Capturing social impacts for decision-making: a Multicriteria Decision Analysis perspective


Correspondence: Mark A. Burgman, Australian Centre of Excellence for Risk Analysis, School of Botany, University of Melbourne, Parkville, Vic. 3010, Australia.




We aim to explore the capacity of MCDA methods to successfully capture social impacts and integrate stakeholders’ participation into environmental decision applications. We follow a theoretical framework that deconstructs the concept of social impact into two components: human impacts and social change processes.




We systematically reviewed the literature on MCDA in the Web of Science (ISI) database, finding 119 papers that meet our search criteria. For each paper, we identified the social change processes or human impacts objectives, along with the attributes that measured them. We also recorded the degree of stakeholder participation in each phase of the MCDA stages.


We found that MCDA practitioners have increasingly integrated social concerns in the analysis of environmental problems, estimating the potential impacts, and developing participative procedures for stakeholders. We identified 252 objectives that represent human impacts or social change processes. Constructed attributes were the most commonly employed (56%), although natural (20%) and proxy (24%) attributes were also relevant. Estimating human impacts or social change processes can involve public participation, but is not a requirement for MCDA: 42% of papers (n = 50) include stakeholder engagement in one or more stages of the decision-making process. We found that stakeholders engage differently across case studies, demonstrating that this process is neither simple nor homogenous.

Main conclusions

Our review suggests that MCDA methods are appropriate techniques to integrate a wide range of social objectives and stakeholder engagement, supporting well informed and democratic decisions.


Environmental planning decisions generally involve complex couplings between human and natural systems (Liu et al., 2007), requiring sound methodologies dealing with potential social impacts and public participation (Lahdelma et al., 2000). In a deliberative setting, decision analysts seek to integrate stakeholder values with technical judgments to find acceptable solutions (Gregory & Keeney, 1994; Belton & Stewart, 2002); decisions made without social acceptance may be fragile, reducing the viability of proposed alternatives (Grimble & Wellard, 1997; Gutrich et al., 2005).

Capturing social impacts is a difficult task, partly because social criteria are not evidently positive or negative in themselves, but rather depend on changing perceptions (Burdge & Vanclay, 1995; Vanclay, 2002). Contingent valuation is one of the most common methods used to capture non-market values in a traditional cost–benefit analysis, whereby people indicate their willingness to pay an amount to obtain a non-market benefit (Carson et al., 2001). Although this is a valid and widely used technique, its capacity to capture complex social impacts is not clear (Gregory & Slovic, 1997). Experiments show that respondents sometimes find it difficult to value environmental and social benefits in monetary terms, leading to error or bias in outcomes (Gregory et al., 1993). In addition, elicitation processes may be sensitive to seemingly innocuous variations in methodologies and framing (e.g. Kahneman & Knetsch, 1992; Gregory et al., 1995).

Multicriteria decision analysis (MCDA) is a family of decision support techniques which analyse multiobjective problems (Vincke, 1992; Belton & Stewart, 2002). These methods have been applied to many environmental decisions (Huang et al., 2011) including conservation biology (Moffett & Sarkar, 2006), sustainable energy (Wang et al., 2009), forestry and natural resource management (Mendoza & Martins, 2006), management of contaminated sediments (Linkov et al., 2006), bioenergy systems (Buchholz et al., 2009) and agriculture (Hayashi, 2000). MCDA methods are sufficiently flexible to incorporate quantitative and qualitative information in decision models, estimating impacts when monetary values or single criteria are inappropriate or difficult to conceptualize reliably (Gregory et al., 2001; Munda, 2006). At the same time, MCDA provides formal structure to incorporate people's values and participation in the decision process (Banville et al., 1998; Hajkowicz, 2008).

While there is a clear emphasis on stakeholder engagement and social participation in the MCDA literature, the extent of these processes in real case studies is unknown. In addition, it is not clear whether MCDA applications successfully capture social impacts when they formalize the decision problem itself. The identification and construction of robust descriptors of social impacts is a critical aspect in contentious decisions – applying sophisticated aggregation procedures is useless if the criteria fail to capture the decision problem or are improperly built (Bouyssou, 1990). Also, more precise definition and distinction of the possible social impacts would be beneficial for MCDA practitioners, which would generally require an operational set of indicators or attributes for inclusion in a decision-making model. Effective capture of social impacts and stakeholder engagement is among the biggest challenges for decision analysts today (Failing et al., 2007).

