. Transportation Planners Guide to Using the 1980 Census. Federal
Highway Administration, Department of Transportation, January 1983.
6. Arthur B. Sosslau and Michael Clarke. Case Studies - Applying the Urban Transportation
Planning Package (UTPP) In Transportation Modeling. Federal Highway Administration, Department of
Transportation, January 1984.
7. Use of Census Data in Transportation Planning. ITE Publication No. IR-011B.
Institute of Transportation Engineers, Washington, D.C., 1987.
8. Content Determination Reports: Place of Work and Journey to Work: 1990 Census of
Population and Housing. Report 1990 CDR-4. Bureau of the Census, Department of Commerce, October 1989.
9. Alan E. Pisarski. Commuting in America: A National Report on Commuting Patterns and
Trends. Eno Foundation for Transportation, Inc., Westport, Connecticut, 1987.
10. Dwight Briggs, Alan Pisarski and James McDonnell. Journey-to-Work Trends Based on 1960,
1970 and 1980 Decennial Censuses. Federal Highway Administration, Department of Transportation, July 1986.
11. Michael A. Rossetti and Barbara S. Eversole. Journey-to-Work Trends in the United
States and its Major Metropolitan Areas, 1960-1990. Federal Highway Administration, Department of
Transportation, November 1993.
12. Alan E. Pisarski. New Perspectives in Commuting: Based on Early Data from the 1990
Decennial Census and the 1990 Nationwide Personal Transportation Study. Federal Highway Administration,
Department of Transportation, July 1992.
13. Decennial Census: 1990 Results Show Need for Fundamental Reform. Report
GAO/GGD-92-94. General Accounting Office, June 1992.
14. Planning the Decennial Census: Interim Report. Committee on National Statistics,
Commission on Behavioral and Social Sciences and Education, National Research Council, Washington, D.C., 1993.
15. Erik P. Berman, Ed J. Christopher, William H. Ma, Jay J. Nam and Matthew J. Rogus. 1990
Census Transportation Factors for Residents of Illinois by County. Chicago Area Transportation Study, Chicago,
Illinois, October 1993.
16. Charles L. Purvis. Bay Area Population Characteristics: 1990 Census: Working Paper
#1. Metropolitan Transportation Commission, Oakland, California, April 1992.
17. Charles L. Purvis. Bay Area Travel and Mobility Characteristics: 1990 Census: Working
Paper #2. Metropolitan Transportation Commission, Oakland, California, August 1992.
18. County-to-County Commute Patterns in the San Francisco Bay Area: 1990 Census: Working
Paper #3. Metropolitan Transportation Commission, Oakland, California, December 1992.
19. San Francisco Bay Area Interregional County-to-County Commute Patterns: 1990 Census:
Working Paper #4. Metropolitan Transportation Commission, Oakland, California, January 1993.
20. Charles L. Purvis. The Journey-to-Work in the San Francisco Bay Area: 1990 Census:
Census Transportation Planning Package (Statewide Element) Working Paper #5. Metropolitan Transportation
Commission, Oakland, California, April 1993.
21. Charles L. Purvis. Detailed Commute Characteristics in the San Francisco Bay Area:
Census Transportation Planning Package (Urban Element) Working Paper #7. Metropolitan Transportation
Commission, Oakland, California, March 1994.
22. Bay Area Place to Place Journey to Work Characteristics: 1980 - 1990: Electronic
Publication Documentation. Metropolitan Transportation Commission, Oakland, California, April 1993.
23. Journey-to-Work Trends in the Delaware Valley Region, 1970 - 1990. Delaware Valley
Regional Planning Commission, Philadelphia, Pennsylvania, June 1993.
24. David F. Pearson. Disaggregating zonal households by Size, Income and Auto
Ownership. Paper presented at the Third National Conference on Transportation Planning Methods Applications,
Dallas, Texas, April 1991.
25. Charles L. Purvis. Using the 1990 Census Public Use Microdata Sample (PUMS) to Estimate
Demographic and Auto Ownership Models. Paper presented at the 73rd Annual Meeting of the Transportation
Research Board, Washington, D.C., January 1994.
