0826-C2
Jason R. Jabbou, David Balsillie and Shashi Kant 1
Due to the elusive and non-market nature of many social and environmental values, determining the cost-benefit implications of different management options to society is particularly challenging. This paper investigates the application of an emerging valuation technique known as Choice Modelling (CM). We argue that this multicriteria decision-support approach offers promise as a potential instrument for analysing and integrating social values into forest policy. We discuss limitations with conventional valuation methods, illustrate the conceptual development and theoretical framework of CM, and examine the applicability of CM in the context integrating environmental, economic and social values related to the decision-making phase of sustainable forest policy.
In the last two decades, the practice of Canadian forestry has witnessed some of the most profound and rapid changes since its inception nearly a century ago. It is not surprising, therefore, that the process of public policy for today's forest institutions is fraught with difficulty. Contributing to the challenges of forest policy in Canada are a variety of issues relating to: aboriginal land claims, provincial ownership of forested land, tenure systems, non-market values, differing goals, and geographic and temporal separations of benefits from costs (Stanbury and Vertinsky 1998). Certainly, one of the most fundamental and difficult tasks involved in any environmental or resource policy-making is the effective integration and synthesis of various conflicting, disproportionate and incompatible values.
In the context of forest management, where competing demands for resources differ widely, policy-makers face a tremendous challenge in balancing the immense spectrum of societal preferences. Towards this end, governments and forest institutions have placed increasing emphasis on soliciting public input into the management and planning process. While significant advances in social science have improved the structural quality and implementation of public participation (Duinker 1998, Jabbour and Balsillie 2003), few studies have specifically addressed techniques for measuring and analysing public inputs. We argue that the absence of effective policy instruments, that evaluate and integrate societal values, is central to the public cynicism that surrounds forest-policy and ultimately is a major impediment in achieving sustainable forest management (SFM).
Today, dissatisfaction with forest policy has extended well beyond the realm of public scrutiny. In fact, a broad consensus among forestry scholars has emerged that, in many critical respects, forest policies have been largely unsuccessful. Tollefson (1998) and others suggest that the dimensions of this shortfall centre on the perceived deficiency of existing policy to protect multiple and non-timber values. Moreover, there is a strong consensus that contemporary policies in Canada have essentially failed to promote and protect various non-consumptive, less tangible, and social forest values.
Generally, forest management decisions are concerned with changes in the levels of attributes that constitute a particular good or service. As such, managers and policy-makers are interested in determining the economic value of alternative management strategies, which ultimately impose different implications on the levels of various forest related goods and services. However, due to the elusive and non-market nature of many forest values and objectives, determining the associated cost-benefit implications of different management options to society is a particularly troublesome exercise. Traditionally, governments and forest institutions have addressed this problem through non-market Contingent Valuation (CV) techniques, which attempt to estimate the changes in the non-consumptive values associated with forest resources. However, due to inherent weaknesses relating to in theoretical and conceptual design, CV methods have recently been the subject of considerable criticism. A particular source of contention is the fact that CV seeks to derive dollar estimates of economic value based on respondents stating their willingness to pay (WTP) in obtaining or avoiding a situation based on a hypothetical market. These subsequent value estimations are not only contingent on a hypothetical market but are also merely a reflection of the scenario as a whole, rather than the individual attributes that might comprise a particular scenario. Therefore, the argument follows, that the application of CV in analyzing implications for alternative management options is likely to result in socio-economic distortions.
In recent years, an emerging technique known as Choice Modelling (CM) has gained support as a successful alternative method for non-market valuation. Rather than directly asking respondents to state their values in dollars, respondents state preferences between one group of services or characteristics, at a given cost to the individual, and another group of characteristics at a different cost. The essential point of departure therefore, is that CM infers values from the choices or tradeoffs that people make. Arising from conceptual limitations encountered in CV studies, the CM method also has the distinct ability to model separately yet simultaneously multiple resource use-allocation scenarios and the changes in attribute levels that comprise a resource. More specifically, CM allows for the simultaneous analysis of several influences on respondents' choice (Rolfe et al. 2000). Because CM focuses on multiple-attribute and attribute level tradeoffs, we argue that the technique is well suited to policy-making situations where a set of possible actions results in different social, economic and environmental impacts, as is predominantly the case in forest management. The following sections of this paper provide an brief overview of the conceptual development and theoretical framework for CM, the methods for its application and a discussion on the appropriateness of CM as an instrument to assist policy-makers with the integration of environmental, economic, and social values related to the decision-making phase of SFM.
