Using Commodity Flow Survey Microdata and Other Establishment Data to Estimate the Generation of Freight, Freight Trips, and Service Trips: Guidebook (2016)

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.

31 This chapter provides the reader with actionable information about how to use the guidebook models. It describes potential uses of the guidebook models and summarizes the various steps to apply them in the context of practical applications. Practical Uses (When to Use What) The guidebook models can help provide answers to a number of important questions pertain- ing to FSA at several levels of transportation decision making. Table 5 shows a sample of potential applications, together with the types of models that could help provide the information needed. Typical Applications Applications of the guidebook models are similar to those of traditional trip generation models; only the independent variables are different. When discussing the various applications, the term metric of FSA (or FSA metric) is used to designate all possibilities (FG, FTG, and STG). Essentially, using the models requires: • assembling the data needed for the type of establishments involved; • running the corresponding models; and • aggregating the results (if the estimates are for a conglomerate of users). The most common applications can be categorized on the basis of the number of establish- ments and the type of data available to quantify the FSA. In terms of number of establishments, two situations can apply: 1. Single commercial establishment. In this situation, the analyst is interested in quantifying the FSA for a single commercial unit (e.g., a restaurant). 2. Conglomerate of commercial establishments. Here, the interest is on estimating FSA for groups of commercial establishments (e.g., establishments at a commercial center, on a com- mercial street, in a neighborhood, grouped by ZIP code, or in a TAZ). It should be noted that commercial developments such as malls typically house multiple establishments. In these cases, the FSA is the summation of the FSA taking place at the establishment level. In terms of data availability, two cases also can apply: • Case 1: Establishment-level data are available. In this situation, estimates of the industry sector and data about the number of employees for each establishment are available (e.g., when the establishments are already in operation and the analyst is interested in quantifying FSA to determine how many parking spaces should be allocated to FSA). The data required are the establishments’ industry sectors, and the corresponding numbers of employees. C h a p t e r 7 How to Apply the Models

32 Using Commodity Flow Survey Microdata and Other establishment Data to estimate the Generation of Freight, Freight trips, and Service trips • Case 2: Only aggregate data are available. This situation occurs with official employment statistics that are only released as aggregate numbers (e.g., the ZIP Code Business Patterns data- base, and in planning applications for which it usually is not possible to forecast establishment- level estimates of employment). The aggregate data required are, for each industry sector of interest, the number of establishments and number of employees. Table 6 groups the various combinations of these situations into scenarios. Data Inputs The guidebook models only need two inputs: the industry sector and the establishment’s employment in full-time-equivalents (FTE). As shown in Equation 17, the FTE is the summa- tion of the number of full-time employees plus 0.45 times the number of part-time employees. FTA FTP FTG STA STP STG FA FP FG Traffic impact analysis Number of parking spaces needed by freight vehicles * Number of parking spaces needed by service vehicles * Number of parking spaces needed by commercial vehicles * Analysis of trends in freight activity * Analysis of trends in FSA * Estimation of freight trip generation * Estimation of service traffic generation * Estimation of freight generation * Freight Trip Generation Service Trip Generation Freight Generation Description * These analyses can be conducted at any geographic level. Table 5. Practical uses of the models. Type of Data Available Number of Establishments Establishment-Level Data Are Available Only Aggregate Data Are Available Single Scenario 1: Single Establishment - Decide on the desired metric(s) of FSA. - Select the model(s) of interest for the estab- lishment’s industry sector. - Run the model(s) with the establishment data. Applicability: Linear and non-linear models (This scenario is the same as Scenario 1.) Multiple Scenario 2: Complete Enumeration - Decide on the desired metric(s) of FSA. - Run model(s) with data for each establishment. - Aggregate results. Applicability: Linear and non-linear models Scenario 3: Sample Estimation - Decide on the desired metric(s) of FSA. - Run the model(s) with the data for each estab- lishment in the sample. - Compute mean values of FSA. - Expand results to entire population. Applicability: Linear and non-linear models Scenario 4: Only Aggregate Data - Decide on the desired metric(s) of FSA. For each industry sector: - Select the correct aggregation procedure. - Apply the aggregation procedure to obtain aggregate results. Applicability: Only linear models Table 6. Outline of estimation process for typical application scenarios.

