| Bottom | | Front page |

Environmental Project, 863

Market information in life cycle assessment

Contents

Preface

This report was prepared within the Danish LCA methodology and consensus creation project during the period from 1997 to 2003.

The report is one out of five technical reports to be published by the Danish Environmental Protection Agency and dealing with key issues in LCA. The reports were prepared as background literature for a number of guidelines on LCA, planned to be published by the Danish Environmental Protection Agency during the autumn of 2003. The reports present the scientific discussions and documentation for recommendations offered by the guidelines. The reports and guidelines developed within the project are presented in the overview figure below.

A primary objective of the guidelines has been to provide advice and recommendations on key issues in LCA at a more detailed level than offered by general literature, like the ISO-standards, the EDIP reports, the Nordic LCA project and SETAC publications. The guidelines must be regarded as a supplement to and not a substitution for this general literature.

It is, however, important to note that the guidelines were developed during a consensus process involving in reality all major research institutions and consulting firms engaged in the LCA field in Denmark. The advice given in the guidelines may thus be considered to represent what is generally accepted as best practice today in the field of LCA in Denmark.

The development of the guidelines and the technical reports was initiated and supervised by the Danish EPA Ad Hoc Committee on LCA Methodology Issues 1997-2001. The research institutions and consulting firms engaged in the development and consensus process are:

COWI, Consulting Engineers and Planners (Project Management) Institute for Product Development, the Technical University of Denmark dk-TEKNIK ENERGY & ENVIRONMENT
The Danish Technological Institute
Carl Bro
The Danish Building Research Institute
DHI - Water and Environment
Danish Toxicology Institute
Rambøll
ECONET
National Environmental Research Institute

This technical report was prepared by Bo P. Weidema, based on research and draft material from different research teams:
For chapter 2: Claus Petersen¹, Bo P. Weidema², and Anne-Merete Nielsen²,
For chapter 3: Bo P. Weidema², Henrik Wenzel³, and Klaus Hansen⁴,
For chapter 4: Bo P. Weidema², and Anne-Merete Nielsen²,
For chapter 5: Bo P. Weidema², Henrik Wenzel³, Klaus Hansen² and Claus Petersen¹,
For chapter 6: Bo P. Weidema², and Nina Caspersen³.



Click on the picture to see the html-version of: Guidelines and technical reports prepared within the Danish LCA-methodology and consensusproject

____________________________________________________________
¹ Eco-net, Denmark
² 2.-0 LCA consultants, Denmark
³ Institute for Product Development, Technical University of Denmark
⁴ Danish Building Research Institute,

1 Introduction

⁵ This report provides the background for the two guidelines “The product, functional unit, and reference flows in LCA” (Weidema et al. 2003a) and “Geographical, technological and temporal delimitation in LCA” (Weidema 2003). It provides further documentation of the examples provided in these guidelines, as well as additional examples, further explanatory text, scientific background and reference to earlier methodological guidelines. It also expands on specific issues, which were not found to be of sufficient general interest to merit inclusion in the guidelines.

This report and the two guidelines that it supports, carry two key messages:

1. The fundamental rule to apply in all methodological choices in life cycle assessment is that the data used must reflect as far as possible the processes actually affected as a consequence of the decision that the specific life cycle assessment is intended to support. Thus, there is a close link between the goal or application area of the life cycle assessment and the methodological choices. This is elaborated in section 1.1.

2. Life cycle assessments, insofar as they deal with comparing potential choices between alternative products, rely heavily on market information, i.e. information on how the market affects the potential choices and how the markets will react to these choices.

Whenever possible, the above understanding has been converted to practical, step-by-step procedures for including market information when:

• defining the functional unit (chapter 3),
• defining the geographical and technological scope (chapter 4),
• handling co-products (chapter 5),
• forecasting data for processes taking place in the future (chapter 6).