26. 1990 Census Transportation Planning Package: Urban Element - Parts 1, 2, and 3:
Technical Documentation for Summary Tape. Journey-to-Work and Migration Statistics Branch, Population
Division, Bureau of the Census, Department of Commerce, September 1993.
27. William W. Mann. Converting Census Journey-to-Work Data to MPO Trip Data.
In ITE Journal. February 1984.
28. Raymond G. Deardorf and Jerry B. Schneider. A Comparison of Census Journey-to-Work and
Model-Generated Transportation Data in the Puget Sound Region. In Transportation Research Record
1090, TRB, National Research Council, Washington, D.C., 1986, pp. 43-51.
29. Hanna P.H. Kollo and Charles L. Purvis. Regional Travel Forecasting Model System for
the San Francisco Bay Area. In Transportation Research Record 1220, TRB, National Research
Council, Washington, D.C., 1989, pp. 58-65.
30. Development of "Observed" Home-Based Work Person Trip Tables from 1980 Census Urban
Transportation Planning Package Data: Assorted Staff Memos (1984-1985). Metropolitan Transportation
Commission, Oakland, California, 1985.
31. W. Thomas Walker. Testing and Adjusting Regional Travel Simulation Models with 1980
Census Data. In Transportation Quarterly, Vol. 42, No. 1, January 1988, pp. 63-88.
32. W. Thomas Walker. Method to Synthesize a Full Matrix of Interdistrict Highway Travel
Times from Census Journey-to-Work Data. In Transportation Research Record 1236,TRB, National
Research Council, Washington, D.C., 1989, pp. 50-58.
33. Michael Batty. A Chronicle of Scientific Planning: The Anglo-American Modeling
Experience. In Journal of the American Planning Association,Vol. 60, No. 1, Winter 1994, pp.
7-16.
34. Michael Wegener. Operational Urban Models: State of the Art. In Journal
of the American Planning Association, Vol. 60, No. 1, Winter 1994, pp. 17-29.
35. Criteria and Procedures for Determining Conformity to State or Federal Implementation
Plans of Transportation Plans, Programs, and Projects Funded or Approved Under Title 23 U.S.C. or the Federal
Transit Act. 40 CFR Parts 51 and 93. Environmental Protection Agency, 1993, § 51.452(b)(1)(viii).
36. Charles L. Purvis. Sample Weighting and Expansion: Working Paper #2: 1990 MTC Travel
Survey. Metropolitan Transportation Commission, Oakland, California, June 1993.
37. Applied Management and Planning Group. 1991 Southern California Origin-Destination
Survey: Summary Findings. Southern California Association of Governments, Los Angeles, California, February
1993.
38. Jing Li, Ashish Sen, Siim Soot and Ed Christopher. Factoring Household Travel
Surveys. Paper presented at the 72nd Annual Meeting of the Transportation Research Board, Washington, D.C.,
January 1993.
39. Lawrence V. Hammel. Nontransportation Uses of the Urban Transportation Planning
Package. In Special Report 206, TRB, National Research Council, Washington, D.C.,1985, pp.
74-79.
40. Philip N. Fulton. Estimating the Daytime Population with the Urban Transportation
Planning Package. In Transportation Research Record 981, TRB, National Research Council,
Washington, D.C., 1984, pp. 25-27.
41. Access to Opportunity: A Study of Reverse Commuting Programs. American Public
Transit Association, Washington, D.C., September 1993.
42. Mark Alan Hughes and Julie E. Sternberg. The New Metropolitan Reality: Where the Rubber
Meets the Road in Antipoverty Policy. The Urban Institute, Washington, D.C., December 1992.
43. Planning Practice: Turnabout is Fair Play. In Planning, December
1993, pp. 17-22.
44. Penelope Lemov. The Impossible Commute. In Governing, June 1993,
pp. 32-35.
Converted to HTML by the author, Dec-14-1998. Last Reviewed: Dec-07-2004.