Elsevier

Landscape and Urban Planning

Volume 169, January 2018, Pages 160-177
Landscape and Urban Planning

Research Paper
Providing information to respondents in complex choice studies: A survey on recreational trail preferences in an urban nature park

https://doi.org/10.1016/j.landurbplan.2017.09.003Get rights and content

Highlights

  • Different ways of presenting choice information on recreation setting are discussed.

  • Omitted variable bias and cognitive burden could be balanced by attribute aggregation.

  • With partial profile designs respondents can handle greater number of attributes.

  • Using design with less specific attributes could be beneficial in certain contexts.

  • Non-visual sensory trail experience was more important than visual experience.

Abstract

This paper examined the recreational trail preferences of visitors in the Medvednica Nature Park, a protected forest area on the outskirts of the City of Zagreb, the capital of Croatia. A discrete choice experiment (DCE) was conducted to get the insight into relative importance of different resource, social and managerial conditions in the park. Accounting for multiple site conditions requires a relatively large number of choice attributes, which may impose too high cognitive burden on respondents. On the other hand, ignoring relevant attributes may lead to the omitted variable bias. A split sample approach was used to find the balance between the possibility of omitted variable bias and cognitive burden; one version of the questionnaire used DCE with the lower number of attributes, of which some were multidimensional, while the other version used DCE with the greater number of more specific attributes. By using partial profile design in the latter experiment, the number of attributes in the choice task was identical in both experiments. Perceived difficulty of the choice task, self-reported choice certainty and choice consistency were similar across the two experiments. Heterogeneity in preferences and scale was detected in both experiments. Indications of non-compensatory behavior, and greater error variance among less experienced trail users were found in the partial profile experiment with more specific trail attributes, but not in the experiment with multidimensional attributes. Based on the research results, important managerial implications were derived. Non-visual sensory experiences of nature, namely fresh air and soundscape, were generally more important to trail users than visual experiences. Crowding was detected as an important characteristic of trail experience; however, trail users were willing to tolerate relatively high levels of crowding.

Introduction

Increased crowding, noise, air and visual pollution are some of the adverse social and environmental impacts often detected in urban environments (Kil, Stein, & Holland, 2014; Parumog, Mizokami, & Kakimoto, 2006). Green areas such as urban forests often serve as spaces to counter the stress of urban life and provide relief from high population densities (Bakhtiari, Jacobsen, & Jensen, 2014; Karjalainen, Sarjala, & Raitio, 2010). Such areas are popular places for outdoor recreation activities; however, recreational benefits which they provide might be threatened by excessive levels of visitor activity (Arnberger & Mann, 2008; Mieno, Shoji, Aikoh, Arnberger, & Eder, 2016). Outdoor recreation activities contribute to biophysical and aesthetic changes in environment, often reflected in degradation of land cover, soil erosion, disturbance of wildlife and littering (Kil et al., 2014, Manning, 2011). High use levels and undesirable user behavior can also lead to crowding and user conflicts (Arnberger and Mann, 2008, Manning, 2010). Understanding visitors' preferences for resource conditions (e.g. trail condition, surrounding landscape), social conditions (e.g. type and level of use, visitor behavior) and managerial conditions (e.g. vegetation and litter management) of natural environment is valuable in designing effective landscape management strategies (Reichhart & Arnberger, 2010).

Discrete choice experiments (DCEs) have become increasingly popular in outdoor recreation studies to investigate the preferences of visitors and tradeoffs in recreational conditions they are willing to make (e.g. Arnberger & Eder, 2011; Bullock & Lawson, 2008; Hanley, Wright, & Koop, 2002; Kainzinger, Arnberger, & Burns, 2016; Lawson & Manning, 2002; Manning, 2011; Newman, Manning, Dennis, & McKonly, 2005; Reichhart & Arnberger, 2010). Their popularity owes to the possibility of simultaneously evaluating different site conditions and underlying trade-offs, which was not possible with normative and univariate approaches used in the earlier studies. However, simultaneous evaluation of different conditions (called attributes in DCE methodology) imposes greater cognitive burden on respondents, and there is a limit on the number of attributes respondents can process at the same time (Zhang, Johnson, Mohamed, & Hauber, 2015). When asked to process too much information when making choices among competing alternatives in a DCE task, respondents often alter their decision rules (e.g. ignore some of the information presented to them) to simplify the choice task (Colombo & Glenk, 2014; DeShazo and Fermo, 2002, Erdem and Thompson, 2014). This violates DCE assumption of compensatory behavior or unlimited substitutability between the attributes, and may contribute to an increased error variance and affect the validity of utility estimates (Dellaert, Donkers, & van Soest, 2012). Therefore, information gain from providing a more complete description to respondents may be outweighed by higher cognitive burden from the larger number of choice attributes that describe a setting, which usually forces researchers to omit the attributes that they think are not essential for most of the population (DeShazo & Fermo, 2002; Louviere, Pihlens, & Carson, 2010). This could however lead to the omitted variable bias as respondents may be influenced by attributes that are not included in a DCE, causing bias in the utility estimates, particularly if omitted and included attributes are correlated (Witt, Scott, & Osborne, 2009). Giving respondents more complete information rather than offering only a subset of relevant attributes or their aggregated definition could improve the consistency and confidence respondents have in choices due to a more meaningful interpretation of alternatives (Hensher, 2006a).

