Use of Census Data in Transportation Planning:
San Francisco Bay Area Case Study

by
Charles L. Purvis, Metropolitan Transportation Commission, Oakland, California

This case study discusses the use of census data in transportation planning activities in the nine-county San Francisco Bay Area. From an institutional perspective, this covers the use of census data in the metropolitan planning organization, the Metropolitan Transportation Commission (MTC); the council of governments, the Association of Bay Area Governments (ABAG); state agencies such as the California Department of Transportation (CalTrans); the various transit operators in the region; and local county and city planning, public works, and congestion management agencies. The primary focus is on census activities at MTC and ABAG, with reference to other creative work underway at the county and transit operator level.

This study also serves as an update to the resource paper prepared by Purvis (1) for the 1994 Conference on Decennial Census Data for Transportation Planning. The Purvis resource paper examined the use of census data in several major metropolitan areas, including the Bay Area, using published reports covering the 1970s through 1994. This paper will focus on Bay Area applications, as well as noting new uses of census data in the region between 1994 and 1996.

In terms of census products used in the Bay Area, the case study will discuss the use and application of data from standard Census Bureau data such as the summary tape files (e.g., STF1A, STF3A), the Public Use Microdata Sample (PUMS); special Census Bureau / USDOT products such as the Census Transportation Planning Package (CTPP) including the Statewide Element (CTPP/SE) and the Urban Element (CTPP/UE); and special Census Bureau data files purchased by the MTC and the Santa Clara County Center for Urban Analysis.

Following this introduction, the paper will review the application of census data in several categories, including descriptive analysis; model estimation; market segmentation for travel forecasting systems; model validation; and miscellaneous transportation applications. The paper will conclude with a set of recommendations and expectations for the Year 2000 Decennial Census. This last section is essentially a strategic assessment of the decennial census in terms of the strengths and weaknesses of census data; and the opportunities for improvements and the threats of not getting what is needed to maintain analytical tools and databases for transportation planning.

DESCRIPTIVE ANALYSIS

In the context of this paper, the term "descriptive analysis" refers to the reports, working papers, summary data files, spreadsheets, maps, press releases, trend reports, newsletters, etc., related to the dissemination of information from the decennial census. In the Bay Area these analyses have traditionally been in the form of place-level or county-level "profile reports" (printing out all information from certain census files, say, STF1A, STF3A, or CTPP/SE); or in the form of "working papers" (providing more in-depth discussion and trend analysis of census data).

Information systems and information technology is a rapidly evolving field, and it is apparent that we are also evolving into a new era of data dissemination. While in the past the most common means of dissemination are the "hard copy" or "dead tree-and-ink" issuance of census data reports, the future (and current) nature of data dissemination involves online, perhaps even real-time provision of data needed in our planning activities.

A good example of the use of new information technology for data dissemination is a World Wide Web (WWW) page, developed by ABAG, allowing the user to pick any of the 140 some-odd places in the Bay Area, then pick any of a set number of topical reports, to pull in pre-developed "profile reports." This ABAG WWW page is demonstrated at http://www.abag.ca.gov/bayarea/census90/pickhtml.html .

This rather straightforward WWW interface to "pick place / pick topic / view data" may likely be superseded in the years to come with more elaborate "data-on-demand" query-and-display setups that provide the data analyst with exactly what data is needed for the analysis at hand.

The issue of access to computers and the Internet is in part addressed by the development of Public Access Networks, an example of which is the "PAN Islands" network hosted by government agencies such as MTC and ABAG, and sponsored by a consortium of private sector companies in the Silicon Valley (e.g., Smart Valley, Inc., Pacific Bell, 3COM, Yahoo, Surf-Watch, Arthur D. Little.) (http://www.svi.org/planning/PAN/ISLND/ ). The principal actors involved in the public access to the Internet arena are local public libraries, many of which receive funds from the federal Library Services and Construction Act. Public access to the Internet via the public library system is critical to ensure universal access to census information.

Many clients and patrons still prefer the paper-and-ink versions of census reports. In the Bay Area, this is fulfilled by a comprehensive set of profile reports developed by the regional data center for the Bay Area, ABAG; and a set of transportation-related working papers authored by the staff at the MTC. A list of these MTC 1990 Census working papers is provided in Table 1 and are included in the references (2 - 12).

