Long-Distance and Rural Travel Transferable Parameters for Statewide Travel Forecasting Models (2012)

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.

37 This chapter of the Guidebook identifies alternate approaches and contexts to developing parameters that are more fully developed in later chapters. Although some of the suggestions in this chapter were not feasible to undertake in the final analysis for this study, these are still impor- tant considerations should additional analysis be warranted in the future, either at the national level or for the purposes of individual statewide model development efforts. This chapter provides general guidance on when and when not to transfer model parameters by identifying conditions and parameters conducive to transferability, depending on available data sources and other considerations, and expanding on discussions about this topic found elsewhere in this report. This chapter also describes procedures for consideration in conducting analysis, both for this study and future research by others at the state or regional level. Transfer- ability, analysis procedures, and typology topics covered in this chapter include the following: • Conditions conducive to transferability—Demographic and geographic considerations as well as availability of local data; • Parameters to be considered for transferability—Which parameters are easiest to estimate and have sufficient data to support transferability; • Temporal analysis considerations—Daily, monthly, annually, time of day, seasonality, week- days, weekends, etc.; • Other aspects of “trip” definition—Consideration of intermediate stops, trips versus tours, etc.; • Proposed process to use datasets for developing transferable parameters—Identifying what model parameters can be estimated, where applicable, from these datasets and the level/type of effort involved; • Limitations of all datasets—Identifying geographic, trip type, or modal limitations in the accuracy of the data; • Minimum amount of local data required to assess the reasonableness of parameters—How the resulting parameters can be tested or compared for reasonableness; and • Long-distance and rural typologies—How to stratify households, trips, and areas. Later chapters of this report quantify some of the general guidance and analysis procedures provided in this chapter by model step. 3.1 Conditions Conducive to Transferability A set of draft rules must be established indicating conditions conducive to transferability and those that are not. Such considerations will include, but not be limited to, the following: • Population densities; • Median income; C h a p t e r 3 Transferability and Typologies

38 Long-Distance and rural travel transferable parameters for Statewide travel Forecasting Models • Available transportation modes; • Key employment types/industries; • Proximity to tourist destinations; and • Source of model parameter. Population density is a potential indicator of model transferability. This is particularly the case with mode choice for long-distance travel, as private passenger vehicles predominate in long-distance travel in smaller-sized urbanized areas and rural areas while long-distance travel is more common on alternate modes in large metropolitan areas. Clearly there is a relationship between population density and available transportation modes that also explains the mode choice issue. Density explains some differences in trip rates, trip lengths, and auto occupancies of urban versus rural trip-makers. With respect to analysis of median income impacts on trip-making, it stands to reason that lower-income households make fewer long-distance trips than higher-income households. Like- wise, household decisions on transportation modes for long-distance travel include an income component. The Guidebook chapter on trip generation provides additional understanding on the relationship between income and rural trip-making. Key employment types and industries can impact rural trip-making. A good example of this is tourism and lodging, which has a large need for low-income workers who cannot afford to live in proximity to resort developments. Such areas are also magnets for long-distance travel since visitors to resorts usually reside outside of the region. The source of the model parameter is a key decision point in parameter transferability because there is a wide variety of sources considered in establishing such settings, including state DOT surveys (both household and intercept), surveys from adjacent or similar states, national surveys, MPO surveys, NCHRP Report 716 and other model guidance documents, as well as other statewide models. Furthermore, smaller states (e.g., Rhode Island) might have more in common with urban and regional models than statewide models, with a smaller percentage of long-distance trip activity and dominated by urbanized land. Clearly, long-distance model parameters should be derived from surveys with a statistically valid sample of such trip-makers. Rural model parameters require a survey with both urban and rural resident components in order to ensure that the resulting rates are in fact the result of differences in residential and/or work location and not just due to error in survey execution or design. Although reported statistics from statewide models and documentation of general guid- ance are useful to provide context, such comparisons are no substitution for analysis of travel survey data. The limitations of the data sources must also be considered, especially as these relate to geo- graphic limitations or trip definition. The minimum amount of data needed for the geography intended (national, regional, state, or metro area) must be assessed for each of the parameters. 3.2 Parameters to Be Considered for Transferability Some potentially transferable parameters that are important to properly estimating long- distance and rural travel patterns and comparative benchmark statistics for statewide models are described in this section. Potentially transferable parameters include the following: • Daily (weekday and weekend) rural trip rates per household by household characteristics (e.g., number of workers by industry) and by trip purpose; • Monthly or annual long-distance trip rates per household by household characteristics (e.g., median income) and by type of trip (trip purpose); • Friction factors, gamma functions, or utilities for rural travel by trip purpose;

