Census Data for Transportation

Census Data for Transportation:

Preparing for the Future

May 11-13, 2005, Irvine CA

Proposed Poster

Adding a Land Use Class Dimension to Correct, Improve, and Enlarge Employment Data in the American Community Survey.

INTRODUCTION

The purpose of the proposed poster is to advocate the inclusion of the concept of land use in the American Community Survey (ACS), in order to correct major errors in, and extend the reach and usefulness of, Census employment data. Both the decennial census and the ACS recognize three dimensions of kind of work or job activity: industrial class, occupational class, and class of worker. However, there is a fourth dimension of job activity, land use, that plays a key role in transportation planning, but that is ignored by both decennial census and ACS.

LAND USE DEFINITION

In the present context, the term “land use” has a restricted meaning. It does not include all the “land use” inputs needed by transportation models, nor is it the inclusive, umbrella concept implied in the phrase “land use and transportation”. Rather, the proposed poster will use the term “land use class” to mean the nature and characteristics of the activities and functions occurring at specific individual locations, and the buildings and associated open land they occupy. Land use class encompasses the general activities, their purposes, and physical activities.

For economic activities, the commonly recognized types of activities include office, commercial, institutional, and industrial, as well as others. The function or purpose of economic activities is designated by industrial classes. The facilities or workplaces are described in terms of building type, e.g. store, factory, and kind of development, e.g. shopping mall, industrial park. The various kinds of land use, arranged in their typical spatial patterns, are what we see when walking or riding through or flying over urban areas.

PROBLEMS AND SOLUTIONS

The poster will use text, tables, and graphics to illustrate three key problems, and their solutions. These are:

Erroneous Industrial Classes

Employment in establishments that under SIC were called auxiliaries – consisting of central administrative offices, research and development facilities, warehouses, and vehicular storage facilities – were supposed to be given their own NAICS codes in Census 2000. However, the questions asked did not elicit the correct responses, so such workers were given incorrect NAICS codes. There are possibly five million such misclassified workers. In addition, these establishments are not ubiquitous but rather are spatially concentrated, thereby negatively impacting small-area data. The same questions are being used in ACS, so the misclassification problem will continue.

The solution involves using the Census Bureau’s business establishment list to identify auxiliaries, by type, and to approximate their employment. Next, these workers would be given the correct NAICS class, as well as the corresponding land use class. Lastly, they would be subtracted from the industrial class to which they were assigned incorrectly.

Inputing Land Use Class

Land use is a missing dimension in Census 2000, and in the ACS. No questions about land use were in the Census 2000 questionnaire, nor has ACS included such questions, nor will it through 2010. However, although the questions are not being asked of respondents, the questions can and should be asked of the data. This can be done via imputation, in the form of post-processing. At the 1996 Census Conference, my paper “Improvement of Decennial Census Small-Area Employment Data : Method to Assign Land Use Classes to Workers” demonstrated how a special tabulation, in the form of a place of work crosstab of workers by industry by occupation, was used to impute general land use classes. The poster would summarize the procedure and results, and would emphasize that the same techniques applied to Census or ACS post-2004 data could provide a similar land use class assignment.

Employment Data/Trip Generation Data Incompatibility

Census place of work employment data, and non-Census data on trip generation, cannot be interrelated in a satisfactory fashion. This noncomparibility has three causes:

1)the two kinds of data use incompatible classification systems. Census uses industrial class to identify kind of activity, while trip generation data sources typically use a composite system, partly industrial class, partly land use class;

2) the Institute of Transportation Engineers’ (ITE) Trip Generation manual, the pre-eminent source of trip generation rates, uses a detailed composite system of what it calls “land uses.” However the ITE manual draws its data cases mainly from developments occurring in the newer portions of urbanizing areas, which typically have limited or nonexistent accessibility via transit, walk, and bicycle. In addition, the manual does not record transit, walk, and bicycle trips even when they do occur. Therefore the ITE rates cannot be applied to older urban areas in which land use patterns are more tightly structured, and automobile travel is less predominant;

3)in any given year household travel surveys for several individual metropolitan areas are being conducted, and their household samples capture trip characteristics throughout the metro area and beyond. However, the classification system used to identify the types of origins and destinations typically consists of a few general types, that also vary area to area..

The solution to noncomparability, Census employment vs. trip generation rates, can be achieved using a five-step process:

1)create a crosswalk that links Census industry codes and ITE land uses, and produce a set of composite Census/ITE classes;

2)Using Cernsus worker records, (after the correction for NAICS misclassification, and land use assignment, as described earlier), assign a composite Census/ITE code to each worker;

3)in each future household travel survey, code trip ends using Census/ITE codes;

4)develop a set of general subarea-types, based upon land use, age of development, population, jobs, and transit and highway accessibility;

5)calculate trip ends per 1,000 workers, by each Census/ITE code, resulting in generation rates by subarea type by Census/ITE code.

With the completion of each metro area travel survey, the accumulated “fund of knowledge” about trip generation by kind of job activity would improve and expand. Trip generation rates would be available not just for new development, but for all expanded development by subarea-type.

Edward Limoges

Ann Arbor, Michigan

January 21, 2005