8th INTERNATIONAL CONFERENCE ON SURVEY IN TRANSPORT
ANNECY, FRANCE - MAY 25-31, 2008
Alternative Ways of Measuring Activities and Movement Patterns of Transients in Urban Areas: International Experiences
Aloys Borgers, Chang-Hyeon Joh, Astrid Kemperman, Shigeyuki Kurose, Dick Saarloos, Junyi Zhang, Wei Zhu, and Harry Timmermans
Aloys Borgers, Astrid Kemperman, Harry Timmermans and Wei Zhu: Eindhoven University of Technology, Urban Planning Group, PO Box 513, 5600 MB Eindhoven, The Netherlands, phone +31 40 247 3315, fax +31 40 243 8488, e-mail: a.w.j.borgers@ tue.nl / a.d.a.m.kemperman@ tue.nl / / w.zhu@ tue.nl
Chang-Hyeon Joh: Kyung-Hee University, Department of Geography, 1 Hoegi-dong, Dongdaemun-gu, Seoul, 130-701, Korea (South), Tel 82 2 961 9264, Fax 82 2 961 0251, E-Mail:
Shigeyuki Kurose: Fukuoka University, Department of Architecture, Fukuoka, 810-0018, Japan, Phone: +81 92 871 6631, Fax: +81 92 865 6031, Email:
Dick Saarloos and Junyi Zhang: Transportation Engineering Laboratory, Graduate School for International Development and Cooperation, Hiroshima University, 1-5-1 Kagamiyama, Higashi-Hiroshima, 739-8529, Japan, Phone & Fax: 81-824-6919, E-mail:
ABSTRACT
Five methods of measuring pedestrian movement can be distinguished: tracking, observing by means of cameras, interviewing, questionnaires and using high tech equipment like RFID, GPS, cell phones and VR. This paper reports on and discusses the collection of data of pedestrian movement and activities in central shopping areas in cities in the Netherlands, China, Japan, and South Korea. The data have been collected by means of interviews or questionnaires. Aspects regarding sample, survey instrument and design, administration, data cleaning, and experiences have been dealt with. Next, questions concerning the timing and location of intercepting respondents, how to elicit destinations visited and routes walked through the study area have been discussed. It has been concluded that in order to improve data quality pedestrians should be intercepted when leaving the study area and immediately be interviewed on the spot.
Keywords: Pedestrian movement, Interviews, Questionnaires, Central shopping areas
INTRODUCTION
The commercial viability of city centers often depends on shoppers, spending money in the area. Some of these are tourists, more are out-of-towners, most are transients in the sense that they do not live in these city centers. In order to estimate the impact of these shoppers and examine their activity-travel patterns, data on their activities, timing, destination choices and route choice is required. Especially if the focus is on city centers, detailed information is needed to develop a model of shoppers’ activity-travel patterns and estimate their impact on turnover and viability. The question is how data on pedestrians’ activities and movement patterns can be collected.
In the literature, different methods to measure activities and movement of pedestrians in shopping environments have been used. These methods can be categorized into five types: (i) tracking pedestrians (e.g. Zarachias, 2000), (ii) observing pedestrians by means of cameras (e.g., Teknomo, 2002; Bierlaire, et al., 2003), (iii) interviewing, (iv) questionnaires (e.g. Lorch Smith, 1993), and (v) using high tech equipment like RFID, GPS, cell phones and even virtual reality (e.g. Tan, et al. (2006), Tanaka, & Shibasaki, 2005). The main advantages of tracking and observing pedestrians is the reliability of the data. If a pedestrian can be traced or observed unobtrusively during the full route through the shopping area, all activities and route choice decisions as well as duration can be recorded. However, the major disadvantage is that additional information regarding intentions, planned activities, expenditures, etc. cannot be observed. If the focus is on tourists, the lack of such motivational and attitudinal information may be quite disturbing. Even if the focus is on regular shoppers coming in from a larger area, the need for additional information will still be present. Furthermore, these methods are time consuming and risky in the sense that trackers or observers might loose the pedestrian while recording the shopping trip.
In contrast, interviewing pedestrians or asking them to fill out questionnaires does offer the opportunity to collect such additional information. On the other hand, some scholars (see e.g. Brown, 1992) found that respondents are not able to correctly report all their activities conducted in the shopping area. This problem of forgetting activities will increase with increasing time between pedestrians starting the shopping trip and being interviewed or asked to fill out the questionnaire.