how to apply the Models 33 If no data are available, the FTE for comparable establishments from the same industry sector and size can be used. FTE Number of full-time employees + Number of part-time employees 0.45 (17)( )= × An establishment’s industry sector is indicated by its 2-digit or 3-digit NAICS code. Generally speaking, when 3-digit codes are available, the models will generate more precise estimates than models that use 2-digit codes. Whenever possible, 3-digit models should be used. It should be noted that the 2-digit NAICS models represent wide ranges of economic activities. For this reason, 2-digit models could be useful in cases for which the type of economic activity to be performed at a given location is not known with certainty. Model Outputs The guidebook models were estimated using data collected from the demand generators (i.e., the entities that create the need for vehicle trips). By taking this approach, the guidebook models decompose the generation of the demand from the generation of the FSA traffic, a stan- dard practice in passenger transportation modeling. Multiple reasons exist for decomposing the problem. To start, doing so leads to models that are more realistic and flexible, and thus better able to capture the nuanced behaviors observed in real life. Even though demand generation is the manifestation of the economic process conducted at the establishment, the generation of vehicle trips results from the logistics of the distribution. Two establishments—identical in all respects except location—can generate the same amount of shipments while generating different amounts of vehicle trips. For example, if one establishment is located in an isolated area, each delivery made by a vendor may require a vehicle trip. If the other establishment is close to similar businesses, that establishment may generate fewer vehicle trips because the vendors may be able to consolidate deliveries to various businesses. These behaviors can only be explained if the generation of demand is treated separately from the generation of vehicular traffic. Moreover, collecting the data that the demand generators can accurately provide leads to data of better quality that can be linked directly to the establishments’ attributes. A store manager can provide very solid information about the number of deliveries they receive, the amount of cargo received, or the number of service calls in a week; however, the manager may not necessarily know how many vehicle trips these activities actually produced. The decisions about the vehicle trips are made by someone else—either the vendor of the supplies or the carrier. As a result of this decomposition, the demand estimates must be complemented by separate models—albeit simplified ones—that convert demand (measured in deliveries/day, shipments out/day, service calls/day, or pounds/day) into vehicle trips. Ideally, these complementary models provide a reasonable approximation to the decision rules that vendors and carriers use to determine the vehicle trips needed. Estimating the necessary conversion factors is not trivial. The models are likely to depend on multiple factors, including: • the degree of competition among the vendors (the more competition, the more difficult it is to consolidate trips); • the density of the destinations to be visited (the lower the density, the more difficult it is to consolidate trips); and • the urgency of the activity to be performed (the more urgent the activity, the more difficult it is to consolidate trips). Regrettably, it was not possible to collect data to study these effects as part of NCFRP Project 25(01). Further research is needed.

34 Using Commodity Flow Survey Microdata and Other establishment Data to estimate the Generation of Freight, Freight trips, and Service trips To gain insight into the practical range of conversion factors, the team conducted informal interviews with carriers. The suggested values, together with the units of the outputs produced by the different types of models are shown in Table 7. The use of the conversion factors, fA and fP, deserves discussion. To facilitate understanding, situations in which separate models exist for attractions and productions are discussed next, followed by situations in which only a single generation model exists. Separate Models Are Available for Attractions and Productions The models estimated for this guidebook (i.e., the FTA, FTP, STA, and FP models) fall into this category. The most distinguishing feature of these models is that separate conversion factors can be used. FTA: Conversion Factor Between Deliveries/Day to Vehicle Trips/Day The FTA conversion factor takes into account the fact that a vehicle trip could be used to make multiple deliveries. This is standard practice for most parcel carriers, the U.S. Postal Service (USPS), and courier 90000 services, because they can consolidate deliveries given the scale of their operations. Notwithstanding the large size of these operations, however, in reality they transport only a small portion of the total cargo. According to the CFS, parcel/USPS/couriers transport 2% of the total ton-miles and 0.4% of the tons transported by truck-only modes. The carriers interviewed for NCFRP Project 25(01) indicated that, because of stiff competition, it is very difficult to make multiple deliveries from the same location. This information suggests that the conversion factor, fA, should be close to 1. FTP: Conversion Between Shipments/Day to Vehicle Trips/Day This conversion factor takes into account the fact that shipments going out of an establishment can be consolidated to reduce transportation costs. The interviews conducted for this project suggest that this form of consolidation is relatively frequent. The team’s best estimate is that a conversion factor of fP = ½ (2 shipments out = 1 vehicle trip) is appropriate. Obviously, in the case of establishments that rely on parcel carriers, the conversion factor could be even smaller. STP: Conversion Between Service Calls/Day to Vehicle Trips/Day This conversion factor takes into account the fact that a vehicle making service calls could execute more than one service call from the same location. This situation resembles that of the Metric of Freight and Service Activity (FSA) Output unit Conversion Factor to Vehicle-Trips (multiply by) Freight Trip Generation (FTG) (HARTGEN) shipments/day (γ φ A + γ P φ P ) Freight Trip Attraction (FTA) deliveries/day φ = 1.00 Freight Trip Production (FTP) shipments/day φ P = 0.5 Service Trip Generation (STG) - - Service Trip Attraction (STA) service calls/day φ A = 1.00 Service Trip Production (STP) - - Freight Generation (FG) (RPI) - - Freight Attraction (FA) pounds/day Table 20, 21 Freight Production (FP) pounds/day Table 22, 23 Aφ and Pφ are the conversion factors for attractions and productions of the various metrics of FSA. Aγ and Pγ are the ratios of attractions and productions with respect to the total generation. A A Table 7. Summary of output units and conversion factors.