For all these elements of the life cycle assessment methodology, the inclusion of market information leads to improvements, which also reduces the uncertainty of life cycle assessment results. While the methodological improvements are described in this report, the consequences for uncertainty are the topic of a separate report: "Reducing uncertainty in LCI. Developing a data collection strategy" (Weidema et al. 2003).

1.1 The relation between application areas and methodology

The methodological elements listed above are fundamentally determined by the temporal and spatial aspects of the studied systems and by the products and interest groups affected. On this basis, six well-defined application areas can be distinguished (see figure 1.1).



Click on the picture to see the html-version of: Figure 1.1

The decision-maker‘s potential influence on the different processes in the product systems increases towards the top left of the diagram, i.e. as the decision horizon becomes more long-term and as the decision relates to more specific products and geographical areas. This is further illustrated in figure 1.2. For retrospective studies (area A in figure 1.2), there is no choice to influence. For medium-term, tactical studies, high influence on specific processes throughout the life cycle is limited to studies where the product systems are very well-defined and where the decision-maker already at present has a high influence on the other actors in the life cycle (illustrated by area B in figure 1.2). Tactical aspects (i.e. contacts to be made in the product chain) may also be part of the considerations in product development, and the more long-term the development, the more ambitious one may be with respect to obtaining influence (area C). Even on a societal level, it may be possible to influence specific choices throughout the life cycle, when the products are relatively well-defined and have well-defined interest groups (including producers and users), and when the time horizon is long enough to allow the necessary regulative and technical infrastructure to be developed (area D). For the rest of the applications (area E in figure 1.2), the products are either too generic (i.e. includes several products or a group of products) or involve too many interest groups to allow a decision-maker to influence specific choices throughout the life cycle.



Click on the picture to see the html-version of: Figure 1.2

With respect to methodological choices, the most important distinction is that between the retrospective, attributional life cycle assessments of the accountancy type⁶ (typically applied for hot-spot-identification, product declarations and for generic consumer information) and the prospective consequential life cycle assessments, which study the environmental consequences of possible (future) changes between alternative product systems (typically applied in product development and in public policy making) (Tillman 1998, 2000). This distinction is further elaborated in section 1.2.

The application areas (as outlined in figure 1.1) affect the methodology in the following ways:

• The functional unit, which delimits which product alternatives can be included in the study, is affected by the time horizon of the study and by the degree of specification of the studied product (specifically defined products and long time horizons allow more alternatives to be included). This point is elaborated in section 1.4 and further in chapter 3.
• The processes to include in the product systems studied are affected by the distinction between attributional and consequential applications (including either those processes which can be associated with the product according to a chosen rule or those which are affected by a product substitution). This is elaborated in section 1.5.
• Within consequential studies, the technologies to consider and whether to include capital goods, maintenance etc., is affected by the distinctions between small/large and short-term/long-term changes. These distinctions (which are defined in section 1.3) are related to the parameters in figure 1.1, but do not exactly follow the divisions between application areas given there. The way these distinctions affect the technologies to consider is elaborated in section 4.2.
• The method for handling co-products is also affected by the distinction between attributional and consequential applications (attributional applications require economic allocation while consequential applications require system expansion). This is elaborated in section 1.6 and further in chapter 5.
• The methods to use for forecasting future processes is affected by the time horizon and complexity of the studied system (determining whether forecasts should be made by extrapolation, modelling or scenario methods), and by the amount of stakeholders affected (determining whether participatory forecasting is relevant). Furthermore, exploratory and normative forecasting may be relevant for specific applications in product development. This is elaborated in section 1.7 and further in chapter 6.

Thus, for each application area of figure 1.1, we can outline the conditions for the methodological choices to be taken:

• For attributional life cycle assessments:
  
  • As attributional LCA does not apply to comparison of alternative product systems, the functional unit does not play any important role for the assessment, and may therefore be chosen at will.
  • The processes to include are those that are deemed to contribute to the studied product.
  • Co-products are handled by economic allocation, since attributional LCA does not involve changes, which is a necessary condition for applying the system expansion procedure.

