OVERCOMING THE DESIGN AND FIELDWORK CHALLENGES FOR MODELLING THE CHOICE OF MODE OF TRAVEL TO THE ISLES OF SCILLY

Transcription

1 OVERCOMING THE DESIGN AND FIELDWORK CHALLENGES FOR MODELLING THE CHOICE OF MODE OF TRAVEL TO THE ISLES OF SCILLY Peter Burge, Charlene Rohr, Marco Kouwenhoven, Stephen Miller RAND Europe Rachel Risely Accent 1. INTRODUCTION The Isles of Scilly are located 28 miles off the south west coast of England. The Isles comprise of five inhabited islands, with a combined population of about 2000, and many smaller uninhabited islands and rocky islets (Figure 1). The main industry on the Isles is tourism, which is highly dependent upon the transport links to and from the mainland. At present there are three commercial services operating between the Isles and the mainland: a sea ferry, a helicopter service and fixed-wing aircraft services. However, the boat used for the ferry service, The Scillionian, is now nearing the end of its operational life and it is planned that the sea ferry will be taken out of service after Airplane Airplane Airplane Isles of Scilly Helicopter Ferry Figure 1 The Isles of Scilly and the local transport links Cornwall County Council submitted a bid to the UK Department for Transport for capital funding support for improved transport links between the Isles of Scilly and the mainland. Within this bid a number of different investment schemes were developed to replace the current ferry with a new (and possibly faster) one as early as To support this bid, the Cost Benefit Assessment (CBA) for these investment schemes required predictions of travellers demand and mode shares to be made. A model of travellers mode choice between ferry, helicopter and fixed wing aircraft was therefore required. In this paper we describe two particular groups of challenges facing the data collection required to support the development of a mode choice and a trip

2 frequency model (Kouwenhoven et al 2006). The first are the technical design challenges resulting from the differences in the nature of the trips and modes to those considered in the majority of mode choice models. The second are the fieldwork challenges that undertaking the surveys introduced. We conclude the paper by discussing the lessons learnt with the intention of providing guidance to those designing future exercises to collect data for demand and mode choice models for islands and tourism destinations. 2. THE DESIGN CHALLENGES In this section we discuss the challenges faced in designing the surveys. In the first two sections we describe the general structure of the surveys, before moving on to discuss the specific challenges that required addressing. 2.1 Understanding existing travel behaviour In order to be able to estimate a robust model, it was first necessary to obtain a better understanding of what is driving the current decisions that travellers (both visitors and residents) make when choosing between the services available. As a result, it was decided to undertake a series of revealed preference interviews, which would provide insight into the mode choices that travellers were already making. For the purposes of interviewing, two separate questionnaires were developed to understand existing trip making behaviour, one for visitors to the Isles of Scilly and one for island residents. These questionnaires contained similar questions on trip making, but were tailored to the specific context of the visitors or residents. The non-resident visitors were interviewed during their trip to or from the Isles. They were sampled both on the ferry and at the airports for the airplane and the heliport for the helicopter alternative. In order to understand existing travel behaviour we asked visitors to report on the circumstances of their trip to the Isles during the interview, including: the origin and destination of the journey; currently used mode of travel, i.e. ferry, helicopter or fixed-wing aircraft; (detailed) purpose of travel; party size, including whether there are children in the party; journey details, such as access/egress times to terminal; journey access information for the modes that were not used; fare/ticket type; trip/holiday duration; frequency of travel to/from the Isles.

