Click for next page ( 33


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 32
32 Responses to these questions can be enlightening. Among How Effective Are Agency Communications? the 11 choices provided by Broward County Transit, responses clustered around three improvements related to Finally, on-board and intercept surveys on occasion measure more service (more frequent bus service, more bus routes, awareness of advertising or agency websites, providing feed- and more evening and weekend service), and two other back to the agency marketing and public information depart- improvements (better on-time performance and bus stop ments. As illustrated in Table 15, questions in this topic area shelters and benches). Agency staff felt that the results may include recall of agency advertising and media used, demonstrated opportunities to increase ridership among cur- channels used to obtain service information, and visitation to rent riders. The results thus provided clear direction toward the agency's website. providing additional service for existing customers rather than adding routes in lower density areas, which had been done previously. QUESTIONNAIRE DESIGN AND LAYOUT Similarly, a LYNX survey found that night and weekend Questionnaire design is a critical element of survey devel- service, and more frequent service, to be the service improve- opment. Although Dillman (2000) reports that mail survey ments that riders felt were most needed. clients often overlook the importance of design, surveys provided by transit agencies display considerable care in question ordering, lay out, font selection, and other design Lists of service attributes may pose a substantial respon- aspects. dent burden because respondents must compare each answer choice with the others and then make a selection Transit staff surveyed cited the use of short question- (Dillman 2000). The respondent burden can be reduced in naires and simple, carefully worded questions as keys to two ways. First, it is important that response choices be as successful on-board and intercept surveys. Surveys pro- concrete as possible and mutually exclusive. Second, vided by transit agencies are often one to two pages long. respondents can be asked to choose the most important (or They typically employ a simple, consistent, linear flow of three most important) items rather than be asked to rank the questions that clearly defines the desired navigation path items. Ranking is a more difficult task. The Broward through the questionnaire. Questions on the same topic are County Transit and LYNX surveys illustrate questions that grouped together. Questions are usually numbered sequen- follow these guidelines. tially, with few if any skips. Instructions are kept to a mini- mum and placed within the question to which they apply. Choosing from lists of service attributes requires that Notably, Dillman (2000) advises following these same respondents be thoughtful and reflective about the reasons design principles in mail surveys. that they use transit and other modes. Responses are subject to rationalizations and a desire by some respondents to answer in a socially acceptable manner. Question Ordering An alternative approach, illustrated by questions devel- Question ordering is a key aspect of questionnaire design. oped from focus group research in New York and Chicago, Dillman (2000) recommends that the first question be easy to is to ask respondents to simply report their experience answer, apply to all respondents, be interesting, and be of using transit. Focus groups in these cities found that clearly connected to the purposes and topic of the survey. riders choose the mode that presents the fewest or least severe problems in terms of travel time, on-time reliabil- Nearly all transit O&D surveys begin with questions ity, safety, cost, comfort, etc. The focus groups also found about the current trip, usually with a question about where that different types of trips (e.g., work and non-work, peak, the trip began. Transit staff starts with O&D questions pri- and non-peak) encounter different types of problems marily to be sure to obtain the trip information, even if (Schaller 1999). respondents do not complete the rest of the survey. This approach appears to be consistent with Dillman's advice to Survey questions were designed to measure the incidence start with questions that apply to all and are clearly connected of problems with transit service. In a CTA on-board survey, with the survey topic and purposes. riders were asked to indicate how often problems occurred for "the type of trip you are making today, considering the On the other hand, trip questions may be of relatively less reason for the trip and your destination." Survey results interest to riders and are not necessarily easy to answer. Dill- identified strengths and weaknesses of CTA service for man suggests starting with attitudinal questions, which are work and non-work trips for different geographic areas. likely to have high salience with respondents, rather than fac- From this information, the agency was able to assess alter- tual questions of less interest. However, it has not been tested native strategies to capture a greater share of total travel whether starting transit surveys with attitudinal questions (Miller et al. 2002). would increase response rates.