Inviting pedestrians to carry high tech equipment to register their behavior may generate highly accurate spatio-temporal data and save much effort in data collection, however, at the expense of limited other information and lengthy time. Inviting pedestrians to participate in virtual reality experiments allows developing various scenarios to investigate different responses with the major question how consistent is their behavior in a virtual world with their behavior in the real world (Zacharias, 2006; Tan, 2006). However, it is unlikely that such experiments can be conducted for large samples, unless a huge budget is available. Moreover, as people need to come in to a laboratory, this means of data collection is not an option for infrequent visitors.
In this paper, we will concentrate on interviewing and questionnaires, as these methods offer the opportunity to collect additional information. These methods involve typical problems like (i) What is the best way of intercepting respondents to guarantee the representativeness of the sample? (ii) Where should the survey points be located, at the entrances/exits of the area or distributed within the area? (iii) Should one choose to conduct interviews when they start their activities, when they finish their activity agenda, or somewhere in between, or should one choose to hand out questionnaires that are to be completed elsewhere? (iv) In both cases of on-street interviews and return-mail questionnaires, how can we limit the time that a respondent will need to participate and, thus, how much detail can be asked? (v) On which days and for how long should the survey go on? (vi) What is the best way of locating destinations and eliciting routes? When the layout of the center allows this, a cordon would allow one to count the number of pedestrians and also to interview them when they arrive or leave. Obviously, in the first case, data collection would reflect their plans for activities, routes, etc. In the latter case, one would opt for obtaining completed patterns.
Collecting data on activity-travel patterns means gathering information about which activities are conducted where, for how long, with whom, when, the route that was followed, how much money was spent on what, etc. In case of questionnaires, this may take more time and therefore response rates may decrease. Moreover, especially for respondents who are less familiar with the center, it may be difficult to report such detail as they can rely less on memory, internalized cues and names, etc. In this sense, if practically possible, intercepting and interviewing pedestrians directly when leaving the area would improve reliability. The same holds for the issue of how to ask for route information. Maps may be appropriate, but some people have difficulty in reading and understanding maps, and this potential problem may be more severe if they are relatively new to the city. In case of questionnaires, there will be no interviewer to assist in correctly locating destinations and drawing routes.
Recently, the authors have collected data of pedestrian activities and movement patterns in a variety of cities, including Maastricht and Eindhoven (The Netherlands), Hiroshima and Fukuoka (Japan), Busan (South Korea), and Shanghai and Beijing (China). These cities differ in a variety of ways: some are largely pedestrianized while others are not or only partially; there are substantial differences in area size, crowdedness and functional characteristics; and they also differ in terms of morphological pattern and accessibility by car and public transport. Also some of these cities attract tourists, whereas others do not really. Survey designs for these data collections were in some cases very similar, but in other cases different choices were made. Much experience has been gained in designing, administering, cleaning and analyzing these data collection efforts. The reason for data collection differed: in some cases, data were collected for academic reasons only; other studies were commissioned or conducted in collaboration with local planning authorities.
The current paper will (i) report similarities and differences in data collection; (ii) give basic statistics describing the various samples, and (iii) address the specific questions raised above. Based on these experiences, guidelines for collecting data of shoppers’ behavior in city center areas will be developed. Finally, we will reflect on the future improvements that can be made through the use of modern technology such as RFID, cameras, virtual reality and GPS-enabled phones.
CASE STUDIES
East NanJing Road, Shanghai and WangFuJing Street, Bejing, China
Area. East NanJing Road (ENR) is located in the city center of Shanghai. Chinese people call it “The No. 1 shopping street” in China both for its historic position and its current symbolic meaning for Chinese retailing. It is also a famous tourism site so that there is a saying “You don’t really have been in Shanghai if you haven’t been to NanJing Road”. Its attraction and importance continue to grow since part of ENR was pedestrianized in 1999 and more modernized large scale retail stores appear. The street is about 1,600 meters long, and 1,000 meters of this is pedestrianized. People’s Square, a multifunctional place for gathering, leisure, shopping, and museums, is located in the western end of the street. The eastern end locates the Bund, an internationally famous tourism site featured for buildings of the early 20th century. There are two metro stations in the area, one near People’s Square, and the other in the eastern end of the pedestrianized section. Most of the stores are located along both sides of the street, which is largely shaped as a linear shopping space.