3 The survey also collected background details about the travelling party, including: age; gender; employment status; household income; home ownership on the Isles of Scilly. Island residents received a similar questionnaire by mail. They were asked to describe the details of their most recent trip made for business purposes (for self-employed people a business trip was defined as a trip for which they were able to claim tax rebates for the fares for their journey) and also the details of their most recent trip for non-business purposes. 2.2 Understanding future travel behaviour The revealed preference surveys were complemented with research into the decisions that travellers to and from the Isles of Scilly may make in the future when the ferry service may differ from that currently operating. The key concern here was to obtain insight into how the demand for different modes may vary if the characteristics of the modes varied, e.g. increased fares, reduced travel times, increased comfort, etc., or whether the available modes were different, specifically the impact on demand if the ferry were no longer available. An appreciation of the impacts of such changes was necessary to inform the various options in the Cost Benefit Assessment (CBA) for the bid for capital funding support. To achieve this, an additional survey was designed incorporating both stated preference discrete choice exercises and stated intention questions. Travellers were recruited for this survey from the respondents to the revealed preference surveys that were used to examine existing travel patterns. They were re-contacted by telephone at their home address, subsequent to their trip, and sent show material by post to allow a detailed questionnaire to be undertaken over the telephone. In this survey respondents were asked to undertake three stated preference exercises: a within-mode experiment in which respondents were asked to choose between two hypothetical ferry alternatives for travel; a between-mode experiment in which hypothetical ferry, airplane and helicopter alternatives were compared; in addition a wouldn t make a trip alternative was available; a between-mode experiment in which hypothetical airplane and helicopter services only were available (no ferry was available) again a wouldn t make a trip option was available. At the completion of the stated preference choice exercise, stated intention questions were asked to inquire about the respondent s current journey

4 frequency to and from the Isles and how this may vary under certain scenarios, specifically with an improved ferry service and with no ferry service. The within-mode experiment was used to examine the preferences of travellers with respect to different ferry services. This was important because some of the proposed options within the CBA would allow for improvements in the existing ferry, both in terms of reductions in ferry journey time and the potential to increase ferry sailing frequency. In this experiment all respondents were asked to compare two hypothetical ferry alternatives. The ferry alternatives were described by six variables: sailing frequency, journey time, journey ambience (i.e. characteristics of the ferry), quayside facilities in Penzance and at St. Mary s and cost. Figure 2 shows an example choice situation from the within mode experiment. Choice 5 If only a ferry service was available when you travelled to the Isles of Scilly, which of the following would you have preferred? Ferry A Journey time is 2 hours 40 minutes Ferry B Journey time is 2 hours Penzance - St. Mary's 07:00 Penzance - St. Mary's 08:00 St. Mary's - Penzance 10:30 St. Mary's - Penzance 17:30 Penzance - St. Mary's 14:00 St. Mary's - Penzance 17:30 Existing boat, passengers may be prone to seasickness in rough conditions Modern boat, passengers less prone to seasickness New passenger terminal building and covered walkway to the vessel at Penzance New shelter canopy at St Mary's The return ferry fare for your journey is No passenger terminal building and exposed quayside area at Penzance New shelter canopy at St Mary's The return ferry fare for your journey is Prefer Ferry A Neither Prefer Ferry B Figure 2 Example within mode choice The levels in the within-mode experiment are summarised in Table 1. It is noteworthy that a large number of fare levels were investigated (8 levels around the actual fare) in order to examine a wide-range of valuations for ferry travel time savings. Nine choice situations were presented to each respondent in the within-mode experiment. One of these nine was a dominant question, i.e. where all the attributes of one option were better than the other. This question was included to allow the identification of respondents who did not appear to answer the questions in a sensible manner. However, the dominant question responses were not intended to be used in the model analysis, leaving eight choice situations per person for analysis.

5 Table 1: Variables Ferry journey time Ferry sailing schedule Ferry ambience Penzance quayside facilities St. Mary s quayside facilities Fare Within-mode experiment variables and levels Levels 1. 3 hours 2. 2 hours 40 mins 3. 2 hours 20 mins 4. 2 hours 1. There is 1 sailing per day Leaving Penzance 09:15 Leaving St Mary s 16:30 2 There is 1 sailing per day Leaving Penzance 08:00 Leaving St Mary s 17:30 3. There are 2 sailings per day Leaving Penzance 08:00 Leaving St Mary s 11:15 Leaving Penzance 14:30 Leaving St Mary s 17:30 4. There are 2 sailings per day Leaving Penzance 07:00 Leaving St Mary s 10:30 Leaving Penzance 14:00 Leaving St Mary s 17:30 1. Existing boat, passengers may be prone to seasickness in rough conditions 2. Modern boat, passengers less prone to seasickness 1. No passenger terminal building and exposed quayside area at Penzance 2. New passenger terminal building and a covered walkway to the vessel at Penzance 1. Limited shelter at St Mary s 2. New shelter canopy at St Mary s % of base fare % of base fare % of base fare % of base fare % of base fare % of base fare % of base fare 8. 90% of base fare Respondents were then asked to participate in a between-mode experiment containing hypothetical choices between the three modes available to travel to the Isles of Scilly ferry, helicopter, and fixed wing aircraft. Respondents were presented with 8 choice scenarios. Figure 3 shows an example choice card from the between-mode experiment. It is noted that the order of the presentation of the modes was varied between respondents, in order to reduce the possibility of a response bias, e.g. that respondents would always prefer the left-side alternative. The levels of the variables used within this experiment are presented in Table 2. The access time for each mode was also presented in the choice exercise (as reported for each mode by the respondent), but this was not varied in the experiment. Helicopter and airplane journey times were also presented, but these were not varied.