TABLE 1
1990 Census Working Papers - Metropolitan Transportation Commission - 1992-1996

An alternative to the hard copy working papers disseminated by MTC are electronic versions of these working papers and spreadsheets, as well as special "electronic publications" (13 - 15), summarized in Table 2:

TABLE 2
1990 Census Electronic Publications - Metropolitan Transportation Commission-1992-1996

The MTC electronic publications #1 and #2 provide comparisons of place-to-place commuters using data from the 1980 Census Urban Transportation Planning Package (UTPP) and the 1990 Census Transportation Planning Package (CTPP). The electronic publication #3 is MTC's version of area-level profile reports using place-level data from the CTPP/SE. These place-level profiles were only issued in electronic format given a 13 page profile report for 140 places in the Bay Area, for a grand total of 1,820 pages of information included in 7.9 megabytes of data files.

The need and value of these types of descriptive census reports, electronic or otherwise, should not be underestimated. They serve a vital role in informing the public, the media, the policy-makers, and the planning professionals on many of the most relevant demographic and travel characteristics needed for informing public policy development. The census working papers have been some of the most well received products issued by the MTC, and have been of great assistance to the librarians and public information staff at MTC and throughout the Bay Area.

Looking towards the future, it is unlikely that MTC or others will completely abandon the practice of issuing major paper-and-ink reports on census data. It is highly probably, on the other hand, that the rapid changes in information technology will greatly enhance the ability of planners to electronically disseminate and exchange census and other planning databases with our clients and partners.

MODEL ESTIMATION

The previous resource paper by Purvis provides a fairly detailed discussion on the use, or potential misuse, of census data for estimating demographic and travel models. This section will report on new applications and insights for the 1994 to 1996 period.

It bears repeating that the best sources of data for the estimation of travel demand models are household travel surveys. Survey data is essential, disaggregate data that should provide the transportation planner with the necessary data for the estimation and calibration of demographic and travel demand models. Decennial census data can, at best, be used in the estimation of household-level models to predict workers-in-household or auto ownership levels (16), or perhaps in the estimation of other types of aggregate travel models. For model validation, on the other hand, decennial census data is a critical and invaluable database for the aggregate validation of various demographic and travel behavior models.

Two new examples of use of census data in model estimation are efforts at MTC in estimating aggregate, trip end mode share models; and work in progress at the Santa Clara County Center for Urban Analysis in estimating aggregate logit work destination choice models, stratified by household income quartile.

Trip End Mode Share Models for Predicting Bicycle and Walk Commuters

One set of new trip-end mode share models developed at MTC are aggregate regression models estimated on zone-level shares of bicycle and walk commuters, at the zone-of-residence and the zone-of-work (17). The models predict the percent of workers that commute via bicycle or walk modes based on aggregate zonal characteristics such as employment density, share of multi-family dwelling units of total units, local workers/job balance within the travel analysis zone, and dummy variables to reflect proximity to the university campuses in Stanford and Berkeley. Typical mode share models are estimated using zone-to-zone network levels-of-service and related demographic and land use characteristics. Typical mode share models are either logit in form or are of the "diversion curve" model style that were popular in the 1960s. The reason behind developing these atypical, aggregate, zone-level, "trip end" based models is the concern regarding too few sample walk and bicycle trips in the 1990 MTC household travel survey. Of the 18,300 sample total home-based work trips in the 1990 MTC household travel survey, only 478 sample trips (2.6%) were by walk mode and just 222 sample trips (1.2%) were by bicycle. Such a small and sparse dataset on walk and bicycle commuters may prove a challenge in the model estimation process. Final decision on whether to use aggregate trip end mode share models or a disaggregate trip interchange mode choice model depends on current work in progress at the MTC to estimate best practice nested work trip mode choice models.

Aggregate Work Destination Choice Models

Another example of travel model estimation using census data are ongoing efforts at the Santa Clara County Center for Urban Analysis (SCCCUA) to estimate aggregate logit work destination choice models. These logit destination choice models are estimated separately by household income quartile using data from a special census file purchased from the Census Bureau by MTC and SCCCUA. This special census file, denoted as STP-214 (Special Tabulation Product) by the Census Bureau, provides one cross-tabulation of block group-to-block group workers (within the nine-county Bay Area) stratified by 12 categories of means of transportation to work, by four household income levels (less than $25000; $25000 - $45000; $45000 - $75000; and greater than $75000 annual household income). MTC's analysis of this special tabulation product is included in Census Working Paper #11 (12). Work at the SCCCUA on these aggregate work destination choice models is continuing, and will be completed in 1996.