OCR for page 32
TABLE 15 SURVEY QUESTIONS RELATED TO AGENCY COMMUNICATIONS AND INFORMATION Communications Pace (suburban Chicago) CTTransit (Hartford, CT) LYNX (Orlando) VTA (San Jose, CA) Do you recall any Where do you receive bus How do you get What are the best ways for advertisements about Pace information? information about LYNX? VTA to get information to service during the past 2 you? (Check up to three.) On bus On the bus months? Telephone Mail VTA Take One Yes--go to 26A. Sales outlet Work VTA website/e-mail No--skip to 29. Work Phone Inside bus advertising Not sure--skip to 29. Online Internet VTA telephone customer Where did you see, hear, service At bus stop (Guide-a- Word of mouth or read them? (Mark all ride) Downtown customer Newspaper/magazine that apply.) service center Shopping center/ TV (specify) Mail convenience store Radio (specify) Information at Other __________ stops/stations Newspaper (specify) Radio (which station?) Other (specify) Newspaper (which newspaper?) Other (specify) Website CTTransit (Hartford, CT) RTD (Denver) visitation Have you visited the Have you ever used RTD's CTTRANSIT website at website to obtain bus trip www.cttransit.com? information? Yes Yes No No 33

OCR for page 32
34 Non-O&D transit questionnaires typically begin with fac- naires. Agencies rely on several simple formatting devices to tual questions about ridership frequency, trip purposes, or aid navigation: customer attitudes about service. Questions begin in the upper left corner; The ordering of attitudinal questions should take into Vertical one or two-column formats; account biases that may be introduced by the order in which Sequential questionnaire numbering from beginning to questions are asked. One source of bias is from the tendency end; of respondents to make their answers to each question con- Bold questions and light answer choices to clearly dis- sistent with answers to previous questions. In an example tinguish questions from answer; from the literature, a study of happiness found that for mar- Check-off boxes used to indicate answer choices; and ried persons, a question on general happiness received more Minimal use of lines between columns or sections of positive responses when asked after a question about marital the questionnaire, thus avoiding visual clutter. happiness than did the general happiness question by itself. This result suggested that "asking the marital question first These practices are consistent with Jenkins and Dillman tended to increase positive responses to the general question (1993) and Dillman's recommendations for mail surveys by causing it to be defined in terms of marital happiness" (2000). (Schuman and Presser 1996). In survey research parlance, this is known as a partwhole question combination in that Transit agencies sometimes use shading to highlight one question is more general and in some sense contains part answer spaces. A Sun Tran O&D survey used dark shading of the more specific question. for questions, light shading for the background to answers, and white boxes for answer choices (see Figure 3 for a por- Analogous transit survey questions pertain to satisfaction tion of the Sun Tran questionnaire). However, write-in with service overall and with specific aspects of service. To answer spaces for addresses and open-ended questions were avoid biases, it is advisable to obtain a rating for overall ser- shaded. The practice of other carefully developed question- vice before ratings questions for reliability, speed of service, naires, such as the 2000 U.S. Census, is to use shading for routing, etc. questions and white space for all answer areas including any fill-in answers (U.S. Census Bureau 2000). Another consistency effect can occur when respondents answer questions in light of an earlier answer on a different Two areas that show a range of different transit practices but related topic. An example in the literature showed slight involve the layout of answer choices and use of matrixes. changes to respondents' self-classification of their subjec- tive social class (upper class, middle class, working class, Answer Choices: One, Two, or Three Columns? or lower class) depending on placement of a question on their education and occupation (Shuman and Presser 1996). As Dillman (2000) notes in the context of mail surveys, It is possible that respondents answered the social class designers of transit surveys sometimes decide to save space question to conform to their educational and occupation by listing answer choices in two or three columns, particu- levels when the education and occupation questions were larly when a large number of choices are involved. Surveys asked before social class. This possible effect can be provided by transit agencies frequently adopt multi-column avoided if the "objective" questions are asked after other formats or list choices horizontally on the same line. Ques- questions or, at minimum, if the questions are separated on tions 13 to 15 in Figure 4 (taken from a GCRTA survey) the survey by other topics. An application to transit surveys provide examples of both horizontal and multi-column would suggest putting automobile ownership and availabil- answer formats. ity questions after questions about alternative modes for this trip. Although multi-column or horizontal formats save space, they also interrupt the vertical flow of the questionnaire. Navigation Guides These practices tend to increase the burden of the respondent, which tends to increase item nonresponse. In addition, Questionnaire layout and formatting provide visual clues to respondents may overlook answer choices in the second or guide respondents in navigating from one question to the third column. The order of choices may therefore influence next and properly marking answers to each question. the results of the survey. The degree to which results are affected by format in transit surveys has apparently not been Most on-board and intercept surveys provided by transit tested; however, the possibility of an effect certainly exists. agencies use simple, clean-looking layouts. The question- naires typically use black type on white or other light-colored The issue is whether these costs in respondent burden and paper, and there is minimal use of shading in most question- impact on results are worth the gain in reducing questionnaire

