Rationales for and limitations of preferred solutions for multi-functionality problems in LCA: is increased consistency possible?

The first tier solution in the ISO 14044 multi-functionality hierarchy is to avoid allocation by either subdividing the process (subdivision) or by “expanding the product system to include the additional functions provided by the co-products.” Alternative interpretations of the system expansion definition are possible. One option, and literal interpretation of ISO 14044, is that system expansion simply requires analysis and reporting at the level of all co-products, i.e., the system is modeled avoiding allocation (for example, as per the ETSI Annex 0.3 illustration for mobile phones that deliver voice, SMS, and internet browsing functions). However, if reporting impacts at the individual co-product level is desired, system expansion according to this definition may not be useful (Guinée et al. 2002; Wardenaar et al. 2012). This may be particularly the case when trying to attribute resource-use/emissions to a specific product to reflect its current share of burdens using the attributional data modeling approach (see below).

Although not referred to in ISO 14044, some practitioners consider that system expansion is commensurate with substitution (for example, see Tillman et al. 1994; Ekvall and Weidema 2004; Lund et al. 2010). Substitution refers to identification and modeling of mono-functional processes external to the considered system, which yield products or functions that are equivalent to those of the co-products of the multi-functional process of interest. These are added to or subtracted from the system model in order to achieve a desired functional unit. Most commonly, these inventories are subtracted from the inventory of the original multi-functional process in order to isolate the remaining inventory attributable to the co-product of interest.

Repeated in a variety of publications (Ekvall and Tillman 1997; Finnveden and Lindfors 1998; Ekvall and Finnveden 2001), the argument regarding commensurability between system expansion and substitution may be traced back to Tillman et al. (1994), who suggested that the two approaches are “conceptually equivalent.” The feasibility and meaningfulness of this approach may, however, depend on the data modeling approach adopted, i.e., attributional versus consequential modeling.

Proponents of physical allocation primarily view attributional LCA as a physical, natural science-based modeling framework, where the (physical) flows of material and energy characteristic of a product system (life cycle inventory) are compiled and modeled in such a way as to support quantification of associated burdens (life cycle impact assessment) on the basis of direct, physical relationships between inputs, outputs (co-products and emissions), and impacts. Where subdivision or system expansion (understood simply as modeling and reporting at the system-level rather than modeling substitution scenarios) is not possible, proponents of this approach hence ascribe to the second tier of the IS0 14044 multi-functionality hierarchy, which stipulates that “where allocation cannot be avoided, the inputs and output of the system should be partitioned between its different products or functions in a way that reflects the underlying physical relationships between them.” The prioritization of this approach is seemingly supported by the further guidance to the hierarchy specified in ISO 14044, which states that “the inventory is based on material balances between inputs and outputs. Allocation procedures should therefore approximate as much as possible such fundamental input/output relationships and characteristics.” It is similarly in line with the natural science approach that is prioritized in ISO 14040 (clause 4.1.8) (ISO 2006b). Other standards that provide more prescriptive guidance than ISO 14044 and also prioritize physical allocation include the European Commission PEF/OEF guides (EC 2013a, b) and the ETSI TS 103 199 (for LCAs of ICT equipment, networks, and services) (ETSI 2011). The latter requires allocation between services and equipment, etc. based on criteria such as use time and traffic.

Contrary to ISO 14040, however, this perspective rejects that where natural science approaches are not possible, then approaches from social and economic sciences or decisions based on other value choices should be used (i.e., the third tier of the ISO 14044 hierarchy) on the basis that the latter may produce model results of questionable physical representativeness. Rather, it is held that attributional LCA model outputs will only be representative, meaningful, and useful for environmental management if they actually reflect the physical reality associated with the modeled processes (Pelletier and Tyedmers 2011, 2012).