Comparisons of regional average and median trip length (in miles), and average and median commute duration (in minutes) is shown in Table 3. This table shows a notable increase in work commute duration and length with increasing household income levels, an indicator that supports the notion for an income-stratified work trip distribution (destination) choice model.

TABLE 3
Mean and Median Average Commute Length and Commute Duration by Household Income Quartile
1990 Census, Special Tabulation Product #214 - San Francisco Bay Area - Regional Totals

TRAVEL MODEL MARKET SEGMENTATION

One of the challenges in travel demand forecasting is the use of disaggregate travel demand models in the aggregate prediction of travel behavior. Methods of aggregation fall into three principal categories: the "naive" method, market segmentation, and sample enumeration (18). In a nutshell, the "naive" method assumes that everyone in a travel analysis zone has the same characteristics: for example, they have the same number of workers in the household, the same household income, and the same number of persons in the household. Zonal mean values are used exclusively in the "naive" method.

The market segmentation method assumes that there are distinct subgroups within each travel analysis zone, for example, households by auto ownership level, by workers in the household, and/or by household income level. In the market segmentation method, the analyst assumes that the average household size, the average workers per household, etc., is the same within each subgroup by each travel zone.

The third aggregation method, the sample enumeration method, does not use any group or subgroup mean of any input variable. Instead, the disaggregate model is applied at the disaggregate (i.e., household, person or trip) level, and then the predictions are aggregated for reporting purposes. The sample enumeration technique is also known as microsimulation, where the forecasting of travel or other activity behavior is made at the discrete individual level, rather than at the zone level.

At the MTC, use of market segmentation is a key feature of the aggregate forecasting system in place and under redevelopment (19). Also, sample enumeration is used at MTC for special analyses, such as evaluating the effectiveness of transportation control measures (20).

To apply disaggregate models in a market segmentation framework, analysts at MTC have used the 1980 and 1990 Census Public Use Microdata Sample (PUMS) databases as "supplementary inputs" to the model application process. Census PUMS data are critical because it provides information the analyst needs to adjust the input model parameters (e.g., household size, household income, percent multi-family) by the desired market segmentation. PUMS data is used because standard census products such as the STF3A or the CTPP do not provide the necessary data at the travel analysis zone level or any other geographic level.

MTC has a nested workers in household / auto ownership choice model (WHHAO). This model splits the households residing in a travel analysis zones into households by three workers in household levels (0, 1, 2+ workers/HH) by three auto ownership levels (0, 1, 2+ vehicles/HH). The input market segmentation to the WHHAO model is households by household income quartile. This means that the outputs of the WHHAO model application are the number of households in each travel zone stratified by household income (4) by workers in household (3) by auto ownership level (3), or four market segmentations into the model, thirty-six market segmentations coming out of the model application.

One of the input variables to the WHHAO model is average household size. Rather than using zonal average household size, the Census PUMS data is used to adjust zonal average household size to zonal average household size stratified by household income level. This is needed because low income households (less than $25,000 per year) are smaller in size than higher income households. PUMS data is used to develop county-level (or PUMA-level) adjustment factors. These county-level adjustment factors are then multiplied by the zonal average household size to yield zonal average household size by household income quartile. These county-level household size by income quartile and adjustment factors are summarized in Tables 4 and 5.

TABLE 4
Average Household Size by Household Income Quartile by Bay Area County
1990 Census Public Use Microdata Sample, 5% Sample

TABLE 5
Household Size Adjustment Factors by Household Income Quartile by Bay Area County
1990 Census Public Use Microdata Sample, 5% Sample

An example calculation follows: the North Beach travel analysis zone in San Francisco has an average household size of 1.719 persons per household. Using the San Francisco County adjustment factors, we estimate that the average low income (quartile #1) household in the North Beach has an average household size of 1.361 persons per household (1.719 * 0.792); the average medium-low income (quartile #2) household has an average household size of 1.709 (1.719 * 0.994); the average medium-high income household has an average household size of 2.020 (1.719 * 1.175); and the average high income household in the North Beach has an average household size of 2.228 persons per household (1.719 * 1.296). Thus, the subgroup mean household size in the North Beach ranges from 1.361 to 2.228 persons per household. The "naive" method would just use the zonal mean household size of 1.719.