Note that when defining the goal and scope of an attributional LCA, one should be aware whether one intends later to use the results for decision making, in which case it should be carefully considered whether it is necessary and worthwhile to perform an attributional LCA or whether a consequential study is adequate and sufficient. See also the discussion in section 1.2.

• For studies of specific products, affecting specific interest groups on a medium (1-5 years) term (for product declarations, hot-spot-identification, marketing claims, and incentives and requirements for suppliers or employees):
  
  • The functional unit shall reflect the current products on the market and their obligatory properties (see definition in chapter 2).
  • The processes to include are those that are affected on short or long term by the decision supported by the results of the study (i.e. choosing the product with the market claim instead of the alternatives, following the incentives or fulfilling the supplier/employee requirements instead of continuing status-quo).
  • Co-products are handled by system expansion.
  • Forecasting of processes is done by extrapolation.
• For studies of generic products (product groups) on a medium (1-5 years) term (for generic consumer information, ecolabelling criteria and product standards, taxes and subsidies):
  • The functional unit shall reflect the current products on the market and their obligatory properties (see definition in chapter 2).
  • The processes to include are those affected on short or long term by the decision supported by the results of the study (i.e. choosing a product with the ecolabel instead of the alternatives, changing behaviour following the taxes or subsidies or fulfilling the product standard instead of continuing status-quo).
  • Co-products are handled by system expansion.
  • Forecasting of processes is done by modelling and participatory methods.
• For studies used to support societal action plans, product legislation and generic performance criteria:
  
  • The functional unit may be broadened to include alternatives assumed relevant under future conditions of availability, price, and product information.
  • The processes to include are those processes, which are affected by the decisions supported by the results of the study (typically large, long-term consequences).
  • Co-products are handled by system expansion.
  • Forecasting of processes is done by modelling and scenario methods.
• For studies used in product development and for enterprise specific performance criteria:
  
  • The functional unit may be broadened to include more alternatives in all parts of the product chain, when assumed to be controlled by the decision maker and relevant under future conditions.
  • The processes to include are those processes, which are affected by the decisions supported by the results of the study (long-term consequences, small or large).
  • Co-products are handled by system expansion.
  • Forecasting of processes is done by modelling and scenario methods. For processes where a large degree of control is assumed, also exploratory and normative methods may be applied (see chapter 6 for definitions).

Unfortunately, the above recommendations are not in complete accordance with the recommendations from the Dutch methodology project (Guinée et al. 2001), which was carried out simultaneously with the Danish project of which this report is a result. In spite of close agreements on many important basic concepts (see Guinée 1999) we did not succeed in reaching consensus on the specific recommendations to be given in our respective guidelines. The main differences between the guidelines are that the Dutch guideline restricts its recommendations to a baseline situation (applications with small, long-term consequences), and recommends an intentional disregard for market mechanisms and their consequences. In relation to the definition of the functional unit, the latter recommendation implies an assumption that there are no changes in consumer behaviour in relation to product substitutions, such as the so-called “rebound effect”, and that differences in consumer prices do not induce the consumer to spend more or less money on other products. In relation to system delimitation, it implies an assumption that all processes will react to changes in demand in proportion to the revenue obtained for the production (i.e. that relative supply elasticities equals relative prices) without any side-effects, so that the affected technology will be the average of the currently installed technology, and so that co-production does not lead to substitution and may therefore be handled by economic allocation. The argument for this intentional disregard for market mechanisms is apparently that a full modelling of market mechanisms is not practicable, and that using an incomplete model of market mechanisms may introduce large uncertainties in the modelling. Thus, the Dutch guideline opts for an incomplete description rather than an uncertain description of the markets. In this way, there is an inconsistency between the Dutch recommended methodology and the application area for which it is suggested (consequential studies), as also pointed out by several of the international reviewers (Guinée 1999). In section 2.4 we continue the discussion on the issue of market modelling, and provide further arguments for intentionally including market mechanisms and their consequences.