transferability and typologies 39 • Friction factors, gamma functions, or utilities for long-distance travel by trip purpose; • Auto occupancy rates for rural vehicle trips by trip purpose; and • Party size for long-distance trips by trip purpose. In addition to the transferable parameters itemized above, and the dynamics noted earlier in this chapter, reasonableness values are documented in later chapters for the following: • Percent of rural trips by purpose; • Percent of long-distance trips by trip purpose; • Average (mean) vehicle trip length of rural trips by mode and purpose; • Average vehicle trip length of long-distance trips by mode and trip purpose; and • Percent of long-distance and rural trips by mode (private vehicle, rail/bus, air, other) and travel distance. Other parameters and benchmarks more difficult to quantify, and less likely transferable, include intermodal connections (e.g., dropped off/drive and park at airports); percent of non- traveling households by type; and percent of trip destinations to locations within the same state or to another state. 3.3 Temporal Analysis Considerations of Transferability Defining what constitutes a statewide model trip is also important to a discussion on trans- ferability. Even this varies among different statewide models, with a few that essentially do not include intra-urban trips (e.g., Louisiana). Trips could be defined by person, household, or even vehicle in some cases. Sometimes, it might make sense to include intermediate stops as trip ends; however, this would seemingly go against the concept of long-distance trips. In fact, what travelers typically think of as a “(round) trip” is what transportation planners consider a “tour.” A few state- wide models (e.g., Ohio, Oregon, and New Hampshire) use the concept of tours instead of trips. For rural travel analysis, average weekday conditions would likely be preferable. Similar to regional models, while it might be best to exclude travel on weekends and holidays, such limitations could result in sample size problems. NHTS staff have indicated that approxi- mately 25–30 percent of 2009 NHTS surveys were conducted on weekend travel; however, weekend travel includes Friday after 6:00 p.m. (teleconference with Adella Santos, FHWA; Vidya Mysore and Frank Tabatabaee, Florida DOT; and Rob Schiffer, Cambridge Systematics, Inc. on August 10, 2011), a timeframe that is similar to other weekday evening peak periods in many regional models. In states with a singular, well-defined peak season, consideration could be given to only including surveys that constitute peak season average weekday traffic instead of annual average daily traffic (AADT), although such a timeframe of analysis would not be recommended for a study on national transferability such as this. Conversely, since long-distance travel is not an everyday occurrence in most households, monthly or annual statistics are appropriate for survey analyses. Also, it is essential to include weekends and holidays in any survey analysis of long-distance travel because these time periods reflect where the greatest amount of such travel takes place. Consideration was given to develop- ing time-of-day factors both for rural and long-distance trips during this study; however, with the infrequency of long-distance trips, use of trip rates by time of day might be overkill. 3.4 Other Aspects of Trip Definition for Transferability In addition to temporal considerations, there are other aspects to be considered in defining a trip for the purposes of research and analysis of transferable parameters. The first of these is consideration of person trip versus vehicle trip analysis. Since the majority of statewide models