Our aim is to explore the capacity of MCDA methods to successfully integrate social values and stakeholder participation into environmental decision problems. First, we deconstruct the concept of social impact into two components: human impacts and social change processes; we discuss a theoretical framework for providing operational definitions. Second, we describe the role of objectives and attributes for informative decision analysis. Third, we present the results of our review of the literature; specifically, we set out to explore the following: (1) the capture of human impacts and social change processes in MCDA applications and the relationships between these concepts, (2) the performance of different types of attributes for estimating human impacts and social change processes, and (3) the degree of stakeholder participation in each phase of the decision-making process.

Operationalizing social impacts

The USA government promulgated the National Environmental Policy Act in 1969, establishing obligations in the identification of potential social impacts as part of the environmental impact assessment process (Burdge & Vanclay, 1995; Taylor et al., 2004). In 1992, the Interorganizational Committee on Guidelines and Principles (ICGP) consolidated social impact assessment as a discipline, establishing standards and tools for social researchers (International Committee on Guidelines & Principles for Social Impact Assessment (ICGP), 1995). Social impact assessment provides guidelines for environmental planning, promoting public participation and identification of alternatives, baseline conditions and a wide range of social impacts (International Committee on Guidelines & Principles for Social Impact Assessment (ICGP), 2003; Burdge & Vanclay, 1995).

Social impacts are formally defined as ‘the consequences to human populations of any public or private actions that alter the ways in which people live, work, play, relate to one another, organize to meet their needs and generally cope as members of society. The term also includes cultural impacts involving changes to the norms, values and beliefs that guide and rationalize their cognition of themselves and their society’ (International Committee on Guidelines & Principles for Social Impact Assessment (ICGP), 2003). Some authors have developed more precise guidelines, identifying main categories of impacts (e.g. Gramling & Freudenburg, 1992; International Committee on Guidelines & Principles for Social Impact Assessment (ICGP) (1995, 2003); Schooten et al., 2003). However, results are not completely satisfactory, presenting substantial discrepancies, omissions and contradictions (Vanclay, 2002).

Slootweg et al. (2001) propose a conceptual framework that allows more precise distinctions in the social context of decision-making, differentiating between social change processes and human impacts. Social change processes are ‘discrete, observable and describable process which change the characteristics of a society’ (Slootweg et al., 2001:27). In contrast, human impacts are experienced by people whether in physical or psychological terms, and are intrinsically subjective. Human impacts and social change processes are generally connected and may be manifestations of the same phenomenon. Slootweg et al. (2001) suggest that, under certain conditions, social changes may lead to human impacts, establishing causal relationships. As we shall see below, these potential links are critical in decision-making models, and they need to be better clarified.

Objectives and attributes

More or less all MCDA approaches identify objectives and performance criteria (attributes), which evaluate a set of alternatives (Keeney, 1992). Objectives are statements that explicitly specify values, establishing a direction of preference expressed as ‘minimize’ (less is better) or ‘maximize’ (more is better) (Keeney, 1988). Identifying and structuring relevant objectives is crucial in any multiattribute elicitation process (Keeney, 1988); failure to do so can lead to a biased value model (Edwards, 1977). A common approach is to construct hierarchies or value trees, establishing links between fundamental and means objectives. Fundamental objectives refer to the elements of ultimate concern, whereas means objectives are subcomponents, or the ways to achieve (or avoid) them (Keeney & Raiffa, 1976; von Winterfeldt & Edwards, 1986).

Attribute selection is a subjective judgment of the analyst, one that typically involves consideration of an imperfect understanding of the causal relationship between alternatives and the objective itself. In decision science, authors generally recognize three types of attributes: natural, proxy and constructed (Keeney, 1992; Keeney & Gregory, 2005). Natural attributes are countable or physically measurable indicators that directly measure an objective; proxy attributes are distal descriptors that indirectly quantify an objective. For example, the objective ‘minimize mortality’ could be represented by the natural attribute ‘number of deaths’ or by the proxy attribute ‘number of car accidents’. Constructed attributes use ordinal scales that describe performance against objectives, linking numerical and/or narrative descriptions (Keeney & Gregory, 2005).