6 Choice 10 If the following alternatives had been available when you travelled to the Isles of Scilly, how would you have travelled? 7 Ferry Helicopter Airplane (Penzance) (Penzance) (Bristol airport) 3 hours 20 minutes travel by car to the quay 3 hours 20 minutes travel by car to the heliport 1 hour travel by car to the airport Journey time is 3 hours Journey time is 20 minutes Journey time is 1 h 10 min Penzance - St. Mary's 09:15 St. Mary's - Penzance 16:30 Existing boat, passengers may be prone to seasickness in rough conditions No passenger terminal building and exposed quayside area at Penzance The return ferry fare for your journey is There are frequent flights all day The return helicopter fare for your journey is There are 1 or 2 flights on a weekday, and 3 on Saturdays The return air fare for your journey is Choose Ferry Choose Helicopter Choose Airplane Wouldn't make trip Figure 3: Example between-mode choice situation Table 2: Between-mode experiment variables Ferry Helicopter Airplane Access Time to Ferry Quay (not varied) Access Time to Heliport (not varied) Access Time to chosen Airport (not varied) Ferry journey time (levels: see Within experiment) Ferry sailing schedule (levels: see Within experiment) Ferry ambience (levels: see Within experiment) Quayside facilities at Penzance (levels: see Within experiment) Helicopter journey time Journey time is 20 minutes (fixed level, not varied) Helicopter schedule: There are frequent flights all day (fixed level, not varied) Airplane journey time: Land s End: 15 minutes Newquay: 30 minutes Exeter: 1 hour Bristol: 1 h 10 min Southampton: 1h 30 min (fixed level, depending on chosen airport in RP): Airplane schedule: Land s End: There are frequent flights all day Newquay: There are 3 flights on most weekdays, and 4 on Saturdays Exeter: There are 1 or 2 flights on a weekday, and 3 on Saturdays Bristol: There are 1 or 2 flights on a weekday, and 3 on Saturdays Southampton: There is 1 flight a day from Monday to Wednesday, and on Friday ) (fixed level, depending on chosen airport in RP):

7 Fare % of base fare % of base fare % of base fare % of base fare % of base fare % of base fare % of base fare 8. 90% of base fare Fare % of base fare % of base fare % of base fare 4. 80% of base fare Fare % of base fare % of base fare % of base fare 4. 80% of base fare Respondents were then asked to consider a hypothetical situation where there was no longer a ferry service available for travel to the Isles. They were then presented with 4 more choices where there was no ferry service. Figure 4 shows an example choice card for the between-mode choice experiment without any available ferry service. Choice 18 If the following alternatives had been available when you travelled to the Isles of Scilly, how would you have travelled? 7 Ferry Helicopter Airplane (Penzance) (Penzance) (Bristol airport) 3 hours 20 minutes travel by car to the heliport Journey time is 20 minutes 1 hour travel by car to the airport Journey time is 1 h 10 min Not available There are frequent flights all day There are 1 or 2 flights on a weekday, and 3 on Saturdays The return helicopter fare for your journey is The return air fare for your journey is Choose Helicopter Choose Airplane Wouldn't make trip Figure 4: Example between-mode choice situation no ferry At the completion of the stated preference experiments, respondents were asked a number of questions about their frequency of travel to (or from) the Isles of Scilly, over the last 12 months, for both business and leisure purposes. They were asked to state how many trips they would make in the next 12 months, again for both business and leisure purposes, in the following scenarios: given the current service characteristics (status quo); in cases where the ferry services to the Islands were improved, specifically that: o a new modern ferry was introduced which could undertake the journey in 2 hours, rather than in 2 hours and 40 minutes, as is the case for the current trip o on the new ferry passengers would be less prone to seasickness