40 Long-Distance and rural travel transferable parameters for Statewide travel Forecasting Models deal with person trips and starting with vehicle trips almost precludes a mode choice process, transferable parameters in this study are provided for person trips rather than vehicle trips. Long-distance trip-making was considered at 2–3 different thresholds to determine how param- eters differ at each threshold. Another consideration is how to deal with intermediate stops and whether these should con- stitute a trip end or not. Clearly, long-distance trips require stops for gas, food, and/or lodging. In the context of a regional model, these intermediate stops for shopping, etc., would each rep- resent a unique nonhome-based trip. In the context of most statewide, multistate, or national modeling, however, these intermediate stops are not of tremendous importance in defining and simulating a trip. On the other hand, it is probably worth considering an intermediate stop at the end of the day for lodging as the end of a daily trip, assuming the analysis is daily rather than monthly or annually. The location of intermediate stops, relative to congestion on Interstate high- ways or crossroads, could result in greater interest about intermediate travel patterns. The number and duration of stops will also be addressed in this research. The topic of intermediate stops also leads directly to consideration of tours versus trips. The previous lodging example might be better addressed as a stop during a tour rather than the endpoint of the trip; however, the majority of statewide models are still trip-based. Those statewide models that are tour-based were developed using statewide travel surveys and, as a result, will not likely have as much use for transferable parameters. However, the preparation of tour-based parameters is beyond the scope of this report. Hence, the number of interme- diate stops for long-distance trips is provided later in this Guidebook based on analysis of the 1995 ATS. 3.5 Process for Developing Transferable Parameters Datasets with the greatest potential for developing transferable parameters for rural travel are the 2009 NHTS and readily available analyses from the Michigan and Ohio statewide travel surveys. New and ongoing surveys in California, Colorado, Oregon, and Utah could also prove useful in documenting rural travel patterns. Transferable parameters for long-distance travel are best derived through a combination of the 2001 NHTS, 1995 ATS, and readily available analyses from the Michigan and Ohio long-distance travel surveys. The development of transferable parameters from any of these household diary surveys requires some level of significance testing to identify what comparisons, typologies, geographies, and time periods best explain or influence the characteristics of rural and long-distance travel. The following dynamics must be considered in the analytical process: • Comparisons—Comparisons should include rural versus urban households and long- distance trips versus routine (urban and rural) travel. • Typologies—Rural and long-distance travel should be analyzed by household characteristics such as income, number of workers, household size, auto availability, etc. “Type” or purpose for long-distance trips can be equated to ATS categories of business, pleasure, and personal business while rural trips can be described using typical home-based and nonhome-based purposes, similar to urban models. • Geographies—Rural areas should be analyzed by type such as proximity to/distance from urbanized boundary, land-use density, and/or roadway density. • Time Periods—This topic is repeated here due to its importance in framing the analysis and results described in subsequent chapters. For rural travel, average weekday conditions would likely be preferable (excluding weekends and holidays, similar to regional models); however, since long-distance travel is not an everyday occurrence in most households, monthly or

transferability and typologies 41 annual statistics must be considered (including weekends and holidays) in analyzing long- distance trip-making. Use of tourist survey data and Canadian travel behavior data would require additional resources for data procurement or access to free sample subsets of these data with limitations and caveats covering their use in this study. Therefore, similar to the Michigan and Ohio sur- veys, analysis was limited to readily available statistics provided by others. In cases where statistics have been summarized from reports, research studies, and data that does not include travel diaries, the best likely achievable product is providing reasonableness values, similar to those provided in later chapters of this Guidebook, that can be used in assessing the validity of statewide, multistate, and national models or to use as a point of comparison or verification against analytical results from this study. The development of transferable parameters and reasonableness values used statistical analysis software (e.g., SAS 9.2) to look at issues of significance, variance, and dispersion. Each dataset was analyzed independently, since estimation of error cannot be done on the combined datasets. The NHTS datasets estimate margin of error using a replicate weight file containing 99 replicate weights prepared for each household and person. Estimates of error in the ATS data are limited to margins of error in published reports, and simple standard deviations, which do not take into account the probability of selection from a PSU sample design. A cluster analysis was considered to develop homogeneous sets or clusters of households, person, and/or trip characteristics using the variance (margin of error where available, simple standard deviation where not). Instead, regression analysis was used to determine what indepen- dent variables are most explanatory in predicting variations in trip rates, average trip lengths, and auto occupancies by type or purpose. This study included coordination efforts with NCHRP Project 25-36, “Impacts of Land Use Strategies on Travel Behavior in Small Communities and Rural Areas” and its development of rural typology using cluster analysis. The pros and cons of using a common typology are addressed elsewhere in this Guidebook. A top-down analytical approach to analyzing rural travel started with the 2009 NHTS national sample. This sample is population proportionate and includes rural households from all parts of the United States in proportion to their incidence, or about 22 percent of sampled households overall. Using those data, rural travel indicators were developed from a national perspective, establishing a benchmark against differences in travel. Such an analysis provides a normative set of values for national urban and national rural travel character- istics. These benchmark values formed the beginning of our understanding of how rural travel characteristics differ from the types of data used in traditional urban travel forecasting models. NHTS also uses Census geography to divide the nation into smaller areas (e.g., Census region or division). This second level of geographic analysis could benefit the process of dividing rural travel characteristics into specific typologies; however, parameter values described later in this Guidebook maximized use of the entire sample, rather than diluting the data into subsamples by Census geography. Finally, in the 2009 NHTS, 15 states purchased supplemental samples and all of them included rural households. The travel of rural households in these Add-On states is a third level of geo- graphic analysis. Smaller, non-Add-On states had a minimum of only 250 households sampled for the entire state. In those states there would be too few rural samples to make robust estimates of rural travel. This process could help determine what final geography the data supports to create

