Characteristics of Premium Transit Services that Affect Choice of Mode (2014)

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.

32 This report and the technical appendices that accompany it use terms that refer to concepts from the field of choice-based conjoint modeling. In addition, this report refers to terms that have been specifically defined as part of this research. Because many readers may not be familiar with these and other terms used throughout the report, some definitions are provided below. • Alternative specific constant (ASC). Unobserved behavior in a mode choice model. See also mode-specific constants. • Bias constants. Another term for alternative specific constants, but also more generally applied to any fixed parameter in the mode choice model. • Boarding penalty. An evaluation of premium service characteristics from the research and are cumulated from individual service characteristics but levied as a single modal penalty for each boarding to a given mode as part of a path. • Choice-based conjoint (CBC). A type of behavioral intention research technique in which individuals trade-off (consider jointly) attributes so that the relative importance, or utility, of the attributes can be determined. • Choice modeling. A method to measure the stated preference of a combination of character- istics with respondents choosing the best alternative. • Choice sets. A modeling term that describes the modal alternatives available in the mode choice model. • Direct walk time. Additional time to access premium modes directly. • Equivalent minutes. Term used to value the importance of attributes other than travel time by converting them into time equivalents. • Factor analysis. A method by which many different variables, correlated with one another to different degrees, may be reduced to a set of manageable factors that are orthogonal (not correlated) to one another. • Independence of irrelevant alternative (IIA). A rule that says if A is preferred to B out of the choice set {A,B}, introducing a third option, X, expanding the choice set to {A,B,X}, must not make B preferable to A. In other words, preferences for A or B should not be changed by the inclusion of X; X is irrelevant to the choice between A and B. • Integrated choice and latent variable (ICLV). A type of discrete choice model that integrates latent variable structural equation models to predict simultaneously these unobserved (latent) variables with modal choices. • In-vehicle travel time (IVTT). A component of transit travel time spent on a transit vehicle (bus or train). • Joint bivariate binary probit. A generalization of the probit model used to estimate two cor- related binary outcomes jointly. The binary probit model is a type of regression in which the dependent variable can only take two values. • Latent variable. A variable that cannot be directly observed. Glossary

Glossary 33 • Level-of-service (LOS) variables. Attributes that vary due to the type of service being experienced (e.g., IVTT, wait time, boarding time, number of transfers, cost). • Log-likelihood. A function of the parameters of the mode choice model. The objective of mode choice models is to maximize the log-likelihood; therefore, higher values of log-likelihood are preferred. • Maximum difference scaling (MaxDiff). An analytical technique for determining the relative preference that a respondent has for a set of alternatives. The result of a MaxDiff exercise is a set of values that indicate the respondent’s top choice and last choice, and where along an interval scale, the middle choices lay. Thus, MaxDiff gives more information than simply asking respondents to order a list of alternatives to show their preference. MaxDiff requires respondents to pick the alternative they prefer most and the alternative they prefer least from a short subset of alternatives (usually 3 to 6). By exposing the respondents to different subsets of alternatives and repeating the exercise, it is possible to infer the relative values or utilities that the respondents place on all of the alternatives. • Mode or modal label. Refers to a segmentation of modes by technology (i.e., bus, light rail, etc.) in the mode choice model. • Mode choice model. A model component that estimates the mode of transport that a traveler (or group of travelers) will use for a specific trip. As a choice model, these estimates are probabilities that a traveler will make a trip using a specific mode. • Mode-specific constants. Values in mode choice models that ensure a mode choice model matches a targeted share of trips by mode. Mode constants are adjusted in the model calibration process. Also frequently referred to as alternative specific constants (ASCs), these values are a primary motivation for this study. According to a discussion piece produced by FTA (2006), Perhaps the largest problem for transit forecasting that occurs in traditional model development is a transit calibration effort that results in adjustments necessary to match current data that are no more than correction factors for errors made elsewhere in the model set. The “calibration” of alterna- tive specific constants is meaningful only when the person-trip tables, highway and transit networks, and observed transit ridership patterns are sufficiently accurate. Errors in person-trip tables, in particular, have frequently led to grossly distorted calibration constants that have nothing to do with travel behav- ior and that lead to useless transit forecasts. • Mode-specific variables. Variable transit attributes specific to a particular mode, which may be perceived or weighted differently in the mode choice model utility equation. For example, the travel time coefficient for time spent on commuter rail may have a different value than the travel time coefficient for time spent on a local bus. • Multinomial (MNL) logit choice. An exercise in which individuals select an alternative from a finite set of alternatives. • Non-traditional characteristic or attribute. A transit service attribute that to date has not been able to be described and utilized in mode choice models. Non-traditional attributes include the following: – On-board amenities (seating availability, seating comfort, temperature, cleanliness of a transit vehicle, productivity features); – Station design features (real-time information, security, lighting for safety, shelter, proximity to services, cleanliness of the station, benches); and – Other features (route identification, reliability, schedule span, transit frequency, transfer distance, stop distance, parking distance, ease of boarding, fair machines). The terms attribute and characteristic are used interchangeably in this report. • Path-building. A process to identify the access, route, transfers, and egress elements of a transit trip. Parameters are used to weight the importance of each element. • Premium transit services. A series of attributes that together represent a higher class of service. These attributes exist over a broad continuum of transit services in operation and

34 Characteristics of Premium Transit Services that Affect Choice of Mode are not necessarily associated with a particular vehicle technology. For example, a commuter coach service offering a seat with Wi-Fi service to all customers and a highly reliable schedule may be perceived as superior to a crowded rapid transit rail line with fewer amenities. In this report, an analytical approach and framework is described to acknowledge that these services often exist as a continuum between premium and non-premium; as a result, they are not easily represented as separate and discrete modes. • Probit model. A type of regression model in which the dependent variable can only take two values (e.g., married or not married). • Revealed preference (RP). Questions on the survey that ask travelers to report on a specific reference trip that they have taken recently. • Robust t-ratio. A t-ratio that makes use of standard errors obtained by using a robust variance-covariance estimator also known as the sandwich estimator. A robust t-ratio uses robust variance-covariance estimator accounts for the panel nature of stated preference data (repeated observations from the same individual) and provides consistent estimates even when there is different variability in the data. • Stated preference (SP) questions. Questions on the survey that ask travelers to evaluate a set of hypothetical (although realistically set) assumptions about a set of modes and select the mode they would choose. These questions also can include a set of scenarios in which the traveler is asked to choose a least favorite option and a most favorite option, identified as MaxDiff above. • Transfer penalty. Additional time spent transferring from one mode to another; the value of the transfer penalty depends on the complexity of the paths. • Unobserved factors. See the definition for alternative specific constant. Unobserved factors include traveler attitudes, as they are not directly observable.