1.2 Discussion of attributional versus consequential LCA

The relevance of attributional LCAs has been questioned (Weidema 1998b, Wenzel 1999), because the ultimate goal even of hot-spot-identification and product declarations is to improve the studied systems:

• If an attributional hot-spot-identification identifies a number of improvement options, a consequential assessment is still needed to assess the consequences of implementing the improvements, so one might as well perform a consequential study in the first place.
• If product declarations are used by the customer to make a choice between several products, this choice should ideally be based on the environmental consequences⁷ of this choice (i.e. a specific, medium-term, prospective study), not on the historical impact caused by the products⁸.
• Likewise, if generic consumer information affects the behaviour of the consumer, this behavioural change should ideally be based on the environmental consequences of this change (i.e. a generic, medium-term, prospective study).

Click on the picture to see the html-version of: Figure 1.3

Even a question that appears retrospective at first sight (like “If I look at the world as it is now, what is the environmental contribution of car driving?”, Guinée 1999, p.5) does not appear to have a meaningful answer, except if we reformulate it as a hypothetical “historical, consequential”: “What would the world have looked like now, if we had removed car-driving?” Such historical consequential questions can be answered by applying the same consequential methodology as for prospective questions, but using current or historical data. As pointed out by Guinée et al. (2001, part 3, p. 14), outside such a consequential context, there is no objective way to separate the system of car driving from the rest of the technosphere (i.e. to draw the dotted line in the left circle in figure 1.3), since all product systems are ultimately linked (i.e. there are so many lines crossing the dotted line in the left circle in figure 1.3 that the drawing of this line will imply a number of normative cut-offs). Thus, outside of a consequential context, any separation of product systems will be inherently normative and will therefore have to be included in the question asked, i.e.: “Providing we use method X for dividing car driving from the rest of the technosphere, what is its environmental contribution?” implying that the question carries the premises for its own answer. Such questions, and the LCAs that are used to answer them, may therefore more correctly be termed attributional (Heijungs 1997, Frischknecht 1998, Hofstetter 1998) than retrospective, since they deal rather with the juridical issue of allocation or attribution of guilt, blame or responsibility, than with the natural science issue of analysing causalities and consequences, and since such questions of guilt, blame or responsibility may pertain to the future as well as the past. The term retrospective, if used at all, should then rather be used for the “historical consequential” applications (see also figure 1.4). The point made here is not that attributional questions are meaningless, but that it is impossible to give meaningful, objective answers to such questions. In terms of uncertainty, this may be considerable in consequential LCA, since current uncertain knowledge is used to assess future consequences. However, this uncertainty can be estimated and controlled, while the error that is inherent in attributional LCA is fundamentally unknowable and uncontrollable.



Click on the picture to see the html-version of: Figure 1.4

In the past, life cycle assessments have primarily been applied to consequential questions, and practitioners have sought to adjust their methodologies to reflect this objective. However, attributional methodologies have often been applied, because adequate consequential methodologies have been missing. We hope that the market-based methods presented in this report will help practitioners to apply a consequential approach more consistently throughout their life cycle studies.

The relevance of attributional LCAs have been defended with a number of different arguments, which will be treated separately here:

• Attributional LCA may be used as a pedagogical introduction to a life cycle study, since at first sight it may appear simpler: All that is needed is knowledge on current or potential suppliers and customers – other market relations may be disregarded, and data need only be collected from enterprises in one’s own supply chain. This may be useful in the early stages of a life cycle study, where there is a need simply to explore the life cycle, to increase the understanding of the product chain (Tillman 2000). An attributional LCA may pinpoint the processes and relations most important to influence in a product system (known as “hot-spot-identification”). However, this could equally well (and maybe even more sensibly) be done with a consequential LCA that tells about the consequences of producing, using, and disposing a quantity more or less of the investigated product. And this would even provide more relevant information on what parameters guide the behaviour of the investigated product systems.
• To operate an LCA-based system for environmental product declarations, there must be a generally accepted set of rules for how to perform such LCAs. Tillman (2000) doubts whether it will be possible to establish the necessary consensus within a consequential approach to LCA since this implies system expansion (see chapter 5) and use of marginal data (see chapter 4), including “an approach as to which marginals and in which way the system should be expanded.” The present report, and the twoguidelines that it supports, is nevertheless a report on an attempt at providing such consensus. And in response to Tillman’s doubt, it appears equally questionable (if not more so) that the necessary consensus and acceptance can be obtained for an attributional approach to LCA that needs to apply and justify arbitrary allocations and choices of which averages to use.
• Additivity between individual parts of a life cycle (enabling a producer to add his own environmental exchanges to those reported by his suppliers) and completeness (in the sense that only negligible parts of the product system are omitted) are both features that Tillman (2000) use as arguments for using attributional LCA. However, in the sense described here by Tillman, both additivity and completeness are also features of consequential LCA as described in this report.
• Attributional LCA are also said to be applicable in situations where no specific change is planned, as may be the case e.g. for hot-spot-identification, for setting priorities that do not immediately involve a change, or where the scale or products involved in a substitution are unknown, while it is questioned (also by Tillman 2000) how this could be done in a consequential LCA. However, a consequential LCA may very well assess the consequences of production, use and disposal of a defined quantity more or less of the investigated product. This can be done independently for any product, without prior knowledge on the specific comparison that each assessment may later be used for. Later, when specific comparisons are required, these may be obtained simply by subtracting the individual product systems. These comparisons will be valid as long as the product quantities studied are small. For larger quantities it is of course important to include any influence on the boundary conditions.
• System expansion as an important method in consequential LCA to handle systems with multiple products (see chapter 5) is thought by Tillman (2000) to imply “a larger system and thus more data to collect.” Since most LCA databases are currently based on average data without concern for market mechanisms, an LCA based on available data and default cut-off criteria will of course be easier and less time consuming to perform than a consequential LCA that must rely on not readily available market data. However, in an LCA that involves specific data collection, the procedures for consequential LCA suggested in chapters 4 and 5 specifically reduces the size of the system to investigate by excluding all processes that do not change as a consequence of the change in demand for the product under study. In contrast, a product system in attributional LCA must include, for every step (tier) of the life cycle, all specific suppliers to the previous tier (and even more individual suppliers when average data are used). Our experience shows that for more detailed LCAs that place a large demand on specific, high-quality data, the additional time spent in collecting market data (see section 2.5) will quickly be outweighed by the timesaving in having fewer processes from which to collect detailed environmental data.
• In an assessment of policy options, the decision-maker may be interested in how to change or influence the markets, and is therefore not interested in limiting the analysis to the predicted market reactions to the potential decision, as implied by a study of actual consequences. In LCAs performed for a decision-maker with a long time horizon and a strong influence on the actors and markets in the product chain (such as studies by a market-leader or studies aimed at societal action plans and legislation), the flexibility of attributional LCA to include any process of interest, may better reflect the actual flexibility of the decision maker. However, even in such cases, where the normal market mechanisms are overruled, the market-based procedures of consequential LCA (see the following chapters) will still provide a good framework for explicitly documenting this dominating influence of the decision-maker.
• Because consequential LCAs only look at the consequences of changes, it may misrepresent the “signal value” implied in a demand for an environmentally improved product if this demand does not lead to an immediate change. An example of this may be the initial immature market for ecological foods, where an increase in demand may not lead to an increase in production, because of the transaction costs of the initial small quantities or because of the time it takes to implement the new technology on the farms. An LCA should give credit to such a demand even if it does not lead to changes in production in the short term, because the combined demand of many actors would be able to overcome the outlined constraints. An attributional LCA can give such a credit, since the attribution is not dependent on any assessment of actual changes. The answer of consequential LCA is to expand the scale and time horizon for assessing the consequences so that the long-term reaction of the market to the change in demand is indeed included, and the credit therefore assigned (see section 4.3).
• Attributional LCA may be used in a context where the decision-maker wishes to support, be part of, or otherwise be associated with what is deemed to be a “good” system, or to be dissociated with what is deemed to be a “bad” system (Ekvall 1999, 2000, Ekvall et al. 2001a, b). For example, the decision-maker may wish to be associated with companies that use renewable energy sources, disregarding whether this leads to increased production of renewable energy or not. A consequential LCA would not be able to provide the sought-after information, since it only takes into account the actual consequences and therefore only gives a credit for renewable energy when an increase in the capacity of renewable energy can be expected (see section 4.3). If no change is expected in the composition of the overall output, for example when the renewable energy source is constrained, as is the case with hydropower in Europe, the consequential LCA does not give any credit (see, however, the exception dealt with in the previous bullet). In this situation, an attributional LCA may well give a credit for a supply of hydropower, simply because it is the association with the “good” system that is credited, and not whether there is any overall change in environmental impact. It should be noted that in the opposite example, where the decision maker wishes to dissociate from what is deemed a “bad” system, e.g. one associated with hazardous chemicals or ionising radiation, the consequential and the attributional LCA would both be able to supply the desired information, since it is very few systems that are downwards constrained, which implies that an explicit reduction in demand would indeed have consequences for these “bad” systems. Ekvall (2000, Ekvall et al. 2001b) seek to justify attributional LCA by referring to rule ethics (as opposed to utilitarian or situation ethics which would support consequential LCA) but acknowledges that its application would require an agreement on what is regarded as “being associated with.” This agreement would amount to a rule for allocating or attributing guilt, blame or responsibility, which cannot be made on objective grounds, as noted above. Furthermore, the concept of “being associated with” is hardly meaningful beyond a few steps backwards or forwards in the supply chain, thus rendering LCA too sophisticated a technique for identifying the relevant associations. Nevertheless, it is natural that a commissioner of a life cycle study may feel that it is more relevant to study the processes in the immediate supply chain than those actually affected by the product substitutions. It is important to clarify whether the interest of the commissioner is really in the environmental impacts of products (i.e. in LCA) or more in the environmental impacts of the supply chain as such, since the latter interest may be better handled through supply chain management.
• Ekvall et al. (2001a, b) provide a specific thought experiment where consequential LCAs would lead to an undesirable effect: The lack of credit for using hydropower may provide an incentive to create a separated, sub-optimised market for hydropower where this credit could be justified. However, the thought experiment depends on two conditions being simultaneously fulfilled, namely that the environmentally preferable process or technology is more competitive (cheaper) than the environmentally less preferable, and absolutely upwards constrained in its ability to change its capacity as a result of a change in demand. In practice, we have not been able to identify any other examples than hydropower, where these conditions occur simultaneously. Nevertheless, this is a real and undesirable effect of consequential LCA, which cannot be avoided but only alleviated or internalised by applying in this situation an additional scenario in which the separated, sub-optimised market is assumed realised, implying thus a credit to the users of the environmentally preferable technology. When so applied, this scenario would work counter to its own fulfilment and counter to the described undesired effect. This isolated undesirable effect of consequential LCA does not in itself constitute an argument for a more general use of attributional LCA.