Due to this emphasis on maintaining physical realism in LCA models, some proponents of this approach have been strongly critical of the use of market information in LCA, wherever use of such information results in distortions of the physical relationships that LCA models are thought to be intended to represent (Pelletier and Tyedmers 2010, 2012 and references therein). Examples of instances where the use of market information is thought to distort the physical realism of LCA models include the following: consequential data modeling based on assumed but inaccurate, market-mediated product substitutions; attributional data modeling using economic allocation; and where the distinction between co-product and waste applied in LCA is based on economic value (see respective sections).

For the analytical results from LCA studies to be relevant for environmental management, however, some authors have underscored that it is also crucial that such solutions not be arbitrary (as is often the case when solutions such as mass-based allocation are generically applied to multi-functionality problems without regard to the specific context) (Pelletier and Tyedmers 2011, 2012). Toward this end, it is suggested that the current wording “underlying physical relationship” in the ISO hierarchy be revised to specify “relevant underlying physical relationship.” An example of relevance in this context might be a physical property relevant to the defined primary function of the co-product of interest that is also a reasonable common denominator for allocation between the co-products. In this way, the model will result in a direct quantification of the environmental implications of providing the defined function (or a component thereof) based on physical relationships between the inputs and outputs (Pelletier and Tyedmers 2011, 2012). Attributed burdens can subsequently be compared among alternative means of providing the same functional attribute via competing processes. Where several functional attributes are relevant (for example, protein, energy, or essential fatty acids for seafood processing co-products), each can subsequently be applied in turn as allocation criteria in order to understand the environmental implications of operating the product system for specified ends (i.e., the various functional attributes it provides). This would appear to be consistent with the ISO 14044 recommendation for a sensitivity analysis where several allocation criteria are potentially relevant. Exergy or other thermodynamic measures may provide a basic “common denominator” for physical allocation where necessary and/or appropriate.

The physical approach has been advocated by several authors in a variety of contexts—in particular, with respect to food production systems (Lundie et al. 2007; Schau and Fet 2008; Pelletier and Tyedmers 2010; Ziegler et al. 2012; Pelletier et al. 2014). For example, Lundie et al. (2007) develop a physicochemical allocation matrix for the dairy industry and highlight the importance of industry specific allocation procedures that reflect causal physical relationships. In a similar vein, Schau and Fet (2008) propose that “biological causality,” in direct parallel to the more generic term “physical causality” used in ISO 14044, may be most appropriate in many food LCA contexts.

An example of using a physical relationship as a basis for allocation is apportioning burdens in a soy biodiesel feedstock supply chain between the oil (used for biodiesel) and meal (used for animal feed) fractions produced by processing soy beans based on their respective energy content. The relationship could be deemed appropriate from a physical perspective because (1) the relevant function of the co-product of interest (soy oil) is as an energy carrier, but energy also represents a potential common denominator for the functions of both co-products (i.e., feed energy and fuel energy); (2) for the crop system, the amounts of material/energy inputs allocated by the plants to producing the protein (soy meal) and fat (soy oil) molecules is reflected in their respective energy densities; and (3) allocation based on energy content produces results that reflect the efficiency of the product system in providing the defined function.

Following the ISO 14044 (2006a, multi-functionality hierarchy, “where a physical relationship alone cannot be established or used as the basis for allocation, the inputs should be allocated between the products and functions in a way that reflects other relationships between them.” In practice, allocation in proportion to the economic value of co-products has been the prevalent (if not exclusive) application of this provision. Moreover, despite its current placement (as an example) in the lowest tier of the ISO 14044 allocation hierarchy, economic allocation is one of the most widely applied multi-functionality solutions in published LCA studies across sectors (Werner and Richter 2000; Thrane 2006; Ayer et al. 2007; Beccali et al. 2010; Van der Voet et al. 2010; EC 2010).