These adjustment factors used in the market segmentation process could also be developed using data from local household travel surveys. The problem will be the reliability of these factors based on typically too few sample observations in the small-scale household travel survey. For example, the 1990 Bay Area household travel survey provides sample data on 10,800 households. The 1990 Census PUMS 5-percent sample includes disaggregate data on 108,500 Bay Area households. A valuable research project would be to calculate the standard errors of these adjustment factors comparing regional travel surveys, such as the 1990 Bay Area travel survey, to the 1990 Census PUMS-based adjustment factors.

DEMOGRAPHIC AND TRAVEL MODEL VALIDATION

One of the basic uses of census data is for the aggregate validation of demographic and certain travel demand models. Validation is the process of comparing predicted values to "observed" values, and making the necessary adjustments (calibrations) to each of the component models to produce a valid model simulation. The decennial census data serves a most valuable purpose as an independent, "observed" estimate of various demographic and travel behavior. The following is a list of various uses of census data in demographic/travel model validation:

1. Workers in household models can be validated against CTPP/UE data at the regional, county, superdistrict, district and travel analysis zone level.
2. Auto ownership level models can be validated against CTPP/UE data at the regional, county, superdistrict, district and travel analysis zone level.
3. Analysts can convert the zone-to-zone commuter matrices, derived from part 3 of the CTPP/UE, into "observed" home-based work person trips using trips-per-worker conversion factors. Home-based work trip generation (production) and trip attraction models can be validated against these CTPP-derived observed home-based work trip tables at the regional, county, superdistrict, district and travel analysis zone level.
   
4. Home-based work trip distribution models can be validated against CTPP-derived observed home-based work person trips. In the Bay Area, the census STP-214 file can be used to calibrate and validate work trip distribution models stratified by household income quartile. The validation process for work trip distribution models should include a careful evaluation of model versus observed average trip lengths (and average trip duration) at the regional, county, superdistrict and perhaps zonal level; evaluation of regional trip length, and trip duration frequency distributions; and evaluation of county-to-county, superdistrict-to-superdistrict, and district-to-district predicted versus observed home-based work person trips.
   
5. Home-based work mode choice models can also be validated against CTPP-derived observed home-based work person trip tables by means of transportation. Ideally, a multi-day household travel survey or multi-day employer survey is used to adjust the "usual means of transportation" typology into an "average means of transportation."

In terms of what the census cannot be used for, the decennial census is not used for the validation of non-work trip frequency, non-work trip destination, or non-work mode choice travel demand models. Given that non-work trips may typically encompass 75 percent of a large region's travel, it is imperative that a suitably-sized household travel survey be on hand for the aggregate validation of non-work travel demand models. Also, it is very useful to have two competing sets of "observed" home-based work data. This gives the analyst flexibility in what he or she should be validating against. If the two observed databases (census and survey) are in agreement over a certain statistic, then that provides the analyst a general indication of confidence, or lack of confidence, in a particular census-based or survey-based estimate of work travel behavior.

MISCELLANEOUS TRANSPORTATION APPLICATIONS

Several new applications of census data for miscellaneous transportation planning and transportation research activities in the Bay Area are worth reporting on. The first application of interest is the use of 1990 Census PUMS data as part of the Bay Bridge Congestion Pricing Demonstration Project. This demonstration project, funded by the FHWA, included an analysis of the demographic characteristics of Bay Bridge commuters. MTC consultants were able to extract Bay Bridge commuters based on PUMA (Public Use Microdata Area) of residence and county-of-work. The analysis enabled MTC to understand the income and modal usage characteristics of Bay Bridge commuters in 1990 (21). In other words, who would be affected by a toll increase during peak travel times?

Census PUMS data has also been used by MTC staff to produce a demographic profile of workers working-at-home in Marin County (22). Analysis focused on the industry, occupation, earnings, sex, age, years of schooling, and hours worked "last week." The "typical" work-at-home worker in Marin County is female, highly educated, works part-time, is self-employed, is older, and earns less than workers who commute outside the home. Male work-at-home worker earnings are 140 percent higher than female work-at-home worker earnings ($42,500 male versus $17,500 female). The predominant industry for work-at-home in Marin County is real estate, management and public relations, and professional and business services. The predominant occupation of work-at-home in Marin are writers-artists-entertainers, and managers-administrators. The work-at-home commute share in Marin ranges from a low of 1.4 percent for government workers to a high of 25.1 percent of self-employed Marinites.

