Click for next page ( 14


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 13
13 announcements, or other means. The most frequent means on-board and intercept transit surveys is typically composed were on-board announcements, "take-one" brochures, car of bus, rail, and other transit users. The theoretical popula- cards, and signs. Other means included press releases, print tion may be all riders or a subset such as riders on a particu- and electronic media, newsletters, and the Internet. Several lar route or traveling in a particular area or at a particular time agencies commented that the advance notice helped boost of day. response rates. Study Population Incentives The population that the researcher can gain access to is the Incentives are often offered to riders to encourage participa- study population. In a survey of voter preference, the study tion in the survey and at times to encourage full and accurate population might be voters with telephones. In that case, the completion of key survey questions. One-quarter of the study population differs somewhat from the theoretical transit agency surveys used some type of incentive, ranging population because not all voters have telephones. The study from a free pass to a free pen to being entered in a lottery or population in on-board and intercept transit surveys is often drawing. Transit agency staff interviewed believes that congruent with the theoretical population, because all riders incentives do increase response rates and that incentives can, at least in theory, be reached on board transit vehicles or provided to all respondents at the time of the survey are more at transit stops or stations. This is one of the major advan- effective than lotteries or drawings. However, there is little tages of on-board and intercept surveys--the survey can systematic evidence to document this belief or quantify the reach all riders whether or not they live in the service area, magnitude of the effect for transit surveys. or have telephones, or are literate. Extensive research on the use of incentives in mail surveys Three-quarters of the surveys reported by transit agencies provides useful experience in a similar realm. Dillman (2000) defined the study population as "all riders" in some fashion: reported that including $1 or $2 with mail surveys has been all bus riders, all rail riders, all fixed-route riders, or all transit shown to increase response rates by 12 to 31 percentage riders. The remaining surveys defined the study population points. in a variety of ways based on geography or time period; for example: In mail surveys, incentives are more effective when sent with questionnaires rather than as a later reward; "token Riders within a geographic study area, financial incentives of a few dollars" sent with mail Riders on one route or a group of routes, questionnaires "have been shown to be significantly more Riders traveling within a time period (most often week- effective than larger payments sent to respondents who have days), or returned the questionnaires" (Dillman 2000). Incentives Riders traveling through a selected station. provided with the questionnaire constitute a "goodwill ges- In addition to these distinctions, on-board and intercept sur- ture that puts the sponsor and questionnaire in a positive light veys often qualify respondents by age so that children and and sets the stage for the respondent to reciprocate with an sometimes young adults are not included in the study appropriate gesture of completing the questionnaire." By population. contrast, prizes have "relatively small, if any, effect on response" (Dillman 2000). A fundamental research question is whether the study population is defined as riders (people) or trips. Although at IDENTIFYING STUDY POPULATION first glance the differences may seem trivial, they do have AND DRAWING THE SAMPLE significant implications for how the survey is conducted and interpreted. Survey methodologists distinguish between the theoretical population, study population, sampling frame, and sample Focusing on riders is most appropriate for customer satis- (Trochim 2004). These are important distinctions and, faction, attitudinal, and demographic questions where the although not complicated for most on-board and intercept objective is to obtain information on a cross section of surveys, need to be carefully considered. customers. The objective of the survey is to collect informa- tion from individuals who use transit. In the survey process, Theoretical Population customers who are encountered by surveyors more than once are surveyed only the first time. The population that the researchers wish to generalize to is the theoretical population. In a survey of voter preference, for Focusing on trips is most appropriate when the informa- example, the theoretical population would be persons who tion will be used to profile characteristics of trips such as will vote in the next election. The theoretical population in O&D patterns and trip purposes. The objective is to obtain a

OCR for page 13