42 Long-Distance and rural travel transferable parameters for Statewide travel Forecasting Models transferable parameters that have the widest applicability (e.g., Add-On states already have rural samples for their statewide models, so perhaps the audience for this report is more than likely the non-Add-On states). Parameters described in later sections of this Guidebook were based on analysis that included NHTS Add-On data. 3.6 Limitations of Datasets Each dataset identified in this Guidebook has limitations that impact the estimation of trans- ferable model parameters. Some of the key limitations of each dataset are identified below: • 1995 ATS—Potential concerns over the age of the data, there are no rural samples (the PSUs were all in urbanized areas), and no ability to estimate true error; • 2001 NHTS—Similar concerns over the age of the data, low mode differentiation, especially low air estimates (partially due to 9/11), and extreme values in trip lengths need to be trimmed; • 2009 NHTS—No long-distance trips, large and comparable rural sample but not all states are represented in the Add-Ons; and • Michigan and Ohio Surveys—One could assert that parameters from these surveys might only apply to highly populated states with manufacturing as a key driver of the economy. Availability of other recent and ongoing statewide and superregional household surveys also helped identify areas of compatibility and inconsistency. It is anticipated that some travel modes might be underrepresented in these surveys as none were stratified to achieve a target number of responses by mode. Access to additional modal- based data from Amtrak, Federal Aviation Administration (FAA) enplanements, and possibly Greyhound would be needed to back up this statement. In the interest of not over-specifying based on available data, rural trip purposes have been limited to home-based work (HBW), home-based nonwork/other (HBNW/HBO), and non- home-based (NHB). For long-distance estimates, modes were limited to those most commonly reported, and trip types to the three main trip purposes consistent with ATS definitions (busi- ness, pleasure, and personal business). An understanding of household characteristics that influ- ence the likelihood of making long-distance trips and mode selection for such trips also are addressed in the analysis. 3.7 Minimum Amount of Local Data Required The only local data needed for analysis are general statistics culled from available statewide model reports, as documented in subsequent chapters of this report. In order to confirm the usefulness of the resulting model parameters, testing of parameter settings could be conducted on one or more existing statewide travel demand forecasting models by future users of this document. Reasonableness values can be assessed against different statewide models to document their usefulness. Cautionary statements of transferability were included earlier in this chapter such that model- ers understand the conditions under which transferable parameters are recommended for use in statewide models. Local data collection is always preferable to borrowed parameters as long as sufficient funding is available to conduct these surveys. As noted earlier, this study also included coordination with NCHRP Project 25-36, including cluster analyses to define a rural typology for analysis purposes. The cluster analysis makes use of