While the initial development of qualitative choice analysis arose from the discipline of psychology almost 30 years ago, the use of CM has only recently extended to multi-attribute choice problems in the field of economic valuation. In the last decade, an emerging body of research by Tacconi (1995), Rolfe et al. (2000), Bennett and Blamey (2000), Adamowicz and Boxall (2001) and others, have advanced the concept and application of CM as an innovative tool for estimating environmental values. Results from these studies demonstrate the ability of CM to model the level societal support for various environmental and resource management options. As previously mentioned, CM methods are capable of determining the approximate value of individual attributes that constitute a specific environmental or social good, such as employment, forest area, recreation opportunities, etc. This is particularly important in the arena forest management where decisions are often concerned with changing attribute levels, rather than loosing or gaining values as a whole. By deriving value estimates of the changes in the aggregate levels of environmental and social quality, CM offers promise for understanding implications for multiple forest management alternatives.
The theoretical basis of CM analysis is random utility theory (RUT), which describes the choice behaviour of individuals in a utility maximizing framework. According to RUT, the utility of a good is composed of (1) an observable or deterministic component, which is a function of a vector of attributes, and (2) an unobservable or random error component (Boxall and Macnab 2000). The following equation for an individual's utility, formalises the basic relationship where (Vi) is the observable component and (εi) represents the error component of utility.
[1] Ui = Vi + εi
The equation below disaggregates the systematic component of choice further, where respondent (i) derives utility (Uij) from the alternatives (j) in choice set (C); utility is held to be a function of the attributes of the relevant good (Zij) and the characteristics of the individual (Si), together with the error term (Rolfe et al. 2000).
[2] Uij = V(Zij, Si) + εij
Due to the inherent stochastic or random error component of (Ui), a researcher can never hope to fully understand and predict preferences; hence, choices made between alternatives are expressed as a function of the probability that respondent (i) will choose (j) in preference to other alternatives:
[3] Pij = Prob(Vij + εij > Vih + εih)
for all h in choice set C, j ≠ h
Assuming that the error terms of the resultant utility function are independently and identically (Gumbell) distributed, a multinomial logit model (MNL) results (McFadden 1974). A consequence of this error assumption is the property of independence of irrelevant alternatives (IIA), which states that the probability of choosing one alternative over the other depends entirely on the utility of the respective alternatives. Violations of the IIA which may occur for reasons of close substitutions in the choice sets or the existence of heterogeneous preferences can be detected using mother logit estimation tests (see, McFadden et al. 1977). The MNL model generally takes the following form:
where (λ) represents a scale parameter, inversely proportional to the variance of the error term and normalized to 1 for any particular data set.
Choice Modelling (CM) involves choice experiments (or choice surveys) which are essentially samples of choice sets selected from the universal set of all possible choice sets that satisfy certain statistical properties (Louviere et al. 2000). With each choice survey, respondents answer a series of choice questions, each presenting different combinations and levels of the relevant services, and the cost to the respondent of the action. The application of CM experiments generally involves six steps outlined below (Adamowicz et al. 1998). The initial task is to define the decision problem. For the purposes of developing forest policy, this would include determining exactly what management options are being valued, characterizing the values used to measure the resulting changes, and establishing whom the relevant population is. In the context of analysis, the associated values of interest represent the additional costs and benefits resulting from the implementation of alternative policy options relative to some pre-defined status quo alternative (Bennett and Blamey 2001).
The second step, which is the most important and difficult part of the CM process, is the survey development and attribute selection. Once the estimated values of interest are established, the specific attributes used to describe the outcomes presented in the choice sets and the range (or levels) over which these attributes will vary is determined. An important consideration when developing the attributes and levels is that the respondents involved need to be able to relate or find relevance in answering the survey. Thus, the survey design process usually involves a series of interviews and/or focus groups with the types of people who will be receiving the final survey. Here, researchers ask general questions about peoples' understanding of the issues related to the decision problem, and whether and how they value various characteristics or services associated with the specific resources site in question.
The third step is the experimental design, in which the alternatives' levels are set and structured into choice sets. Because the extensive array of possible interactions of attributes and levels often exceeds a single respondent's ability to cope with all the resulting choice combinations, fractional factorial designs are used to help reduce the number choice sets that each respondent is required to evaluate.