We conducted a systematic literature search of the Web of Science (ISI) database from 01/01/1980 to 31/08/2012. We selected the following 12 terms as keywords: MCDA, MCDM, multi-criteria, multicriteria, multiattribute, MAUT, MAVT, Analytic Hierarchy Process, Outranking, ELECTRE, PROMETHEE and NAIADE. Each of these terms was combined (AND) with the keyword social; all queries were combined (OR) to avoid duplication of articles. We refined results by the following 14 Web of Science categories: environmental sciences, social sciences interdisciplinary, zoology, environmental studies, engineering environmental, agriculture multidisciplinary, water resources, ecology, marine freshwater biology, multidisciplinary sciences, forestry, agricultural economics policy, biodiversity conservation and fisheries.

Our search resulted in 465 papers, which were then screened manually to identify a subset according to the following conditions: (1) Application of MCDA methods to an environmental problem, (2) Identification of human impacts or social change process objectives, (3) Definition of alternatives, estimating their potential consequences and performance. The subset satisfying these three conditions comprised 119 articles (see Appendix S1 in Supporting Information), which were categorized according to year of publication, MCDA method and domain. Furthermore, we recorded the objectives representing social change processes or human impacts, along with the attributes that measured them. Finally, we identified the extent to which stakeholders participated throughout the following MCDA stages: (i) formulation of objectives, (ii) selection of attributes, (iii) identification of alternatives, (iv) estimation of consequence and (v) construction of value models.


The number of MCDA publications meeting our search criteria have increased notably in the last decade, with 91% of all papers published since 2000 (Fig. 1). In Table 1, we present a description of the 119 articles included in the review, categorized by journal, methods and domains. Articles came from 58 journals; only 11 journals had ≥ 3 selected articles. We identified several MCDA methods, different adaptations of compensatory, non-compensatory and programming techniques; more than 80% of articles come from Multiattribute value (Utility) theory (MAVT/MAUT), Outranking (NAIADE, PROMETHEE, ELECTRE) or Analytic Hierarchy Process (AHP) families. The case studies include 15 domains: energy systems, water management, waste management and forest resources are the most common (72%).

Table 1. Articles included in the review (n = 119) according to journal (with 3 or more articles published), MCDA methods (used in 2 or more articles) and research domains (relevant to 2 or more articles)
Name n Name n Name n
Energ Polic14MAVT/MAUT33Energy system27
J Environ Manag11AHP30Water management26
Ecol Econ8NAIADE/Social MCE12Waste management22
Water Resour Manag7PROMETHEE10Forest resources11
Waste Manag5ELECTRE8Marine Resources8
Environ Manag4Outranking (others)4River management7
Nat Hazards3Mathematical Programming4Conservation Biology7
Environ Plann C3MACBETH2Invasive species5
Risk Analysis3Multiple6Contaminated sites2
Eur J Oper Res3Others10Others4
Figure 1.

Year of publication for the articles included in the review (n = 119 articles). A detailed list of these is compiled in Appendix S1 in Supporting Information.

Human impacts and social change processes

We identified 252 objectives that represent human impacts or social change processes. The majority of the papers (68%) include only one or two social objectives in the decision model (mean = 2.1). In Table 2, we present the objectives grouped according the values that they represent. Human impacts were more frequently captured than social change processes. We found 26 types of values relating to human impacts; of these, health impacts (minimize), social acceptability (maximize), aesthetics/scenic impacts (minimize) and recreation opportunities (maximize) were the most common values, collectively representing 53% of the human impact objectives. The breadth of concerns captured by social change processes was considerably narrower, focusing mainly on employment opportunities (maximize) and damage to iconic structures (minimize).

Table 2. Human impact (n = 188) and social change process (n = 64) objectives formulated in the case studies. Some authors formulated two or more objectives for one value in the same case study. In human impacts, we present values with two or more applications
Human impactsSocial change processes
ValueNumber of objectivesValueNumber of objectives
Health impacts (min)31Employment (max)39
Social acceptability (max)30Damage to iconic structures (min)10
Aesthetics/scenic view impacts (min)21Population affected (min)6
Recreation (max)17Public access (max)4
Educational/learning opportunities (max)10Land requirement (min)2
Social benefit (max)10Migration (min/max)2
Social conflicts (min)10Infested houses (min)1
Cultural and values impacts (min)9  
Local participation (max)7  
Community quality of life (max)6  
Social justice (max)6  
Impact on local communities (min)5  
Noise (min)5  
People's perception of risk (min)4  
Development (max)3  
Political cost (min)3  
Community vulnerability (min)2  

In a structured decision-making process, the connection between human impacts and social change processes is generally expressed in terms of links between fundamental and means objectives. Human impacts are generally formulated as fundamental objectives, indirectly measured by social change processes (means objectives), which become proxies attributes in this context (Table 3). Each of these links assumes a cause-and-effect relationship: a reduction in unemployment will increase social acceptability; greater monetary investment in the local economy will improve community quality of life; increasing potable water access will improve social justice; and so on. Generally, these cause-and-effect relationships are self-evident, but stakeholders may vary in their perception of the strength and direction of these associations.