8 o the new ferry would be able to make 2 round trips per day to the Islands; this would mean that people making a day trip to the Isles would get about 6 hours on the Islands. o there would be improvements at the quayside in Penzance and St. Mary's, namely a new passenger terminal building and covered walkway to the vessel at Penzance, and a new shelter canopy at St. Mary's; in cases where the ferry service was stopped. These stated intentions questions were used to quantify trip frequency responses to changes in (ferry) travel service characteristics. Respondents were also asked questions about how their travel may change if there was a regular Sunday ferry services, if the ferry services ran for all twelve months of the year and if there were Sunday services for helicopter and airplane services. 2.3 Dealing with the diversity of journey origins The nature of the travel being undertaken posed some significant challenges. In many mode choice studies the researcher is dealing with a corridor of travel where the majority of journey origins and destinations are within a relatively confined area. In the case of travel to and from the Isles of Scilly the destinations on the islands are quite restricted. However, those on the mainland are incredibly varied, with visitors travelling to the Isles of Scilly from across the UK (and from other countries). As a result there is significant diversity in the journey origins (or destinations in the case of islanders travelling to the mainland) and therefore a wide range of variation in the associated time spent in access modes prior to the 2-3 hour hop over to the islands. It was possible to utilise the revealed preference surveys to ask respondents how they made their current journey, collecting information on their observed mode of travel and the journey time/cost. The difficulty however was in specifying the access/egress information for the alternative modes. For example, if a respondent was making a journey from London to Penzance for the ferry, he/she would be in scope for the survey. Access time for the helicopter (also from Penzance) would be approximately the same if they are travelling by car, and would be slightly longer if they are travelling by train (there is a dedicated Heliport mini bus from railway station in Penzance). However, the problem was to identify which fixed wing service they would use, if, for example, there was no ferry service (one of the possible options), as there are many airports across the south of the UK providing services to the Isles of Scilly. Therefore it was not possible to assume that all travellers would select the services flying from Land s End. As a result it was decided to show respondents a map of the potential airports from which they could fly and ask them, if they were to fly, which airport they

9 would use and approximately how long they believed it would take to get there. It was anticipated that the impact of ferry improvements would be quite localised, i.e. that it would be unlikely that people making fixed-wing aircraft journeys from Southampton airport, for example, would switch to ferry if the journey time was reduced from 2 hours 40 mins to 2 hours. As a result it was suggested that interviews with fixed wing travellers were only necessary for those who are making existing aircraft journeys from Land s End (and possibly Newquay), but that travellers should not be included if they were making journeys to/from Exeter, Bristol, or Southampton. Table 3: Number of trips by purpose and mode (2003) Total Passengers Staying Visitors Day Visitors Residents/VFR/Bus /Other No. % No. % No. % Ferry % % % Airplane from Land's End % 744 1% % Airplane from Newquay % % % Airplane from Exeter % 42 0% 292 1% Airplane from Bristol % 0 0% 412 2% Airplane from Southampton % 0 0% 116 1% Helicopter to Tresco % % % Helicopter to St. Mary's % % % Total % % % Source: Isles of Scilly Route Partnership, 2004 However, in order to understand what was driving the total demand for travel to the Isles of Scilly is was necessary to gain an understanding of the existing choices of all those currently travelling, regardless of journey origin. As a result it was decided to include those travelling to/from the airports at Land s End, Newquay, Exeter, Bristol and Southampton. In order to undertake these interviews in a cost effective manner, it was decided to undertake the interviews at the airports at Land s End (which Table 3 shows is the origin/destination for the largest number of aircraft trips) and at St Mary s airport on the Isles of Scilly (which is the island destination for all flights). For many visitors to the islands, a trip to the islands is just a small part of a holiday in the south west of England. This is reflected in the large number of day visitors. In these cases, the origin for the trip to the islands is better considered as the address at which they are staying for the rest of their holiday rather than their home address. As a result the surveys were designed to collect both the origin and destination terminals of the journey and information on the traveller s home location and final destination. In addition visitors were asked about both the time spent of the islands and the total holiday duration. This information provided some of the inputs required for inclusion in the stated preference exercises; however, it was decided not to vary the access and egress times. Varying these times would have added extra complexity to the exercises but if anything would have reduced their realism it is hard to