If a company or product chain in an expanding market has several production lines, some older more polluting and some new less polluting, it may appear with a below average environmental performance in an attributional LCA that use average data, while a consequential LCA that focus only on the new production lines that will be installed, may show a performance equal to the rest of the market, since all actors on the market typically install the same new technology. It may be argued that an attributional LCA will provide an incentive for improvement of the older, more polluting production lines, in order to better compare with “green” competitors that have only newly installed production lines, while the consequential LCA does not provide the same incentive. The attributional LCA can be said to reward the newcomer that is not burdened with the old technology, while the older factory is punished for having been in business longer. However, there is a way for the older factory to avoid this, namely by separating the old and the new parts of the factory, selling the products from the old production lines to the general customer, and selling the now competitive products from the new production lines on the “green” market. This restructuring would not change the overall environmental impact and the attributional LCA would not have provided any more incentive for improvement than the consequential LCA. In contrast, consequential LCA does provide a real possibility to reward improvements in older production lines even when these are not immediately affected by changes in demand. This is possible if the producer actively links his improvements in the older production lines to increases in sales. Thereby, the customer buys both a product from the new production line and a share of the improvements of the older production lines, which obviously provides a better environmental performance than just buying a product from the new production line. In fact, such cross-subsidising between production lines need not be limited to the production lines within the same company or product chain, if the money is better spent on environmental investments elsewhere. However, to be credible, such cross-subsidies should binding and verifiable (e.g. contractual) and their existence preferably verified by an independent third party. In this way, consequential LCA allows any production to obtain an environmental “credit” when consciously affecting a specific production, while those who only do what everybody else do, obtain the same LCA result as everybody else, no matter how good or bad their average performance.