Figure 1. Hypothetical choice set
The fourth step in the CM process involves sample size and survey implementation. The appropriate sample frame for each study will depend on the nature of the particular application and the statistical power that is required for the estimation process (Bennett and Blamey 2001). Researchers will then divide the sample frame into sub-samples to reflect the resulting blocks or partitions revealed in the fractional factorial design. Ideally, researchers should select participants randomly from the relevant population using standard statistical sampling methods.
Following the completion of data collection, step five requires compilation, coding and analyses of the data. Using advanced econometric software packages (i.e. LIMDEP), researchers are able to generate and extrapolate choice models. As discussed in the previous section, the model estimation approach is the multinomial logit procedure, which models the probability of a respondent choosing an alternative as a function of their attributes and their socio-economic characteristics. Therefore, the probability that a respondent wil choose an alternative "increases as the levels of desirable attributes in that alternative rise and the levels of undesirable attributes falls - relative to the levels of the attributes in the other alternatives that are available" (Bennett and Blamey 2001).
The final step in the CM process involves analysis and policy decision support systems (DSS) development (Adamowicz and Boxall 2001). By analysing the tradeoffs generated by the model, researchers can estimate the average value for each of the services associated with the management area. By extrapolating this information to the relevant population, researchers can then calculate the total costs and benefits under different scenarios and ultimately, simulate outcomes for policy analysis.
Policy decisions regarding the management and allocation of natural resources are much more likely to be informed by improved estimates of economic, environmental and social values. However, the analysis and quantification of certain values are much less conducive to economic valuation than others. Certainly, it is difficult to estimate the real demand and the associated implications of social values. By simply looking at an individual's purchasing behaviour, for instance, one cannot possibly reveal an individual's preferences for social quality. As such, the inclusion and integration of many important societal preferences into decision-making routinely go unnoticed. Consequently, policies often fail to adequately represent the values and objectives of all constituencies, as those whose interests are predominantly economic and monetarily tangible generally prevail. We argue that current forest policy structures do not pay enough attention to the social nature and contingency of choices related to decision-making.
If all management scenarios involve choices and all choices involve tradeoffs; decision-making is about assessing the inherent tradeoffs involved in process of making choices. In the context of forestry, where cost-benefit implications guide policy-making, information on the relative value of choice alternatives is critical in the assessment of tradeoffs. Furthermore, understanding and predicting the behavioural nature of individual and aggregate choice responses is fundamental in determining and evaluating the societal costs and benefits that different management decisions implicate (Louviere et al. 2000).
Standard economic approaches to behaviour assume that individuals base choices on the theory of maximizing returns (or utility maximization), and that individuals have stable and well-defined preferences. In contrast, individuals deviate from these predictions quite regularly and in a consistent manner. These differences exist, in large part, simply because conventional decision and valuation models fail to transmit a host of intangible considerations that affect the choices of real economic factors. We believe that the development of choice modelling (CM) represents a significant advance in the analysis of individual choice behaviour. While CM clearly contains some elements of the traditional economic theory on choice behaviour, the theory also postulates that individuals derive utility from the properties or characteristics that goods or in this case, choice options are comprised of, rather than the choices per se (Bennett and Blamey 2001). This distinct behavioural linkage in CM allows researchers to anticipate changes in choice behaviour with changes in attributes (Adamowicz and Boxall 2001).
With respect to valuation, a critical feature of CM is its ability to determine the relative importance of economic, environmental and social factors associated with non-consumptive and non-market values. As demonstrated in this paper, the application of CM can estimate the implications of environmental quality attributes, employment opportunities and financial factors such as taxes or levies in influencing choices among alternatives. In resource and forest management, where tradeoffs between conservation and development as well as natural and non-natural goods and services are in a constant state of flux, CM techniques could provide a particularly useful means towards adaptive management.