The above examples show but just two of the many applications that the creative transportation planner/analyst can produce using the census PUMS datasets. It is frankly one of the best transportation research databases offered by the Bureau of the Census.

An example of transit applications using census data is recent MTC work with the Central Contra Costa Transit Authority (CCCTA) on a GIS-based analysis of transit dependent population in the CCCTA service area. One of the layers in MTC's GIS system is local bus stops and rail stations. The analyst then used the GIS system to create a buffer zone around each bus stop to represent areas within a certain walking distance of the bus stop. The GIS program then performs a "cookie cutter" technique to split out demographic data within and outside the buffer zone. Demographic variables such as zero auto households, population age 62-and-over, and non-working households, were used as measures of transit dependency. This technique can then provide the transit market analyst the demographic characteristics of all persons residing within walking distance of each of the CCCTA routes, as well as the characteristics of persons not within walking distance of any of the system routes. Similar GIS efforts are underway at other Bay Area transit operators, including BART, Golden Gate Transit, and SamTrans.

Academic researchers at the University of California at Berkeley have made significant contributions to the research literature on urban structure, commuting, residential choice and job location choice behavior. The most recent research on the San Francisco Bay Area is included in a working paper by Cervero and Wu (23). Cervero and Wu provide an analysis of the polycentric commuting patterns in the Bay Area using commuter flow data from the CTPP/UE and housing price data from the STF3A data files. They find an emerging hierarchy of employment centers ranging from the San Francisco CBD to outlying suburban business parks, as well as shorter commute times to suburban employment centers.

SUMMARY

As a summary of the San Francisco Bay Area case study, an evaluation is given of the degree to which census data was crucial or essential to MTC planning and research applications and of the possibility that these applications could have been produced using other databases. The paper concludes with a discussion of recommendations and expectations for Census 2000.

Evaluation of Importance of Census Data and Possibility of Substitution

Descriptive Analysis

Descriptive analyses such as the census working papers and profile reports discussed in this paper are best served by a national census including a sample, "long form" data as well as the 100-percent count "short form" data. A national survey could be conducted to replace the census long form, and perhaps the national survey could be tailored to each metropolitan area's needs. The problem with a national survey is that the costs would probably be higher than the decennial census (approximate cost of $25 per census long form), the sampling rates would probably be substantially lower, and the statistical variance and standard errors would be substantially higher. So, if higher unit costs and lower accuracy are acceptable, then the census long form could conceivably be replaced by a national survey. Given that one component of descriptive analysis is the trend analysis of demographic characteristics at the small area level (e.g., census tract or block group), a smaller size national survey could not be used for trend analysis for finer grains of geographies, say, below place level. This would be a critical loss for city, county and metropolitan area planners who depend on decennial census long form data for neighborhood-level demographic characteristics.

Demographic Benchmarking

Though not discussed in this paper, one of the primary uses of census data at the metropolitan and local area level is for the benchmarking of most demographic variables, including housing units, households, population, household income and auto ownership. Much of the most important benchmarking data is included in the decennial census long form. The census long form could be replaced by national or local demographic surveys, but probably at a higher unit cost, lower sampling rates, and higher statistical variance and standard errors. In addition, total overall costs of a national survey or sets of metropolitan area surveys could conceivably exceed the cost of conducting the decennial census long form survey.

Model Estimation

MTC and others have demonstrated the use of decennial census data in estimating demographic and travel demand models, including workers in household, auto ownership, aggregate work trip mode choice models, and aggregate work trip destination choice models. The best datasets for demographic and travel demand models, however, are still the locally-conducted household travel surveys, oftentimes conducted concurrently with the decennial census (e.g., MTC's household travel surveys in 1981 and 1990).

The utility of future decennial censuses for model estimation could be enhanced by developing a "contextual PUMS" program. This contextual PUMS program is where the Census Bureau hires researchers as "special sworn employees" in order to conduct research using raw, disaggregate census microdata records within the confines of a Census Bureau research station. An example of this is a research data center opened in January, 1994 in Boston to examine topics relevant to current economic issues (24).

Market Segmentation in Travel Models

The decennial census is the largest, most accurate database available to transportation planners for use in determining the demographic characteristics of subgroups of the population. Census PUMS data could be replaced by national or metropolitan area survey data, but at higher unit costs, lower sampling rates, and higher variance and standard errors. Research should be conducted to determine just how less accurate it is to use metropolitan area survey data as opposed to Census PUMS data for market segment adjustments.