Table 3. Links between human impacts (fundamental objectives) and social change processes (means objectives)
Fundamental objectiveMeans objectiveAttributeReferences
Social acceptability (max)Unemployed population (min)Number of people unemployedAchillas et al. (2010), Banias et al. (2010)
Community quality of life (max)Impact on local economy (max)Millions of dollarsApostolakis & Pickett (1998)
Social justice (max)Population with potable water (max)Percentage of total population using potable water systemKang & Lee (2011)
Aesthetic impacts (min)People who view the plant or plume (min)Number of people who daily view the plant or plumeKeeney & Sicherman (1983)
Health impacts (min)Pollution (min)Pollutant emissions per yearDorini et al. (2011), Afgan & Darwish (2011)

Natural, proxy and constructed attributes

Attributes are flexible constructs, which interchangeably measured a wide range of objectives (Fig. 2). For example, minimizing damage to iconic structures is a frequently used objective, and different attributes were formulated, for instance, number of locations impacted (natural) (Apostolakis & Pickett, 1998); proximity to archaeological sites (proxy) (Pavloudakis et al., 2009) or ordinal scale that verbally describes the relative performance of the alternatives (constructed) (Liu et al., 2010).

Figure 2.

Constructed (n = 142), natural (n = 50) and proxy (n = 60) attributes according human impacts (n = 188) or social change processes (n = 64) objectives. Black = constructed, mid-grey= natural, light grey = proxy.

Constructed attributes were the most commonly employed (56%), although natural (20%) and proxy (24%) attributes were also relevant. Human impacts were generally captured with constructed scales, but many case studies showed that quantitative direct (natural) and indirect (proxy) attributes can be used; proxy attributes are useful expressing causal links between fundamental and means objectives, while natural attributes are mainly encountered in health impacts objectives. Social change processes were generally estimated directly with natural attributes: number of new jobs, number of iconic structured damaged and number of individuals affected were normally included. However, in situations where quantitative data are not available, constructed scales were also used.

We found high levels of potential ambiguity in the formulation of constructed attributes; case studies generally failed to develop or communicate adequately well-defined scales. This is particular relevant for aggregative compensatory weight methods (e.g. MAVT/MAUT), because the range of consequences associated with alternatives for any one objective is rescaled using value functions. The case studies we encountered assumed linear value functions for all attributes, but the basis of this assumption was generally unspecified.

Stakeholder engagement in the MCDA process

Estimating human impacts or social change processes can involve public participation, but is not a requirement for MCDA: 42% of papers (n = 50) include stakeholder engagement in one or more stages of the decision-making process. The level of stakeholder involvement varied across the case studies, depending on the characteristics of individual decision problems and the experience and preferences of the analyst (Fig. 3). Construction of value models (establishing preferences between de objectives) was encountered across all case studies. Identification of alternatives and consequences is less frequent: stakeholders may lack the technical background required to make credible estimation of some consequences (Hendriksen et al., 2012), but considering human impacts may be more appropriate if using the knowledge and experiences of those exposed to the benefits and costs of alternatives (Becker et al., 2004).

Figure 3.

The extent of participation across the MCDA process. Fifty case studies included stakeholder engagement at some stage of the process (= 50 articles).

We categorized articles according three general approaches to stakeholder engagement: (1) Value models (n = 9): stakeholders were only involved in articulating the importance of objectives and the elicitation process generally resolved trade-offs or established ordinal preferences, (2) Objectives and value models (n = 19): stakeholders also participated in problem formulation, (3) Whole process (n = 22): stakeholders actively participated in all phases of the decision-making process; they formulated alternatives and estimated their potential consequences. We found these approaches across all MCDA methods.