A variety of closely related rationales have been advanced in support of economic allocation. The most common of these is that the generation of economic value is what motivates production processes; hence, the share of value attributable to the production of co-products constitutes an appropriate basis for apportioning “responsibility” for the associated environmental burdens (Peereboom et al. 1999; Guinée et al. 2004; Mendoza et al. 2008; Weinzettel 2012). With respect to identifying the functional unit, ISO/TR 14049 (2012) states that “the functions are typically related to specific product or process properties, each of which may fulfill specific needs and thereby have a use value, which typically creates economic value to the supplier of the product.” Huppes (1993) is also explicit in this line of argument, stating that “in a social sense, the value created causes the process. If no value is created, the process will soon cease to exist and no environmental interferences will result.” Analogously, Peereboom et al. (1999) argue that “from a chemical engineering point of view, it is more logical to use mass allocation rather than economic allocation. On the other hand, economic allocation better represents the societal cause of the emissions.”

This “social causality” argument is intended to parallel the “physical causality” rationale that may be interpreted to underpin multi-functionality solutions in the first two tiers of the ISO 14044 hierarchy. Central to this argument is the corollary assumption that allocation should reflect the principle of “fairness” (Kloepffer 1996; Frischnecht 1998, 2000).

A closely related argument advanced by Ardente and Cellura (2012) is that economic allocation is merited where co-products/co-services have intrinsic qualities that physical parameters cannot adequately reflect. Here, it is recognized that prices represent aggregate proxies for the complex quality attributes inherent to many products and services (e.g., artistic, cultural, or other subjective properties such as taste or beauty), which cannot be communicated in terms of single physical attributes. Hence, differences in the prices of co-products reflect the balance of the qualities of the products that are decisive to user choice. This argument also corresponds to the previously described rationale for allocation based on social causality and for distributing responsibility for burdens in proportion to benefits conferred (fairness).

Another related argument is that economic allocation may be a preferable option where other (physical) criteria result in the attribution of a large proportion of burdens to low-value co-products (as for example, the joint mutually dependent mining of gold and copper, as described by Weinzettel 2012). The implicit rationale here is that LCAs should not produce results that are counter-intuitive (i.e., because they are contrary to price ratios or to preconceptions regarding the relative importance of co-products, generally). Again, this argument may be related to intuitions regarding social causality and fairness/responsibility.

A final rationale for economic allocation that has been described is that it facilitates producing LCA model outcomes that may incentivize desired behaviors—in particular, increased use of low-value co-products (for example, see Weinzettel 2012) in order to increase industrial ecological efficiencies. Similar arguments have been made in support of basing the distinction between products and wastes in LCA on economic value. Here, products are defined as those output flows with positive market value, whereas wastes are those with zero or negative market value. This criterion has been adopted or advocated in several contexts to date (Huppes 1993; OECD 1998; Guinée et al. 2004; EC 2010; WRI/WBCSD 2011; Ardente and Cellura 2012; Weinzettel 2012).

These related rationales in support of economic allocation point toward specific conceptions of the nature and purpose of LCA, along with the modeling decisions necessary to satisfy them. Following the distinction made by Tillman (2000) between cause-oriented and effect-oriented causality, those arguments emphasizing social causality/fairness might be classified as “cause-oriented” whereas those emphasizing incentives as “effect-oriented.” In the case of arguments based on social causality (and the corresponding link to the principle of fairness), the modeling choice (economic allocation) is made for the purpose of ensuring that the model outcomes fairly apportion burdens in relation to benefits (expressed by revenues) gained. In the case of arguments for incentivizing desired behaviors, the modeling choice is made for the purpose of achieving a pre-ordained desirable social outcome (for example, reduced waste). At root of both is the notion that LCA is not a purely physical modeling framework but rather one that can/should accommodate normative social orientations. In this light, a core purpose of LCA is to provide support to furthering such normative orientations. From this perspective, the current lowest-tier status of economic allocation in the ISO multi-functionality hierarchy is inappropriate. This could only be resolved by modifying the current hierarchy such that economic allocation is prioritized wherever allocation is necessary.