Demographic and Travel Model Validation

It may sound redundant, but the decennial census is the best database that can be used for the aggregate validation of several demographic and travel behavior models, including: workers in household models, residential and job location choice models (land use allocation models), auto ownership level models, and work trip generation, distribution and mode choice models. The decennial census cannot be used for the aggregate validation of non-work travel behavior models, so the best source for aggregate validation of these models are metropolitan household travel surveys.

The census is used to represent "observed" conditions. Household travel surveys can also be used to represent "observed" conditions, but problems with the lumpiness and sparseness of typical metropolitan travel surveys renders it quite difficult to use surveys as an aggregate validation database at any fine level of geography, say, district or superdistrict. With the decennial census, aggregate validation can be performed at almost any geographical level, perhaps even down to the travel analysis zone (neighborhood) level.

It is conceivable that metropolitan areas in the United States return to the 5-percent sample surveys that were more typical in the 1950s and 1960s. A 5-percent sample survey of the 2.465 million households expected to reside in the Bay Area in the Year 2000 would be approximately 123,000 sampled households. At a current dollar cost of approximately $125 per household, this expanded metropolitan household travel survey could cost on the order of $15.4 million in current US dollars. (This is a substantially larger amount than the Bay Area spent on the 1990 household travel survey: $1.0 million.) A national set of metropolitan travel surveys of this size could very well cost more than any conceivable decennial census.

Transportation Research Applications

The transportation planning and research community have made considerable progress in using various census products in research efforts. PUMS data has been used in analyzing markets for congestion pricing, for work-at-home commuters, and for analyzing the demographics of transit users in the San Francisco Bay Area. The CTPP and other standard census products are used in studies on urban structure and economic development. Certainly the microdata records from household travel surveys provide the researcher with the most flexibility in any intended analysis. However, the decennial census data, especially the microdata files, affords the researcher the opportunity to delve into the demographic, household and commuter characteristics of the rare or hard-to-reach populations.

Recommendations & Expectations for Census 2000

Maintain Census Long Form

The census long form is critical to provide the accurate and precise data needed to support demographic analysis and transportation planning and research activities. The likely substitute, in absence of a census long form, would be a set of metropolitan travel surveys that would be more costly and less accurate than a properly conducted national census.

Increase Involvement of Metropolitan and Local Planners To Improve Quality of Workplace Geocoding

Accuracy of workplace geocoding is still an issue to be reckoned with. Improvements in geographic information technology will certainly help, as will a cooperative program between the Census Bureau and local persons knowledgeable of local conditions. Legal barriers that limit the involvement of local planning staffs to assist in census data processing should be liberalized to allow greater involvement of local census partners and stakeholders.

Develop Contextual PUMS Program To Facilitate Improved Planning Research

In order to increase the relevance of transportation planning research it is desirable to create a census microdata research program. This program would allow bona fide researchers the opportunity to add value to census microdata and prepare more in-depth research than would otherwise be doable.

Embrace Changes in Information Technology To Provide Better, More Relevant Data

Rapid changes in information systems and information technology should be dealt with in terms of Census Bureau plans to collect, analyze and disseminate decennial census data. New information technology should lessen the need for paper-and-ink publications in favor of electronic data-on-demand. Public access to the Internet should be a high priority to facilitate collection and dissemination of decennial census data.

Given these changes in information technology, it may not make sense to talk about improvements to the Census Transportation Planning Package for the Year 2000. Things may be changing to the degree that future analysts may get what-they-want, when-they-need-it, and how-they-asked-for-it. This type of chaotic flexibility will likely be a challenge in terms of data consistency and comparability, so the Census Bureau may need to prepare standard census products in order to facilitate a transition from the highly structured products of the past to the chaotic data-on-demand; products of the future.

REFERENCES

1. Purvis, Charles L. The Decennial Census and Transportation Planning: Planning for Large Metropolitan Areas. In Decennial Census Data for Transportation Planning: Conference Proceedings 4, TRB, National Research Council, Washington, D.C., 1995, pp. 55-67.

2. Bay Area Population Characteristics: 1990 Census: Working Paper #1. Metropolitan Transportation Commission, Oakland, Calif., April, 1992.

3. Bay Area Travel and Mobility Characteristics: 1990 Census: Working Paper #2. Metropolitan Transportation Commission, Oakland, Calif., August, 1992.