how to apply the Models 35 FTA in that, generally speaking, it is difficult to consolidate service calls. For this reason, a conversion factor of fA = 1 is suggested. FP: Conversion Between Pounds/Day to Vehicle Trips/Day The outputs of the FP models are in units of pounds/day, so they must be converted into vehicle trips/day. This process is straightforward in the case of large shipments that require full truck- loads, but it is extremely complex for less-than truckloads, which make up the vast majority of shipments in metropolitan areas. The research team attempted to estimate statistical relationships for the latter case using the RPI 2015 survey, which collected both FG and FTG data. The results were not encouraging. As shown in Tables 20 through 23 in Chapter 8, although they are statisti- cally significant, the models have low explanatory power. Additional research is needed to develop better ways to estimate FTG from FG for less-than truckload shipments, which are the norm in metropolitan areas. Only a Generation Model Is Available (Attractions Plus Productions) Because the ER-EB models combine attraction and production in a single estimate, the conversion between shipments and vehicle trips must account for the relative importance of attractions vis-à-vis productions. This can be done using Equation 18: G A A P P (18)( )φ = γ φ + γ φ where: fA and fP are the conversion factors for attractions and productions, and gA and gP are the ratios of attractions and productions with respect to the total generation. In a city where attractions represent 60% of the total generation and productions the other 40%, and fA = 1 and fP = 0.5, the corresponding value of fG will be equal to 0.8 (0.6 × 1 + 0.4 × 0.5). Multiplying the total generation by fG will estimate the corresponding number of vehicle trips. It is important to acknowledge that the conversion factors presented here could be significantly improved. Future research should focus on improving the conversion factors presented. Collecting data that account for the logistical decisions regarding the relations between shipments and vehicle trips would complement FSA modeling. This should be a priority. Step-by-Step Process Table 8 provides an easy way to find the models desired. The rows in Table 8 represent the various metrics of FSA, whereas the columns represent the model type (linear or non-linear). The cells in the table contain the numbers of the tables in this guidebook that correspond to the models. The FP models estimated with the CFS 2007 for the states of California, Ohio, Texas, Wyoming, and the entire United States were submitted for disclosure in November 2015. These models were estimated, in linear and non-linear forms, for 2-digit and 3-digit NAICS, and “all modes” and “road” modes. The flowcharts that end this chapter illustrate a detailed process to help practitioners use the guidebook models to quantify FSA. The discussion starts with cases for which the disaggregate data are available, followed by applications for which only aggregate data are available.