If ENR is the symbol of retailing in Southern China, WangFuJing Street (WFS) is the symbol of the North. Its history can be traced to 700 years ago. Like ENR, WFS was also partly pedestrianized in 1999. The street is located in the historical center of Beijing city. The Forbidden City and Tiananmen Square is 1,000 meters to its west. The cultural center is only several hundred meters to its north. The shopping space is also linear, stretching 1,200m from south to north. The pedestrian section is about 530m. A metro station is located in the southern end.
Sample. In Beijing and Shanghai, 760 and just over 810 respondents were interviewed respectively. Response rates are not available as non-response was not registered. This is quite common in this type of research. In principle, it could be avoided by giving additional instructions to the interviewers.
Survey instrument. Both surveys used on-street questionnaires administrated by 20 undergraduate students during every survey day.
Design issues. The major purpose of both projects was to understand pedestrian spatio-temporal behaviors in shopping environments. In addition to basic personal information, the major questions relate to pedestrians’ activity diaries, where they entered the shopping area, at what time, their sequences of visits in stores, what they bought, how much time they stayed in every store, how much money they spent, the time they were going to end their shopping trip, and exit point.
Administration. Three surveys have been conducted in ENR since 2001. The latest one was carried out during two days of 2007, May 19 (Saturday) and May 22 (Tuesday), each day from 12:00 to 20:00. These days represent normal days in a year and no special events happened. The survey in WFS was conducted May 17 (Monday) and May 22 (Saturday) 2004. Most of the time during the survey the situations were normal but after 17:00 on Monday the survey had to be terminated due to unexpected rainfall, which led to a smaller sample size than planned.
Respondents were selected randomly. In the WFS case, only respondents who reported they were near the end of their shopping trips were invited to participate in the survey. In the ENR cases, respondents were selected randomly as well. However, in 2001, respondents were not selected regarding the status of their trip. Respondents were asked to report their trip performed so far and to describe the plans for the rest of the trip. In 2003, only respondents who (almost) finished their trip were invited to participate. In 2007, the 2001 strategy was adopted again, but now the status of the trip (“Just started”, “In the middle”, or “Almost finished”) was asked for as well.
Because both WFS and ENR are linear in shape, sampling points were distributed along the streets. However, due to practical reasons, almost all the administrators in the ENR case were arranged in the pedestrian section because there are no resting facilities in the non-pedestrianized section, which makes it very difficult to intercept pedestrians when they are moving. As can be expected, most of the time administrators asked pedestrians who were resting in the pedestrian section to participate because the questionnaires require some effort to complete. Such arrangement may cause some biases in the distributions of pedestrian activities in the street. For the same reason, about 2/3 of the administrators in the WFS case were allocated in the pedestrian section since there are some resting facilities in the non-pedestrianized section as well. Each interview cost about 15 – 20 minutes to finish.
Based on reported sequences of store visits, it is not difficult to elicit pedestrians’ routes since both streets are linear. However, this method cannot guarantee that pedestrian did not move to somewhere outside the space between origins and destinations. This limitation was considered in the ENR survey by recording “turning points” where pedestrians changed their walking directions. We also tried to elicit shopping durations for the WFS case because they were not recorded. This was implemented by subtracting the estimated walking time (assumed walking speed: 1 m/s) from the time between reported start time and end time. The remaining time was averaged over the destinations visited by the respondent.
Data cleaning. Data were checked for consistency; no irregularities were detected. Missing answers were coded as such.
Experiences. Comparative analyses showed that different respondent selection strategies may lead to different results and have respective pros and cons. Firstly, by comparing the three ENR datasets (2001, 2003, 2007) regardless of selection strategy, we observed quite stable pedestrian behavior, which shows the value and validity of these methods in general. Except the socio-demographic changes which may be caused by the retail-developments in ENR itself, results showed similar number of store patronages, similar average expenditures, and similar activity times. This might suggest that pedestrians are able to predict the remaining part of their trip rather good. Secondly, selecting respondents only at the end of the trip (the WFS case and the ENR 2003 case), could cause biases in the sample because those pedestrians who arrived relatively late and had not finished their trip when interviewers stopped interviewing, were excluded from the sample. As these respondents probably spend less time and money because of their late arrival, the overall activity time and expenditures are overestimated. Respondents may not be very good at reporting activity durations. In the ENR case, we found quite much underestimation of durations by respondents and accordingly unrealistically long walking times.