10 imagine why these times would vary, and this would certainly be something outside of the immediate influence of any changes to the ferry or its surrounding infrastructure. 2.4 Dealing with the complexity of existing fares As seen earlier, the available modes for the journey to and from the islands are restricted. However, there is a wide range of fares available for these modes, with a wide variety of single and return tickets offered on both the ferry, helicopter and fixed-wing services, each with differing restrictions. As a result, there is a potentially large number of mode-fare alternatives for the traveller to choose between. This is further complicated by the existence of holiday packages, where the traveller may not actually be aware of the fare paid for the trip over to the islands as it may be hidden within the overall cost of the package they have purchased which could include other travel costs and/or accommodation costs. Information on the travel costs paid were collected through a series of questions which asked both about the type of ticket, how many were travelling, whether the ticket was obtained through a special offer, and the cost if known. This information was then used in combination with fare tables provided by the operators to estimate the likely cost of the trip being examined. This was used as the base fare in the within-mode stated preference choice experiment. The fares for each ferry alternative in this experiment were calculated on the basis of percentage adjustments to the base fare, see Table 1 for further details. For the within-mode experiment, and the revealed preference modelling, it was necessary to estimate a base fare for the modes not used. This was done by assuming that the traveller would use the ticket type that was closest to that purchased for the used mode, and that the group size would remain constant. The fare tables were then used to assign the most comparable fare. As in the within-mode experiment, the fares for each alternative in the between-mode experiment reflected the fare for the travelling group and were calculated by percentage adjustments made to the calculated base fare for each mode. In the choices, these fares were multiplied by percentage increases and decreases, in order to understand the importance of cost in choice of mode, as detailed in Table 2. Therefore, in some choices the helicopter option would be more expensive than the airplane and in others the airplane option was more expensive than the helicopter. However, this complexity in fares (and access/egress times) required a high degree of customisation to be introduced to the choice experiments to realistically reflect the specific circumstances of the traveller being interviewed.

11 2.5 The utilisation of customised choice exercises The survey was designed to be undertaken as a two stage process. First travellers were intercepted during their journey and asked to answer questions that provided insight into the revealed preferences influencing the choices that they had made. Secondly a subset of respondents from the revealed preference survey were re-contacted and asked to take part in a more extensive survey that examined how they may behave under changed circumstances. As the visitors to the islands were located across the UK, the second stage of interviewing was undertaken by telephone. However, in order to fully consider the different options that may be available under changed circumstances it was necessary to send show material to the respondents in the post prior to their interview. The choice situations that respondents were being asked to consider were sufficiently complex to require a visual prompt as it was unlikely many individuals would be able to take in all the information and make informed trade-offs if the situations were simply explained over the phone. Typical practice in such circumstances is to produce a general set of choices to send to all respondents in a one size fits all approach. However, as we have seen, there was significant variation in both the access/egress circumstances and the fares and group size of each traveller. As a result, a common set of choices would have highly compromised the quality of the data collected as the options presented to respondents would have been unrealistic. To deal with this issue a computerised card generation system was developed to read in the data from the revealed preference surveys (which provided the details of each individual s existing trip circumstances). This data was then used to determine the equivalent base fares for the non-chosen modes. The incorporation of all of the background information was then used to generate customised, realistic, choice cards for each respondent The individually customised choice cards were then posted to each respondent so that a credible experiment could be conducted. An additional advantage of the approach adopted was the ability to introduce significant variation in the choice experiment designs across the sample. Paper surveys frequently use a limited number of fixed block designs to deal with the issue of splitting a fractional factorial experimental design over a sample. In this case it was possible to generate far more blocked divisions of the fractional factorial, as well as far more variation in the fares across the sample.