An important advantage of CM over traditional economic non-market valuation techniques is its focus on attributes, which Mallawaarachchi et al. (2001) argue, makes the approach more suitable for estimating both the values of attributes and situational changes. With CM, respondents are not required to make direct choices between social or environmental quality and money. Rather, the method allows respondents to think in terms of tradeoffs, which is inherently easier than directly expressing dollar values but also, encourages respondent introspection. Individuals are generally more comfortable providing qualitative rankings or ratings on a grouping of attributes that include price, rather than dollar valuation of the same grouping without prices, by de-emphasising price as simply another attribute. The CM approach not only provides a richer description of attribute tradeoffs that individuals are willing to make, but provides smaller variances in the resulting welfare estimates. Finally, the CM method has the potential to significantly reduce biases and problems such as respondent protest bids, which often arise in open-ended valuation studies because of the unfamiliar and unrealistic task of putting prices on non-consumptive and non-market values. Therefore, respondents in CM studies are able to give more meaningful and accurate answers to questions about their behaviour (i.e. a preference of one alternative over another).
This paper has demonstrated the application and suitability of CM as a potential instrument in developing sustainable forest policy. While the CM approach is still in its infancy and much development needs to occur (Rolfe et al. 2000), its unique potential to model complex and simultaneous tradeoffs in multi-variable situations offers promising application for socio-economic frameworks. In countries such as Canada, where the public have demanded an increasing role in decision-making, the development of economic instruments capable of analysing and integrating social values are in short supply, particularly in forest management. Forest planners and policy-makers need an effective means to solicit and analyze societal preferences in order to make effective multiple-objective decisions. The CM framework has the ability to value a range of management scenarios. More specifically, the application of CM provides an innovative and formal structure within which to investigate the environmental, economic and social tradeoffs that are involved in alternative forest strategies. The per-individual estimates that are derived from CM studies can be easily aggregated to determine the total value accrued to the broader population affected by the resource issue in question. Thus, forest policy-makers can identify an optimal forest management scenario that is both socially desirable and accommodates to the key principles of sustainable forest management.
Adamowicz, W., P. Boxall, M. Williams and J. Louviere. 1998. Stated preference approaches for measuring passive use values: choice experiments and contingent valuation. American Journal of Agricultural Economics. 80: 64-75.
Adamowicz, W. and P.C. Boxall. 2001. Future directions of stated choice methods for environmental valuation. Paper prepared for: Choice experiments: A new approach to environmental valuation. London, England. April 10.
Bennet, J. and R. Blamey. 2001. The choice modelling approach to environmental valuation. Edward Elgar Publishing. Northampton, MA. pp. 269.
Boxall, P.C. and B. Macnab. 2000. Exploring the preferences of wildlife recreationists for features of boreal forest management: A choice experiment approach. Canadian Journal f Forestry Research. 30: 1931-1941.
Duinker, P.N. 1998. Public participation's promising progress: Advances in forest decision-making in Canada. Commonwealth Forestry Review. 77(2): 107-112.
Jabbour, J.R. and D. Balsillie. 2003. The effectiveness of public participation in forest management: A case study analysis of the Morice Innovative Forest Practices Agreement. The Forest Chronicle. In press.
Louviere, J.J., D.A. Hensher and J.D. Swait. 2000. Stated choice methods: Analysis and application. Cambridge University Press, Cambridge UK.
Mallawaarachchi, T., R.K. Blamey, M.D. Morrison, A.K.L. Johnson and J.W. Bennett. 2001. Community values for environmental protection in a cane farming catchment in Northern Australia: A choice modelling study. Journal of Environmental Management. 62: 301-316.
McFadden, D. 1974. Conditional logit analysis of qualitative choice behaviour. In P. Zarembka (ed.) Frontiers in econometrics. New York, Academic Press.
McFadden, D., K. Train and W.B. Tye. 1977. An application of diagnostic tests for the independence from irrelevant alternatives property of the multinomial logit model. Transportation Research Record. 637: 39-45.
Rolfe, J., J. Bennett and J. Louviere. 2000. Choice modelling and its potential application to tropical rainforest preservation. Ecological Economics. 35: 289-302.
Stanbury, W.T. and I.B. Vertinsky. 1998. Governing instruments for forest policy in British Columbia: A positive and normative analysis. In The wealth of forests: markets, regulations and sustainable forest management, ed. C. Tollefson, 42-77. UBC Press.
Tacconi, L. 1995. Rethinking the economic analysis of forests: theory and practice. Forest Ecological Management. 73: 229-238.
Tollefson, C. 1998. The wealth of forests: markets, regulations and sustainable forest management. UBC Press, Canada.
1 PhD candidate, Faculty of Forestry, University of Toronto, 33 Willcocks St. Earth Sciences Centre, Toronto, Ontario M5S 3B3, Canada. [email protected]