36 Using Commodity Flow Survey Microdata and Other establishment Data to estimate the Generation of Freight, Freight trips, and Service trips Scenario 1: Single Establishment (Establishment-Level Data Available, Linear/Non-Linear Models) For each one of the metrics desired: - Use Table 8 to identify the table(s) with the models you want. - Go to these tables and use the establishment’s industry sector to find the corresponding model. - If no model is found, consider using the “All Industry Sectors” model. - Run the model(s) using the establishment employment as an input. (Note: In some industry sectors, the FSA does not depend on employment. It is constant.) Decide on the metric(s) of freight and service activity (FTG, STG, FG) (see Table 8). Data required: Industry sector (2-digit NAICS) and employment. Scenario 2: Complete Enumeration (Multiple Establishments; Establishment-Level Data Available for All; Linear/Non-Linear Models) Sort the establishment data by industry sector. For each industry sector and metric: - Use Table 8 to identify the table(s) with the models you want. - Go to these tables and use the establishment’s industry sectors to find the corresponding model. (If no model is found, consider using the “All Industry Sectors” model.) - Run the model(s) using the establishment employment as an input. - Repeat for all establishments in that industry sector. - When done with all establishments in the industry sector, aggregate results by sector. Repeat for all industry sectors. Aggregate results. Decide on the metric(s) of freight and service activity (FTG, STG, FG) (see Table 8). Data required: Industry sector (2-digit NAICS) and employment. Linear Model Type Non-Linear Model Type Freight Trip Generation (FTG) (HARTGEN) Table 13 - Freight Trip Attraction (FTA) Table 9 Table 10 Freight Trip Production (FTP) Table 11 Table 12 Service Trip Generation (STG) - - Service Trip Attraction (STA) Table 14 Table 15 Service Trip Production (STP) - - Freight Generation (FG) (RPI) - - Freight Attraction (FA) Table 16 Table 17 Freight Production (FP) Table 18 Table 19 FTG as function of FG (RPI) - - FTA as function of FA Table 20 Table 21 FTP as function of FP Table 22 Table 23 Freight Generation (FG) (CFS 2007) - - Freight Attraction (FA) - - Freight Production (FP) Tables 24-27(NY), 40-43(CA), 56-59(TX), 72-75(WY), 88- 91(OH), 104-107(U.S.) Tables 28-39(NY), 44-55(CA), 60-71(TX), 76-87(WY), 92- 103(OH), 108-119(U.S.) Metric of Freight and Service Activity (FSA) Table 8. Tables where models can be found.

how to apply the Models 37 Scenario 3: Sample Estimation (Multiple Establishments; Establishment-Level Data Available for a Sample; Linear/Non-Linear Models) For the establishments in the sample: - Sort the establishment data by industry sector. - Count the number of establishments in each industry sector. If there are more than 20 estab- lishments per industry sector, the analysis could be done by industry sector. If not, regroup the data in clusters of industry sectors with more than 20 observations each. When creating these groups, try to ensure that they include similar economic activities. For each industry sector (or clusters of) and metric(s): - Use Table 8 to identify the table(s) with the models you want. - Go to these tables and use the establishment’s industry sector to find the corresponding model. (If no model is found, consider using the “All Industry Sectors” model.) - Run the model(s) using the establishment employment as an input. - Repeat for all establishments in that industry sector. - When done with all establishments in the industry sector, aggregate results by sector and com- pute a mean and standard deviation. - Repeat for all industry sectors (or clusters of sectors). - To obtain aggregate results, multiply the number of observations in the entire population by the average values obtained from the sample. - If a grand total is needed (all industry sectors), add the values for all industry sectors. Decide on the metric(s) of freight and service activity (FTG, STG, FG) (see Table 8). Data required: Industry sector (2-digit NAICS) and employment. Scenario 4: Only Aggregate Data (Multiple Establishments; Only Aggregate Data Available; Linear Models) For each industry sector and metric of freight and service activity: - Identify the type of model: Constant (C), Employment Rate (ER), or Combination of Constant and Employment Rate (C-ER) - For model type C: The aggregate value of the metric of interest = (The value of the constant) × (Number of establishments in the industry sector). - For model type ER: The aggregate value of the metric of interest = (The employment rate) × (Total employment in the industry sector). - For model type C-ER: The aggregate value of the metric of interest = (the value of the constant) × (number of establishments in the industry sector) + (the employment rate) × (total employment in the industry sector). - For model type ER-EB: FTG = (FTG rate for bin 1) × (total employment in the industry sector and Bin 1) + (FTG rate for bin 2) × (Total Employment in the industry sector and Bin 2) + (FTG rate for bin 3) × (total employment in the industry sector and Bin 3) - Repeat for all industry sectors. - If a grand total is needed (all industry sectors), add the values for all industry sectors. Decide on the metric(s) of freight and service activity (FTG, STG, FG) (see Table 8). Data required: Industry sector (2-digit NAICS) and employment.