Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
sizer. The population synthesizer provides flexibility in use with current and future land use forecasts and it can be adjusted based on validation results. ⢠A population synthesizer is a powerful tool, but it should be used with caution. By design, a population synthesizer may provide misleading details about every individual in the population. A goal in the development of a population synthesizer is to synthesize, as accurately as possible, at as disaggregate a level as possible, with as many variables as possible that determine travel behav- ior. A population synthesizer should be used for the char- acteristics that it accurately represents aggregated to a level at which it is precise and accurate. ⢠The ARC population synthesizer uses object- oriented Java software. It consists of subprograms called classes. Inputs include census data for the base- year and population forecasts for the forecast year. The population synthesizer validates accuracy at multiple aggregation levels, including demographic and geographic levels. The results feed into the activity- based travel demand model. ⢠The population synthesizer creates a synthetic pop- ulation for the base year and for each forecast year. A public use microdata sample (PUMS) is used for the ini- tial distribution for the base year. For the forecast year, the distribution comes from the base- year distribution. The census tables are used for the controls for the base year, while the controls for the forecast years use the land use forecasts. A synthetic population is produced for the base year and the forecast year, along with a validation report that compares the synthetic population character- istics with known characteristics. The base year is 2000 and a back- cast to 1990 is used to validate the synthe- sizerâs ability to generate a forecast population. ⢠Three versions of the population synthesizer were created for the initial testing and validation process. The simplest version of the population synthesizer has 52 household demographic categories. Two versions that are more detailed have 128 and 316 household demo- graphic categories, respectively. More detail from the census tables for the base year and from the ARC demo- graphic and land development forecast for the forecast year can be used as more categories are added. The com- putation time lengthens as categories are added, how- ever, and the increase in the number of sparsely populated categories causes more rounding errors. The validation process helps identify the most appropriate household demographic categories to use. ⢠Household size is controlled at the travel analysis zone (TAZ) level. In the base year, four categories are used for the version with 52 household demographic characteristics and five categories are used with the 128 and 316 versions. Household size is controlled at the TAZ level in the forecast year, but only average house- hold size is available. The base- year distribution is used to translate this information into the controlled cate- gories. Age is controlled only in the version with 316 household demographic categories. Three age cate- gories are used. These categories are whether the head of the household is over or under age 65, and for those households under age 65, whether any children under age 18 are present. In the forecast year, control is used only as the regional sizes of the subpopulations 65 years of age and older and less than age 15. The fore- cast control categories are sized by using relationships in the base- year census PUMS data between the avail- able values and the control categories. Employment in the base year is controlled at the TAZ level as the num- ber of workers in the household by four categories. For the forecast year, the control is used only for the region and only for the average number of workers per household. ⢠The results of the validation process indicate that the use of census data to control for more variables in the base year 316 version produces a more accurate synthetic population. For the forecast year, the addi- tional controls in the 128 and 316 versions provide lit- tle value. The results highlight a number of aspects related to the use of population synthesizers. The accu- racy of the synthesized characteristics depends on the control variables used for population synthesis, with uncontrolled variables synthesized less accurately. Accuracy also declines at more detailed levels of aggre- gation. Accuracy for control variables can be influ- enced by rounding procedures and the use of averages. It is important to validate a population synthesizer by examining these elements on a case- by- case basis. The results of the validation process can be used to improve the population synthesizer. MICROSIMULATION OF SINGLE- FAMILY RESIDENTIAL LAND USE FOR MARKET EQUILIBRIUMS Bin Zhou and Kara Kockelman Bin Zhou described a study examining single- family res- idential developments for housing market equilibrium using microeconomic theory and disaggregate spatial data. She discussed the use of a logit model and notions of price competition to simulate household location choices in Austin, Texas. She summarized the back- ground to the study, the data collection process, the model of location choice, and the market equilibration results. Volume 2 includes a paper on this topic.3 The following points were covered in her presentation. 22 INNOVATIONS IN TRAVEL DEMAND MODELING, VOLUME 1 3 See Zhou, B., and K. M. Kockelman. Microsimulation of Single- Family Residential Land Use for Market Equilibriums. Volume 2, pp. 63â68.