The survey design should find the optimal balance between the interests in various attributes (i.e. omitted variable problem) on the one hand and complexity of the choice task on the other (DeShazo and Fermo, 2002, Witt et al., 2009). This paper discusses survey design features that may help in finding this balance – type of information presented to respondents (multidimensional vs. unidimensional attributes), and type of experimental design used to deliver information to respondents (full profile vs. partial profile design). A split sample approach was used with two different choice experiments that evaluated recreational trail preferences of forest visitors in the Medvednica Nature Park on the outskirts of the City of Zagreb, the capital of Croatia, but differed in the provision of information about the recreational setting. One experiment used full profile design and included multidimensional attributes to keep the number of choice attributes manageable for respondents, while considering all relevant aspects of visitors’ experience. The other experiment included a greater number of more specific, unidimensional, attributes to describe the recreational setting. A partial profile design was used in this experiment to prevent a large increase in choice task complexity. In a partial profile design only a subset of attributes appear in each choice set (Kessels, Jones, & Goos, 2012), thus the number of attributes per choice task was kept constant across experiments. A core set of attributes used in other outdoor recreation studies was expanded by including the influence of non-visual sensory experiences, namely air and noise pollution, which were considered a concerning aspects of visitor activity by the park management. Previous studies have been mostly focused on visual experiences of visitors; however, sensory experiences, such as noise, could have a strong influence on the perception of natural landscapes and visitors' experience (Buchel and Frantzeskaki, 2015, Hallo and Manning, 2009). We compared self-reported choice certainty and perceived choice difficulty across the two experiments and investigated how different presentation of information about recreational setting affected decision behavior of respondents, utility estimates and response error variance. Understanding how respondents attended to information presented in the choice tasks and identifying limitations and advantages of alternative attribute presentation methods allows improving the design of the future choice studies (Colombo & Glenk, 2014). As environmental valuation generally involves trade-offs between complex goods and services that cannot be easily described with a restricted number of attributes, how to optimally describe a choice context to respondents is an important consideration.

The following hypotheses were examined: a) non-visual sensory experiences of nature, namely fresh air and soundscape, could be more important to peri-urban forest visitors than visual experiences, b) different information provision strategies influence propensity to non-compensatory choice behavior, and c) different information provision strategies affect the utility estimates and error variance (or choice consistency).

Section snippets

Methodology

Within DCEs, respondents are presented with multiple choice sets consisting of at least two alternatives, and asked to select their preferred alternative in each set. Alternatives differ in the values of attributes, such as resource, social and managerial trail conditions (see e.g. Arnberger and Eder, 2011, Bullock and Lawson, 2008, Hanley et al., 2002, Lawson and Manning, 2002, Newman et al., 2005, Reichhart and Arnberger, 2010). Experimental designs are used to generate the choice sets (

Medvednica nature park study

Medvednica Nature Park is a protected area located on the Medvednica mountain on the outskirts of the City of Zagreb, the capital of Croatia (Fig. 1). Since Croatia has become the member of European Union, it has become a part of the Natura 2000 network. The park, with an area of 17,938 ha, is rich in biodiversity, and is the habitat of many different protected and endangered species of flora and fauna. The main feature of the park are forests of great biological value. It has improved air

Descriptive statistics

The socio-demographic characteristics of respondents, their perceptions of several experiential conditions, perception of crowding and place attachment are summarized in Table 2. The average age of respondents was 38. Most of them were female and finished higher education. They were dominantly accompanied by their friends and/or family members, and quite familiar with recreational experience in the Medvednica Nature Park. Reported feelings of place attachment were relatively strong. Enhancing