transferability and typologies 43 “commuting zones” established by the U.S. Department of Agriculture (http://www.ers.usda.gov/ briefing/rurality/lmacz/). Variables under consideration with NCHRP Project 25-36 include population density, road density, land-use mixture, and variation in population density. A rural typology is critical in establishing the transferability of model parameters. 3.8 Long-Distance and Rural Typology Considerations This section describes alternative stratifications for long-distance and rural trip-making based on statistical analyses for this project as well as coordination with other research efforts. A key analytical step in this research has been to compare trip generation statistics for house- holds in “rural” areas, using various definitions to assess resulting differences in trip rates. As noted elsewhere, such analyses also must account for urban trip characteristics as a point of comparison. Analytical comparisons necessitate a typology of rural activity, such as defining rural house- holds nearer to urban centers versus those farther from large activity centers. Another unique characteristic of some rural areas, yet more difficult to quantify, is proximity to major rec- reational areas. For example, trip activity is likely different for rural areas in proximity to national parks, beach areas, and casinos, than rural areas that are largely focused on agricul- ture, mining, and forestry. Demographic profiles are also helpful, defining household characteristics such as size, life- cycle, income, and/or number of workers by worker status and occupation. An interesting topic, should such data become available in the future, would be to include comparisons of Internet availability and use this information to impute if rural households are more or less likely to shop online, based on a lack of home-based shop trips. The propensity of rural resi- dents to link trips is another unique factor because those with long daily commutes are likely to do their shopping and other personal business prior to leaving the urban area at the end of the work day. While such intermediate stops might not be important to statewide models, as discussed previously, there might be reason to analyze this with an eye toward regional models covering large rural territories. The selected rural characteristics could be analyzed for households in each state, and in state clusters, such as Census division, to determine what trip generation statistics may be transferable to areas without original data to support a rural trip generation model. The research team has identified opportunities to leverage some of the analysis already conducted for urban transfer- able parameters (NCHRP Report 716) and much of the thinking on new typologies, especially sociodemographic, would be helpful for this effort. The Version 2 NHTS 2009 has a number of enhancements that can be helpful for analytical purposes, including estimates based on the 2008 American Community Survey (ACS) and land- use descriptors for the household and the workplace locations from Claritas/Neilson. This is an important source of comparable land-use definitions when looking at trip generation estima- tion from state to state. Claritas has developed a “floating density” estimate based on contiguous 2-mile-square grids across the country, which evens out some of the variability in density based on Census tract area. Selected characteristics of urbanized areas from the annual “Highway Statistics” publication of the FHWA Office of Highway Policy Information (OHPI) can also potentially assist in defining characteristics that separate rural from urban settings. Trip rates for households in similar sociodemographic classes, such as number of workers or income, were estimated to see how much of the difference can be accounted for by traditional

44 Long-Distance and rural travel transferable parameters for Statewide travel Forecasting Models methods. This is the point at which it could be useful to conduct analyses with and without weekends and holidays. As noted earlier, removing households and persons with weekend and holiday travel days will reduce the sample sizes by approximately 30 percent and increase the margin of error overall. Significance testing, along with measures of variance, could be useful to inform readers of this report of possible future paths and other considerations. Looking at the distribution of trips by purpose might highlight sources of variance. The trip purposes used in trip generation are generalized travel purposes, such as home-based work (HBW), home-based other (HBO), and nonhome-based (NHB). It is certainly possible to look at rural trip generation for these purposes, but the very high VMT per person and per household hints at long work trip lengths. Research on trip chaining has indicated that people with longer commutes are more likely to stop along the way than other commuters (McGuckin, N., Zmud, J., and Nakamoto, Y., 2005). HBW is coded as direct trips between home and work, and so longer commutes with an intervening stop might not be included in the HBW category but pushed into NHB, possibly skewing the resulting percents. Auto occupancy is another important travel parameter that could vary substantially between states, by trip purpose, and by the sociodemo- graphic class of the household. Ultimately, it could be useful to create transferable “types” of rural areas that do not rely on Census or state designation, but truly tie the variations in travel to land-use characteristics of the rural household location, and that can be applied across geography. Evaluating rural areas based on roadway density would be helpful as a step to analyzing location variation. Another consideration would be to classify rural areas based on proximity to urbanization and other readily available characteristics. Another ideal step toward creating a typology of rural areas could include acquiring data on the major industry for households in that rural area. For example, areas dependent on agricul- ture or manufacturing would potentially have different travel characteristics. In the 2009 NHTS, Claritas added detail to the file that might possibly be helpful at some point. For households that report a workplace location, the following information is available for that workplace location at the Census tract level: • Percent of workers Agriculture/Mining/Construction; • Percent of workers Finance/Insurance/Real Estate; • Percent of workers Manufacturing; • Percent of workers Retail; • Percent of workers Services; • Percent of workers Transportation/Communications/Utilities; and • Percent of workers Wholesale Trade. Since these data are tied to the workplace location and not the household, supplemental analy- sis would be required to explore their usefulness for this task. At the household location, only workers by retail or nonretail are coded (at the block group level). In addition to the major industry of a household’s workers, an important question is how far away are desirable destinations, such as entertainment or shopping opportunities? Of course, it is possible to use the reported trip lengths for shopping and entertainment to cluster rural households into groups, but that leaves households where nobody reported a shopping or enter- tainment trip on the travel day outside of the typology. With a complete source of information on MPO and urbanized boundaries, the research team looked at coding rural areas within and outside of these boundaries as a way to get at distance characteristics. Such analyses would need to be sensitive to varying levels of urbanization and the political nature of MPO boundaries throughout the United States.