14 completed survey for each customer trip in the sample frame. Sample Frame Thus, a rider who is encountered by survey workers twice is asked to complete two surveys. The sample frame is the listing of the study population from which the sample will be drawn. In a telephone survey of A further distinction is between trips and trip segments. voter behavior, the sampling frame might be all residential Following the standard practice in O&D surveys, a phone numbers. On-board and intercept transit surveys "customer trip" for this discussion is defined as a journey can define the sampling frame in a number of ways. For between two activities; for example, home to shopping, on-board surveys, the sampling frame is typically based whether one or more than one bus or train is used for the jour- on bus or train routes or vehicle trips (defined as service ney. A "trip segment" is each time the traveler boards a bus from one end of the route to the other). The sample frame is or passes through a turnstile. The distinction is important therefore customers on these listed routes or vehicle trips. because a person making a transfer has twice the chance of On-board bus and rail surveys reported by transit agencies being surveyed than one who does not. The transferring rider generally adopted this approach. can easily be overrepresented in a sample if the intent is to measure customer trips. With an appropriate weighting of One variation to this approach is to survey only high- transferring riders, this issue can be corrected. ridership routes. Broward County Transit's 2003 on-board bus survey was conducted on the 10 routes with the most In theory, then, customer satisfaction and attitudinal ridership. This approach focuses survey resources on surveys would appropriately define the study population as routes that are likely to generate the largest number of "riders." O&D surveys would define the study population as completed surveys. At the same time, there is the possibil- customer trips or trip segments. Many surveys, however, ity that lower-ridership routes not surveyed would produce have both O&D and attitudinal sections. In this case, the different responses. The results should thus be viewed as decision would be based on the importance of the O&D data representative only of riders on the most used routes. and the ability to avoid double counting riders in the attitu- dinal section. Surveys conducted at bus stops or rail stations typically define the sampling frame as stops or stations. For the light One option is to ask riders to complete the O&D section rail portion of the Denver-area Regional Transportation for each trip or trip segment, but to complete the attitudinal District (RTD) Customer Satisfaction Survey, the sampling section only once. This is a workable approach provided the frame was station platforms. Likewise, PATH's O&D survey study population can understand this instruction. defined the sampling frame as all PATH stations. As a practical matter, transit agencies have experienced Another approach is to survey at centralized nodes, such difficulty in enlisting riders to complete more than one as transit centers or transfer facilities. The advantage of survey. Completing multiple surveys may seem redundant this method is that riders from a cross section of routes can to customers and overly burdensome. Therefore, although be surveyed without having to disperse surveyors over O&D surveys may request that customers complete multiple every route. GRPC conducted interviews at a transit center surveys, many do not actually do so. This may not be a sub- and four transfer facilities for this reason. From a data stantial problem in a large agency where customers are accuracy standpoint, this approach is most appropriate if unlikely to be encountered by survey workers more than the large majority of customers pass through one of these once. However, it can be a substantial problem with large central nodes. survey efforts in mid-size or smaller agencies. In either case, the issue of riders not wanting to complete more than one sur- The sampling frame may be further defined by time of vey is one reason that agencies emphasize weighting of O&D day, choice of weekday or weekend, and direction of travel. surveys by route, direction, and/or time of day, as shall be These further refinements are designed to ensure that a rep- discussed later, to attempt to offset the possible biasing resentative sample is drawn for each segment; for example, effects of self-selection in the survey process. each line, direction, or time-of-day combination. The use of this sampling frame in a stratified sample is discussed later A final comment on this issue concerns the importance in this chapter. of being clear in the presentation of survey results. Although it may be tempting to speak generally about the survey showing how riders feel or how they travel, care Sample should be taken to identify the study population. A survey in which the study population is riders, for example, can The sample is the group of people selected to be surveyed. refer to riders who used transit on the specified survey As Trochim (2004) points out, the sample is not the group of dates. A survey that uses trips as the study population can people that actually complete the survey. Rather, the sample be clear that results are specific to trips, with some riders is the group of people that the researcher attempts to contact represented multiple times. or recruit.