Discussion

Considerable share of respondents had difficulty choosing between alternative recreational settings, regardless of the information provision strategy. This raises concerns over compensatory behavior in similar studies, as they often included a larger number of attributes, which may be too much for respondents to process at the same time. In the experiment A, we introduced multidimensional attributes into the choice task to keep the number of attributes relatively low and therefore prevent large

Managerial implications and conclusions

Understanding the relative importance of different trail conditions to visitors can help the park management to appropriately prioritize activities and maximize the visitor experience. In this study, as in many other European outdoor recreation areas, visitors reported relatively high levels of crowding (Arnberger & Mann, 2008). Other studies showed that visitors to high use areas can tolerate more users than visitors to low use areas, and high visitor numbers can be acceptable as solitude is

References (61)

  • M.F. Luce et al.

    Coping with unfavorable attribute values in choice

    Organizational Behavior and Human Decision Processes

    (2000)
  • T. Mieno et al.

    Heterogeneous preferences for social trail use in the urban forest: A latent class model

    Urban Forestry & Urban Greening

    (2016)
  • B. Neuts et al.

    Tourist crowding perception and acceptability in cities: An Applied Modelling Study on Bruges

    Annals of Tourism Research

    (2012)
  • L.B. Pedersen et al.

    Does the inclusion of a cost attribute in forced and unforced choices matter? Results from a web survey applying the discrete choice experiment

    Journal of Choice Modelling

    (2011)
  • T. Reichhart et al.

    Exploring the influence of speed, social: Managerial and physical factors on shared trail preferences using a 3D computer animated choice experiment

    Landscape and Urban Planning

    (2010)
  • B. Rulleau et al.

    Preferences: rational choices and economic valuation: Some empirical tests

    The Journal of Socio-Economics

    (2012)
  • J. Zhang et al.

    Too many attributes: A test of the validity of combining discrete-choice and best-worst scaling data

    The Journal of Choice Modelling

    (2015)
  • A. Arnberger et al.

    The influence of age on recreational trail preferences of urban green-space visitors: A discrete choice experiment with digitally calibrated images

    Journal of Environmental Planning and Management

    (2011)
  • A. Arnberger et al.

    Crowding in European forests: A review of recent research and implications for forest management and policy

    Forestry

    (2008)
  • M. Bech et al.

    Does the number of choice sets matter? Results from a web survey applying a discrete choice experiment

    Health Economics

    (2011)
  • R.K. Blamey et al.

    Attribute Causality in Environmental Choice Modelling

    Environmental and Resource Economics

    (2002)
  • P.C. Boxall et al.

    Understanding Heterogeneous Preferences in Random Utility Models: A Latent Class Approach

    Environmental and Resource Economics

    (2002)
  • S.D. Bullock et al.

    Managing the Commons on Cadillac Mountain: A Stated Choice Analysis of Acadia National Park Visitors' Preferences

    Leisure Sciences

    (2008)
  • K. Chrzan

    Using partial profile choice experiments to handle large numbers of attributes

    International Journal of Market Research

    (2010)
  • S. Colombo et al.

    Social preferences for agricultural policy instruments: Joint consideration of non-attendance to attributes and to alternatives in modelling discrete choice data

    Journal of Environmental Planning and Management

    (2014)
  • M. Czajkowski et al.

    The effects of experience on preferences: Theory and empirics for environmental public goods

    American Journal of Agricultural Economics

    (2015)
  • B.G.C. Dellaert et al.

    Complexity effects in choice experiment-based models

    Journal of Marketing Research

    (2012)
  • S. Erdem et al.

    Prioritising health service innovation investments using public preferences: A discrete choice experiment

    BMC Health Services Research

    (2014)
  • D. Gyrd-Hansen et al.

    The price proxy in discrete choice experiments: Issues of relevance for future research

  • N. Hanley et al.

    Modelling recreation demand using choice experiments: Climbing in scotland

    Environmental and Resource Economics

    (2002)
  • Cited by (30)

    • The performance of full versus partial profile choice set designs in environmental valuation

      2023, Ecological Economics
      Citation Excerpt :

      Davis et al. (2019) use a full profile design in one split but compare results to an ICE, which is a different approach than a partial profile design as understood here. Sever and Verbič (2018) are closest to a straight-ahead comparison of both design types. However, they confound the effects of multidimensional vs unidimensional attributes with the design type and thus cannot differentiate what triggered the differences.

    View all citing articles on Scopus
    View full text