transferability and typologies 45 Proximity to modal alternatives and urban areas impacts the quantity, trip purpose, and travel mode of rural and long-distance trips. The biggest problem with using modal proximity in a rural typology is that the NHTS and ATS do not include information on respondents’ proxim- ity to transportation modes. Therefore, such a typology would require manual coding of modal proximity information into survey records or identifying some sort of access variable based on existing attributes. Proximity to urban areas is easier to address, especially if Metropolitan Statistical Areas (MSAs) or Consolidated MSAs are used to define urban proximity, because these measures are already in the NHTS. The problem with any rural typology based on distance to attractions is the effort and information required to code the location data. In conducting analysis with NHTS and other survey databases, researchers must keep in mind how the transferred rural estimations would be applied—in other words, how easy it would be for statewide planners to obtain the same data for the rural households or areas in their state in order to correctly apply the travel rates. A rural typology could be based on several other factors. The goal is to choose factors that both explain perceived differences in travel rates or type and that can be reasonably quantified for different rural areas in individual states. Following are three more alternate approaches to further refining the definition of “rural” households that could be explored further using the 2009 NHTS. The NHTS variable names and unweighted sample sizes are provided for convenience in capital letters: • Census-defined rural households that fall inside or outside of a CMSA—Of the NHTS households that are classified as “Rural” by Census definition, a slight majority are located within CMSAs (using URBRUR and HHC_MSA<‘0000’ for not in a CMSA). In essence, over 50 percent of Census-defined rural households are located within the boundaries of MSAs or CMSAs. • Housing Unit Density (HBRESDN)—Nearly 30,000 households classified as rural fall in the lowest coded density (zero-99 housing units per square mile). More than 10,000 fall in the next coded level (100 to 499 housing units per square mile). The remainder are spread out across other density codes. These density calculations are appended to the NHTS file by Claritas and are based on Census estimates at the block group level of geography. • Major Employment Type (based on employment of residents at household location, not employment at place of work)—These data are available from FHWA as an Additional Variable File (Claritas) and have the following variables of interest: percent of employees in Agriculture, Mining, Construction (WTINDAGR), percent of employees in Manufacturing (WTINDMAN), percent of employees in Retail and Service (WTINDRET, WTINDSVC), and other designations (WTINDTRN, WTINDFIN, WTINDWHL or Transportation, Finance and Wholesale, respectively). These and perhaps other “typing” variables can be analyzed to see if differences in household travel between various rural areas of the country, or even within a state, can be explained. The same ANOVA cluster analysis that was conducted for NCHRP Report 716 for the rural house- holds in the NHTS dataset was used as part of NCHRP Project 8-84 with an expanded statistical analysis to see if these factors are significant. Below are some specific approaches for developing weekday trip generation rates for rural households that maximize the use of statistical analyses already conducted as part of NCHRP Report 716. • At a minimum, trip rates (daily weekday person trips per person) should be presented in the format of cross-classification trip production rates presented in NCHRP Report 716, which

46 Long-Distance and rural travel transferable parameters for Statewide travel Forecasting Models found that trip rates were not statistically different for households in different population sizes. Therefore the report presents urban trip generation data by different trip purposes. The purposes used in NCHRP Report 716 are as follows: – Home-based work; – Home-based nonwork; – Home-based school; – Home-based other; and – Nonhome-based. • The purposes that came out as significant in a preliminary rural trip length cluster analysis for this study were: HBW, HBShop, HB Recreation, and all others. Since these purposes do not map onto each other, some additional consolidation is recommended (HBW, HBO, NHB). • Percent of trips and trip rates per person per household have been developed, by purpose and for “All.” Should such information be available, consideration could also be given to typical driving ages in rural areas and impacts to school trip lengths. In addition, this study should develop auto occupancy factors for the same purposes and “All.” It was very helpful to “recycle” the SAS code from the development of the NCHRP Report 716 trip generation tables for this study, or at least to be reminded to use the same assumptions, omissions, etc.