OCR for page 32
35 How did you GET FROM that place to the FIRST bus you used for this trip? Walked # Blocks Bicycled Dropped off by someone Drove my car and parked What was the FIRST bus you used for THIS TRIP? This is my first bus on this trip. Cat Tran TICET I transferred from Sun Tran Route Please give the bus route # Where did you GET ON the bus you are riding now? Location of the bus stop. Name the cross streets of the nearest corner OR name of the transit center or park and ride. Cross Streets & Street 1 Street 2 Transit center or park and ride Where will you GET OFF the bus you are riding now? Location of the bus stop. Name the cross streets of the nearest corner OR name of the transit center or park and ride. Cross Streets & Street 1 Street 2 Transit center or park and ride Will you transfer to ANOTHER bus on THIS trip to where you are going NOW? No, I will not transfer to another bus. To Cat Tran To TICET Yes, I will transfer to Sun Tran Route Please give the bus route # Where are you GOING TO NOW? Mark One Work School Shopping College/University Home Medical Social/Church/Personal Other What is the address or location of the place where you are GOING NOW? Name the address OR the cross streets of the nearest corner. Address Street # Direction (N,S,E,W) Street Name Cross Streets & Street 1 Street 2 In the City of Zip Code if known What is the name of the PLACE or BUILDING you are GOING TO? example: BANK FIGURE 3 Sun Tran questionnaire with white boxes for answer choices. length. This may be less of a trade-off than it first appears, for Metrolink, and Metra (Chicago area) successfully fielded two reasons. First, length should be viewed as the number of questionnaires of four to seven pages. questions rather than the number of pages in the question- naire. Dillman (2000) reported that squeezing a given number Spacing questions over a larger number of pages may of questions into fewer pages does not improve response rates necessitate a booklet format for the questionnaire. Once a sur- to mail surveys. vey goes over two pages (which can be printed front and back), the questionnaire is easier to open and close if printed as a Second, a number of transit agencies have achieved booklet on oversize paper. GCRTA and TARC surveys used remarkable response rates with rather long surveys. RTD, 11 in. by 17 in. paper, folded to letter pages in booklet style.

OCR for page 32
36 FIGURE 4 Excerpts from GCRTA (top) and Metrolink surveys with skip questions. GCRTA survey also illustrates multi-column and horizontal answer layouts. Matrix Formats Transit agencies reported mixed success with matrix for- mats, however. Although more logical to some respondents, The primary exception to linear question sequences in sur- the matrix formats produce a large block of type at the top of veys provided by transit agencies is an occasional use of the survey. Intended navigation is also less clear, because matrix formats. Rather than having a question and list of respondents are expected to jump from the bottom of the left answer choices, the questions and choices are arranged in part of the matrix to the top of the right part, then down and some type of grid. to the left to continue with the survey. Transit agencies most often use matrix-type formats for The experience of the U.S. Census Bureau is relevant to O&D questions. An example is shown in Figure 5. this issue. Before 2000, the Census Bureau used a matrix for- mat for long-form questions about each household member. There appears to be a natural logic to using a matrix for Based on testing, it replaced the matrix format for the 2000 O&D questions because each question used for the origin is Census. In 2000, the agency asked all the questions about one repeated for the destination. Questionnaires in a matrix- person in the household in a linear format before moving on type design can align the comparable O&D questions. to the next person. The new approach lengthened the ques- A matrix format might also convey the meaning of a one- tionnaire from 20 to 28 pages but, in testing, improved the way trip. response rate by 3 to 4 percentage points (Dillman 2000).