A representative example of economic allocation is apportioning burdens at a fur farm between the co-products, fur (0.5 kg) and meat (2 kg), based on their relative market values (3,333 euros/kg versus 5 euros/kg, respectively). In this case, 99.4 % of burdens are allocated to the fur and 0.6 % to the meat. The core argument in support of such a solution is that the relevant properties of the co-products that are determining for consumer preferences (and hence producer motivations to run the fur farm) are very different (for example, taste/texture in the case of meat versus softness/prestige in the case of fur). They hence cannot be accommodated by system expansion (if co-product level reporting is desired and/or appropriate substitutes cannot be identified) or physical allocation strategies (which may have the counter-intuitive effect of attributing the majority of burdens to the lower value meat co-product). In the case of the latter, one could imagine a scenario where a reduction in demand (and revenues) for the meat co-product due to high attributed burdens would result in its disposal rather than productive use: in this case, all impacts would be attributed to the fur. In contrast, economic allocation in this context apportions responsibility for burdens to co-products using a proxy (revenues) that proponents of this approach believe more accurately and fairly reflects these complex, socially causal attributes.

For each of these three schools of thought regarding preferred multi-functionality problem solutions, avoiding allocation via subdivision is a preferred solution. Beyond this initial point of agreement, however, we suggest that none actually map directly with the existing ISO 14044 hierarchy. Instead, three alternative hierarchies are implied, one of which corresponds to consequential data modeling applications and the other two to attributional data modeling applications (Table 1).

It is difficult to conclude with any certainty to what extent researchers choose to ignore the ISO 14044 hierarchy based on specific personal interpretations of the nature of LCA and the preferred multi-functionality solutions that support them. In some cases, it may equally be the case that researchers interpret the “wherever possible” and “should” terms employed in the descriptions of each tier of the ISO hierarchy as having precedence over the preceeding “shall” prescription to observe the hierarchy in solving multi-functionality problems. Also unclear is the extent to which the ISO 14044 requirement that modeling choices reflect the goals and scope of the analysis is consistent with prescription of a hierarchy, regardless of its substantive content.

We posit that, at least in some instances, choice among these hierarchies described above will, indeed, primarily reflect a practitioners pre-existing orientations and understanding of the nature/purpose of LCA as opposed to simple observance of the ISO 14044 hierarchy. For consequential LCA, system expansion + substitution is the only modeling choice that supports the corresponding defined nature and purpose of LCA (i.e., change-oriented, system-level modeling, taking into account indirect, market-mediated effects). Allocation should, in theory, never be necessary. From the physical perspective, maintaining the natural science basis of the model constitutes the core criterion for selection of a multi-functionality solution. In this light, assumed market-mediated substitutions are questionable in terms of the extent to which they reasonably represent physical reality, and economic allocation does not, in any circumstance, constitute a defensible modeling choice. From the socioeconomic perspective, in contrast, fairly reflecting social causality and/or incentivizing particular behaviors are emphasized; hence, allocation in proportion to the economic value of co-products is preferred. In corollary, arbitrary applications of physical criteria which do not reflect such social causality are thought to produce arbitrary results and are hence viewed as undesirable. Each of these positions is internally consistent but seemingly mutually exclusive. We hence question the extent to which they can actually be reasonably accommodated in the context of a single hierarchy, as in ISO 14044, unless one approach is clearly prioritized, along with a supporting rationale, over the other.

Some LCA practitioners may claim to confer equal validity to the different supporting understandings of the nature and purpose of LCA we describe here, and instead advocate adherence to a more general “causality” principle for the partitioning of impacts on a case-by-case basis. In other words, the allocation strategy would be chosen with respect to context, with the practitioner selecting from among several possible alternative causality principles (including physical, economic, or possibly others) related to the selected functional unit(s) and/or the goals of the LCA. However, as we demonstrate, arguments for particular forms of causality are based on very different underpinning assumptions regarding the nature and purpose of LCA. It is hence difficult to see how advocating causality in general as a basis for allocation would contribute to improved clarity and consistency in practice.