OCR for page 13
15 The sample may consist of the entire sample frame. For are charged with distributing surveys to attempt to distrib- example, Pace Suburban Bus distributed its Customer ute the surveys to a cross section of riders. The integrity of Satisfaction Index/User Survey on all Pace buses over a the sample can be affected by this approach, although 3-day period. More often, however, a subset of the sample clearly there may be logistical and budgetary reasons frame is selected, typically through simple random sample necessitating it. procedures. For example, Los Angeles County Metropolitan Transportation Authority (Metro) created a digital file of all bus trips and used computer software to randomly select bus Stratified Sampling trips to assign to survey workers for an on-board customer satisfaction survey. Similarly, for a systemwide survey, Although most on-board and intercept surveys reported by MARTA randomly selects 1,000 bus and rail trips to be transit agencies used a simple probability sample, approxi- surveyed each year. mately one-quarter of these agencies reported using a stratified sample. The objective of stratification is to ensure that key Simple random samples, as in these examples, are method- subgroups of the population are represented in the overall ologically attractive because they maintain the basic principal population. Among on-board and intercept surveys, stratifica- of probability sampling: that each unit has an equal chance of tion is most often applied to time of day and route. The objec- being selected. Simple random samples can be difficult to tive is to ensure that peak and non-peak hours, or each route, field, however. Random selection of routes (or bus stops or are represented in the sample and to control for variations in stations) can result in survey workers spending an inordinate response rate within strata. amount of time moving between assignments. In a stratified sample, the sample frame is divided into MARTA circumvents this problem by filling in surveyor mutually exclusive and exhaustive subsets. A simple random time with other assignments through the course of the year. sample of elements is then chosen independently from each Another approach is to group bus trips, for example, to form group or subset. A San Diego Association of Governments coherent itineraries for survey workers. Each "package" of (SANDAG) On-Board Transit Passenger Survey, which bus trips is then run through the random selection procedure. covered 164 bus and rail routes in a large geographic area, Santa Clara Valley Transportation Authority (VTA) used was stratified based on route, direction, and time of day. this approach for its 2000 On-Board Survey. Surveyors were assigned trips within each of four time periods for each route and direction combination. Although the sampling frame may consist of bus trips, subway stations, or light rail platforms, the sample itself In deciding how to stratify, a main goal is to divide the inevitably consists of riders. The transition from vehicle trips sample frame into relatively homogeneous groups. In the or stations to riders is generally straightforward. In self- case of bus routes, agencies believe or assume that bus riders administered surveys, the sample typically consists of all on a given route and rail riders at a given station and time persons riding the buses or trains that are in the sampling period are relatively similar. frame. However surveys are distributed, the transit agency attempts to enlist participation from every rider. Some agencies stratify at a more generalized level. For example, the Regional Transportation Commission of Southern At times, however, it is necessary to select from within the Nevada stratified based on quadrants of the agency's service group of people boarding the bus or otherwise falling into the area. The Ann Arbor Transportation Authority (AATA) strati- sample frame. If survey workers are conducting personal fied for five groupings of bus routes. interviews, for example, it is probably not possible for them to interview every person who boards the bus or passes The same sampling fraction can be used within each through a station. Without a carefully implemented method strata; for example, sampling one in every 10 bus trips of selecting which persons are approached, there is an acute for each route, time of day, and direction combination. chance that surveyors will gravitate toward persons who Almost universally, transit agencies reported using the same appear friendly and/or are similar to the surveyor. Transit sampling fraction for each strata, which is called a propor- agencies typically attempt to maintain the randomness of the tionate stratified random sample. sampling procedure by selecting every nth person. Thus, the King County Metro Ride Free Area survey selected for inter- An alternative is to use different fractions to produce a viewing every third rider that boarded during non-peak times disproportionate random sample. This approach provides and every fifth rider that boarded during peak times. statistically significant results for relatively small subgroups; MARTA surveyors interviewed every fifth person to board for example, people using transit during late night hours. the bus or train for a 5-min interview on travel patterns. A proportionate stratified sample would likely produce too few surveys from this group to be statistically meaningful. Not all agencies adopt a strict sampling procedure. TriMet oversampled light rail riders to obtain a meaningful Some agencies reported encouraging bus operators who number of responses from light rail riders.