12 3. THE FIELDWORK CHALLENGES As discussed in the previous section, this study involved collecting information from two very diverse groups (Isles residents and visitors). However, both of these groups could be described as hard to reach. In addition, this research required significant samples for both the revealed preference survey and sufficient numbers (with an appropriate socio-economic mix) agreeing to participate in the follow-up stated preference survey. As a result significant fieldwork challenges were thrown up. 3.1 Interviewing hard to reach groups residents Interviewing in remote geographic areas tends not to be considered when the market research industry talks about hard to reach audiences. Traditionally, hard to reach is a term that is applied to socially remote groups, e.g. young people, ethnic minorities and other groups who are, or who have the potential to be, profitable consumer markets, or which are large enough groups with particular issues which, once addressed, could show an obvious return on investment, e.g. political or social returns. Consequently, smaller geographically remote groups (as opposed to socially remote) are often overlooked and are rarely targeted for consumer research, the logic being that these groups are both too difficult to reach and not big enough to return any meaningful commercial, social or political results. In short, they are all too often considered to be not worth the effort. On the rare occasions when geographically remote communities become the subjects of market research studies, it is usually as the result of localised, specific issues which will have a potentially big impact on respondents lives (housing, transport etc). As a result, the way in which the target community is approached becomes of vital importance if market researchers are to avoid political opinions and heightened community concerns clouding the research at the initial point of contact. Researchers must also be careful to avoid raising expectations or appearing to make promises about what impact the research may have, even if the research may indeed impact directly on the quality of life of respondents. This was an issue that the research team was sensitive to from the start of the project, and the interviewers were given detailed briefings prior to the fieldwork to ensure that any adverse impacts from (and on) the research would be minimised. A postal system was used to canvas the residents of the islands for the revealed preference element. This in itself should be a straightforward process. However, response rates to postal questionnaires are often low (often in the region of 10%) which posed issues for ensuring adequate responses for the survey segmentation and recruitment for the follow-up stated preference interview bearing in mind the low number of households on the Isles. In addition, there was concern that residents were already suffering from interview fatigue resulting a number of local and central Government consultations arising from ODPM Best Value surveys. It was believed that this would have a further knock-on effect on the response rates for this

13 project. Following discussions with the client it was arranged for letters signed by the Leader of the Isles of Scilly Council to be mailed to all households with the questionnaires setting out the survey objectives and listing why residents views were vital to the process. This, together with an incentive to win travel vouchers to the mainland, ensured sufficient responses were achieved to enable the robust analysis of the revealed preference data. 3.2 Interviewing hard to reach groups - visitors While the demographic make-up of the Scilly Isles is known and the population relatively easy to target, the same cannot be said of the islands visitors for whom no accessible records are kept. Traditionally, this second group would have been researched through a telephone methodology utilising purchased sample of a pre-defined group. However, due to the lack of information and sample regarding visitors to the Isles of Scilly, it was decide to utilise a face-to-face methodology to collect revealed preference data from visitors at the following locations: Land s End and St Mary s airports Penzance heliport The harbours at both Huw Town and Penzance On-street in Huw Town On ferry between Penzance and St Mary s. Fieldworkers also needed to ensure that respondents were comfortable with providing details of their home address. Many felt uncomfortable with providing such details as they were often in the middle of a week or two week holiday and had issues with security (i.e. giving out their home details to someone who knew that they were away from this location for at least a few more days). It was vital that this unease was overcome as the completion of the stated preference phase was dependent on recruiting at the revealed preference stage. As well as being used for modelling purposes the revealed preference interviews were also used to develop personalised show material to be mailed to respondents for the follow up stated preference exercises. The inputs to these for visitors was more complicated than for island residents. It was for this reason that self completion questionnaires were ruled out for visitors. Fieldworkers needed to ensure that all information was accurately recorded i.e. fare information, details on access and egress cost and duration etc. as this information affected the quality of both the revealed and stated preference data. 3.3 Timing of the fieldwork Originally, the tight project schedule set out that all of the fieldwork should be completed by mid-september. In practice this worked well for the visitor element of the project as the best time to complete the revealed preference study on a face to face basis was during the busy tourist months of July, August and September. However, for the resident element this caused