OCR for page 32
37 COMING FROM GOING TO 1a. Where did you come from on this one-way trip? 3a. Where are you going on this one-way trip? (check one best answer) (check one best answer) 01 Home 05 Personal business 08Visiting friends/relatives 01 Home 05 Personal business 08 Visiting friends/relatives 02 Work 06 Shopping 09 Medical appointment 02 Work 06 Shopping 09 Medical appointment 03 School 07 Other: ________________________________ 03 School 07 Other: ________________________________ 04 Recreation: ______________________________________ 04 Recreation: ______________________________________ 1b. Where was that located? Fill in: 3b. Where is that located? Fill in: a street address and the city OR a street address and the city OR a street with the nearest cross street and the city OR a street with the nearest cross street and the city OR a landmark with specific location details, a landmark with specific location details, for example, John's Grocery on Hawthorne. for example, John's Grocery on Hawthorne. (circle one) (circle one) NE SE NW SW NE SE NW SW Street: N S E W __________________________________________ Street: N S E W __________________________________________ Nearest Cross Street: __________________________________ Nearest Cross Street: __________________________________ City: __________________________ Zip Code____________ City: __________________________ Zip Code____________ 2. How did you get to the stop/station for this bus/MAX? 4. How will you get to that location from this bus/MAX? (check one) (check one) 01 Walked _______# blocks 04 Transferred from MAX 01 Walk _______# blocks 04 Transfer to MAX 02 Drove and parked 05 Dropped off by someone 02 Drive 05 Be picked up by someone 03 Transferred from bus #: _____ 06 Other: ______________ 03Transfer to bus #: _____ 06 Other: _________________ FIGURE 5 TriMet 2000 O&D survey using matrix formatting. TriMet moved from a matrix format for O&D questions Branching and Skips in its 2003 and earlier Ridership Surveys to a vertical format in the 2004 survey (see Appendix C). In pretests, TriMet Although transit surveys tend to ask respondents to answer found that the vertical format produced a better response rate. every question, occasionally the need arises to ask a follow- up question to a subset of riders, based on their answer to a On the other hand, Metrolink has been satisfied with a previous question. This need introduces the challenge of similar format for its O&D section. It is possible that the asking respondents to follow instructions for skipping cer- favorable conditions of a commuter rail line (long trips, good tain questions. lighting, an upscale demographic) contribute to this positive experience. For example, Metrolink asked customers who accessed a Metrolink station by bus to report bus company and bus line Aside from O&D questions, transit agencies have used information. GCRTA asked an open-ended question of matrix formats successfully for attitude and fare payment respondents who currently use transit less often than they did questions. Customer attitude questions are often formatted in a year ago (see Figure 4). matrixes with service attributes down the left column and the ratings scale repeated horizontally (see the LYNX Question The difficulty with skips is that respondents may not read 16 in Appendix D). This format saves space and has proven the instructions or follow the arrows. Therefore, respondents easy to complete. Dillman (2000) noted that the format who rode the same amount or more often than they did a year encourages respondents to place each item in a comparative earlier answered the GCRTA skip question, as well as those framework. Respondents think about their rating for on-time who ride less often. In the analysis, GCRTA filtered out the arrivals, for example, relative to their rating for cleanliness. first two groups and focused on open-ended answers from Often, this is exactly the intention of survey designers. If not only the "ride less" group. In this way the agency was able to intended, questions should be asked separately. achieve its research objective. The increasing complexity of many transit agencies' fare Use of Other Languages structures has led some to use matrices for fare payment questions, largely to save space. TriMet reports that cus- Transit agencies often serve a multi-ethnic population that has tomers have become accustomed to this format in the varying proficiency in the English language. A number of agency's annual survey. Ease of responding is probably agencies take for granted that customer communications helped because respondents are being asked to check only should offer customers a choice of English and other lan- one box and by showing the price of each fare option next to guages, primarily Spanish. In the survey of transit agencies, each box. 43% reported using both English and Spanish in the survey

OCR for page 32
38 TABLE 16 RESPONDENT USAGE OF FOREIGN LANGUAGE OPTION Percentage Completed in Other Completed Completed Completed in Languages Agency Survey Project Surveys in Spanish Spanish (no. and percentage) Metro (Los Angeles) 2004 On-Board Customer Satisfaction Survey 14,265 5,125 35.9 Santa Monica (CA) Big Blue Bus Line-by-Line Analysis 4,709 895 19.0 CTA (Chicago) Douglas Branch Blue Line Passenger Survey 1,756 319 18.2 VTA (San Jose) 2000 On-Board Survey 18,351 2,953 16.1 Vietnamese 40 (0.2%) CTA (Chicago) West Side Customer Travel Survey 5,200 398 7.7 Polish 6 (0.1%) CTTransit (Connecticut) Bi-annual passenger survey 4,500 300 6.7 TriMet (Portland, OR) Origin Destination Survey--Systemwide 2000 81,100 4,000 4.9 TriMet (Portland, OR) Annual Fare Survey 15,179 594 3.9 LTD (Eugene, OR) 2004 Origin/Destination Study 5,528 185 3.3 RTD (Denver) Customer Satisfaction Survey 3,654 120 3.3 Metrolink (L.A. area) 2004 Onboard Survey 13,470 323 2.4 Pace Suburban Bus (Chicago area) CSI/User Study 7,937 160 2.0 TARC (Louisville) Project Gobility 4,211 85 2.0