We found that MCDA practitioners have increasingly integrated social concerns in the analysis of environmental problems, estimating the potential impacts, and developing participative procedures for stakeholders. We encountered this trend in a wide variety of journals, methods and domains, which underlines a broadly based maturation of the treatment of social values in MCDA. Several methods have been formally proposed to incorporate stakeholders; the most commonly used are social multicriteria evaluation (MCE) (Munda, 2004), deliberative multicriteria evaluation (Proctor & Drechsler, 2006), multicriteria approval and voting theory (Fraser & Hauge, 1998; Laukkanen et al., 2004), participatory AHP (Antunes et al., 2011), and several variants of MAVT/MAUT (e.g. Liu et al., 2012).

Although public participation can occur in all decision-making stages (Becker et al., 2004; Jonsson et al., 2007), we found that stakeholders engage differently across case studies, demonstrating that this process is neither simple nor homogenous. Applications show that MCDA clarifies stakeholders’ positions, emphasizing social learning, understanding and facilitating public acceptance (Hobbs & Meier, 2000; Stagl, 2006). A participatory process helps to identify impacts that decision-makers or experts may not have considered (Lahdelma et al., 2000); likewise, iterative processes that refine objectives and alternatives according to stakeholders’ values may improve final solutions (Nordstrom et al., 2010).

Following Slootweg et al. (2001), we deconstruct the social impact concept into two components: human impact and social change process. This is useful to distinguish fundamental and means objectives and to encourage explicit understanding of cause-and-effect relationships. We found that MCDA applications focus mainly on human impacts associated with fundamental objectives, rather than social change processes, associated with means objectives. This is contrary to past observations in social impact assessment literature, suggesting a majority of studies use measurable and politically convenient indicators, generally represented by social change process (Vanclay, 2002; Schooten et al., 2003). Although social sciences have developed several conceptual frameworks and tools to analyse the social dimensions of decisions (Becker & Vanclay, 2003; Taylor et al., 2004), our review of the literature encountered a dearth of applications demonstrating clarity in the prediction of human impacts, social change processes and potential causal relationships between them. This lack of a sound theoretical basis is also common in social impact assessment applications (Dietz, 1987; Becker et al., 2004).

We found that constructed attributes, based on ordinal scales, are the most common approach to estimate human impact objectives. Analysts need to be mindful of the potential for ambiguity in the assessment of performance alternatives when using constructed scales, especially in elicitation of marginal value functions and trade-offs among objectives (Keeney & Gregory, 2005). Potential ambiguities are prevalent in these scales, mainly because of the variable interpretation of ordinal scales (Keeney & Gregory, 2005). Assuming linear value functions for constructed scales requires that differences between each consequence level be equidistant (von Winterfeldt & Edwards, 1986). Natural attributes use plainly understandable scales (e.g. number of new jobs, number of accidents), reducing cognitive errors associated with different interpretations of the attribute (Keeney & Gregory, 2005). Proxy attributes, which establish links between human impacts and social change processes, may not always be well understood, because decision-makers must process additional information associated with cause-and-effect relationships (Keeney & Raiffa, 1976).


Our analysis of the published literature demonstrates the versatility and capacity of MCDA methods to capture social impacts and integrate stakeholder perspectives. MCDA applications are flexible in their capacity to adapt to various types of problems, including those characterized by high uncertainty and a strong imperative for stakeholder participation. Greater rigor and communication in the selection of social attributes could substantially improve the broad application of MCDA methodologies to environmental problems.


This work was supported by the Australian Centre of Excellence for Risk Analysis (ACERA). R.A.E. is funded by a PhD scholarship (BECAS-CHILE, Chile) and receives support from the Melbourne Sustainable Society Institute at the University of Melbourne. We thank Yung En Chee, Sana Bau and three anonymous reviewers for valuable comments on a draft manuscript.


Rodrigo A. Estévez is a sociologist, MS Biology and PhD Science candidate at the University of Melbourne. His research focuses on decision analysis in a range of areas, including conservation biology, environmental planning and pest management.

Terry Walshe teaches environmental risk assessment at the University of Melbourne. He has contributed to risk and decision analysis in a range of settings, including forest management, water resources, public health and biosecurity.

Mark A. Burgman is Director of the Australian Centre of Excellence for Risk Analysis and the Adrienne Clarke Chair of Botany in the School of Botany at the University of Melbourne. He works on ecological modelling, conservation biology and risk assessment. He worked as a consultant ecologist and research scientist in Australia, the United States and Switzerland during the 1980′s before joining the University of Melbourne in 1990.