14 problems. Tourism related industry is the main employer on the Isles and as such there was concern that residents would be less inclined to respond to mailed questionnaires (and offer further time for the telephone stated preference interview) during their business period. With this in mind, and reflecting earlier concerns about response rates for this element, it was decided to delay this element until early Autumn. This did, however, throw up some challenges for the analysis and reporting timescales. 3.4 The impact of weather Weather patterns in West Cornwall and on the Isles are notoriously unpredictable and localised. Fieldworkers needed to be adaptable and able to move between locations. For example, fixed wing flights may well have been grounded due to fog at Land s End, however helicopter flights could still be running from Penzance. As a result, a system was put in place where fieldworkers would move between the locations if and when required. In addition, the sea state can mean uncomfortable crossings for both interviewers and respondents. Hit rates needed to be calculated to take these factors into account to ensure that the fieldwork was completed within the project schedule. 4. LESSONS LEARNT Clearly designing and undertaking a series of surveys to support the development of a mode-choice model for an island destination, where the majority of trips are made by non-residents visiting on tourism, brings with it challenges beyond those in a conventional mode-choice study. The biggest challenge on the design front was the specification of surveys and choice experiments that were realistic for the very diverse set of circumstances of the travellers being interviewed. The staged interviewing approach used for locating those travelling to the islands, generating sensible choice exercises for examining the impact of alternative futures, and conducting subsequent telephone interviews, provided a practical solution for dealing with many of the challenges faced. Separate surveys were developed for visitors and residents, as the purposes behind the trips being made, the flexibility around travel circumstances, and the values of time and price elasticity of the two groups were anticipated to be significantly different. In practice, this was found to be particularly important as the sensitivities of those resident on the islands are significantly different to those of visitors to the islands. The staging of the interviews provided a practical solution to the key design challenges, but brought with it additional challenges for the fieldwork. The two stage approach required that sufficient numbers of respondents to the first revealed preference survey agreed to participate in the second stated preference survey. In addition, it was necessary to ensure that the sample

15 retained for the second survey sufficiently represented all of the key socioeconomic groups. The major concerns for sampling residents was that of ensuring that sufficient numbers were recruited, from what is a relatively small population, and that the responses to the surveys were not clouded by reactive political responses, but rather objectively described the resident s existing travel behaviour and their expectation of how that may change. The sponsorship of the local council significantly helped with both of these issues, and highlights the importance in these situations of securing the support of local groups prior to any interviewing. Visitors were recruited locally whilst visiting the islands, but this was complicated by the variability of the weather, requiring significant flexibility to be built into the interviewing schedules and constant monitoring of the fulfilment of the various pre-defined quotas. The sensitivity to being recontacted in a subsequent telephone was initially underestimated, but with careful interviewer briefing and the use of appropriate reassurances, this issue was largely overcome. Whilst designing and undertaking the surveys to collect data for residents and visitors to a small group of islands brought a number of significant challenges, these were largely overcome and produced datasets that were able to support the development of a mode choice model. This in turn will allow a complete cost-benefit analysis to be undertaken for a range of transport options to and from the islands, with a level of detail that is usually only seen on the bids for funding for much larger schemes. This investment in data collection and modelling, however, reflects the perceived importance of the transport links to the sustainability of the islands. Acknowledgements This research study was funded by Cornwall County Council and the Department for Transport. The views expressed in this paper are those of the authors and do not necessarily reflect those of project sponsors. We would also like to acknowledge the contribution of James Laird and Andrew Daly who provided valuable input throughout this research study. Bibliography Isles of Scilly Route Partnership (2004) Isles of Scilly Major Link Scheme Bid. Report no: RT/DV01104/25A/023 Kouwenhoven, M., Rohr, C., Miller, S., Daly, A. and Laird, J. (2006) Evaluating a Replacement Ferry for the Isles of Scilly using a Discrete Choice Model Framework. European Transport Conference, Strasbourg.