4. County-to-County Commute Patterns in the San Francisco Bay Area: 1990 Census: Working Paper #3. Metropolitan Transportation Commission, Oakland, Calif., December, 1992.

5. San Francisco Bay Area Interregional County-to-County Commute Patterns: 1990 Census: Working Paper #4. Metropolitan Transportation Commission, Oakland, Calif., January, 1993.

6. 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, Calif., April 1993.

7. Disability, Mobility Limitation, and Self-Care Limitation Status in the San Francisco Bay Area: 1990 Census: Summary Tape File 3A: Working Paper #6. Metropolitan Transportation Commission, Oakland, Calif., October, 1993.

8. Purvis, Charles L. Detailed Commute Characteristics in the San Francisco Bay Area: 1990 Census: Census Transportation Planning Package (Urban Element): Working Paper #7. Metropolitan Transportation Commission, Oakland, Calif., March, 1994.

9. Purvis, Charles L. Detailed Interregional Commute Characteristics: 1990 Census: Census Transportation Planning Package (Urban Element): Working Paper #8. Metropolitan Transportation Commission, Oakland, Calif., May, 1994.

10. San Francisco Bay Area: County & Regional Profiles: 1990 Census: Census Transportation Planning Package (Statewide Element): Working Paper #9. Metropolitan Transportation Commission, Oakland, Calif., September, 1994.

11. San Francisco Bay Area Detailed Household Characteristics: 1990 Census: Census Transportation Planning Package (Urban Element, Part 4): Working Paper #10. Metropolitan Transportation Commission, Oakland, Calif., April, 1995.

12. San Francisco Bay Area Commuters by Household Income Characteristics: 1990 Census: Special Tabulation Product #214: Working Paper #11. Metropolitan Transportation Commission, Oakland, Calif., January, 1996.

13. Bay Area Place to Place Journey to Work Characteristics: 1980 - 1990: Electronic Publication #1. Metropolitan Transportation Commission, Oakland, Calif., April 1993.

14. Bay Area Place to Place Journey to Work Spreadsheets: 1980 - 1990: Electronic Publication #2. Metropolitan Transportation Commission, Oakland, Calif., March, 1994.

15. San Francisco Bay Area: Place-Level Profiles: 1990 Census: Census Transportation Planning Package (Statewide Element): Electronic Publication #3. Metropolitan Transportation Commission, Oakland, Calif., September, 1994.

16. Purvis, Charles L. Using Census Public Use Microdata Sample to Estimate Demographic and Automobile Ownership Models. In Transportation Research Record 1443, TRB, National Research Council, Washington, D.C., 1994, pp. 21-29.

17. Purvis, Charles L. Estimation and Validation of Aggregate Home-Based Work Trip Walk and Bicycle Share Models - Technical Memorandum: HBWWALK/BIKE #1. In San Francisco Bay Area 1990 Travel Model Development Project: Compilation of Technical Memorandum: Volume I. Metropolitan Transportation Commission, Oakland, Calif., March 1995.

18. Horowitz, Joel, Frank S. Koppelman and Steven R. Lerman. A Self-Instructing Course in Disaggregate Mode Choice Modeling. Report DOT-T-93-18. Federal Transit Administration, Washington, D.C., December 1986.

19. Kollo, Hanna P.H. 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.

20. Harvey, Greig W. and Elizabeth Deakin. Description of the STEP Analysis Package. Metropolitan Transportation Commission, Oakland, Calif., forthcoming.

21. Harvey, Greig W. and Elizabeth Deakin. Incomes of East Bay Households and Transbay Commuters. Memo to Congestion Pricing Management Board, Metropolitan Transportation Commission, Oakland, Calif., April 11, 1994.

22. Purvis, Chuck. Marin County and the San Francisco Bay Area: Demographic and Commuting Patterns. Metropolitan Transportation Commission, Oakland, Calif., November 1994.

23. Cervero, Robert and Kang-Li Wu. Polycentrism, Commuting, and Residential Location in the San Francisco Bay Area. Working Paper 640. Institute of Urban and Regional Development, University of California, Berkeley, April 1995.

24. New Research Center Opens in Boston. In Census and You. Bureau of the Census, Washington, D.C., March 1994, pp. 9,12.

Converted to HTML by the author, Dec-14-1998. Last reviewed, Dec-07-2004.