The understanding of consequential life cycle assessment as a tool for decision-support as opposed to a tool for documentation and attribution of guilt, blame or responsibility, implies a focus on the importance of system boundaries and market data. This implies also a focus on the problems involved in verifying such information, including the involvement of stakeholders, critical assessment of sources, and peer review. This is common to other decision-support tools, which are also not expected to result in unambiguous information, but rather different scenarios where the different assumptions are documented.

1.3 Product substitution

In a consequential, comparative life cycle assessment, the object of study is the environmental impacts of a potential product substitution, i.e. the replacement of one product or group of products with another product or group of products, fulfilling the same needs of the customer. We define a product also in terms of its production process, which implies that a product substitution will always imply one or more process substitutions (understood as process changes or complete replacements), and that a process substitution can also be seen as a product substitution, even when the product itself is unchanged (e.g. in terms of its physical properties).

Product substitutions may occur anywhere in the life cycle, from raw material substitutions, over substitutions in the production and use stages, to substitutions between alternative waste handling options.

In this context, several authors (Clift et al. 1998, Frischknecht 1998, Tillman et al. 1998) have suggested that a distinction between foreground and background processes can be useful. However, we have found it necessary to define these terms more strictly⁹, to understand that:
- a foreground process is a process whose production volume will be affected directly by the studied change,
- a background process is a process whose production volumes will not be affected or only be indirectly affected (i.e. only through the market) as a consequence of the increase or decrease in demand as a result of the studied change.

Life cycle assessments are typically limited to study the effects of substitutions at one specific stage in the life cycle, the range of possible substitutions at that stage being delimited by the functional unit (i.e. the functional unit typically does not specify what choices to make at other stages). The reason for this is that life cycle assessments are typically aimed at situations where the influence of the decision-maker is limited to the specific substitution studied. (i.e. most processes are in the background).

However, if the decision-maker is able to affect substitutions at different stages in the life cycle (i.e. using foreground processes for these), these substitutions may - both in principle and in practice - be specified by the functional unit, thus including simultaneously all possible choices in the study.

Even when the decision-maker is not able to directly influence any substitutions elsewhere in the life cycle (i.e. when most processes are in the background), the studied substitution at one stage in a life cycle (the foreground) may still lead indirectly to product substitutions in other life cycle stages (in the background), due to the change in demand implied by the initial substitution. These substitutions are then not included in the functional unit, but the expected result of the substitutions (in terms of affected processes and their technologies) is simply included when modelling the product systems.

Put very briefly, using the terminology of foreground and background processes: Product substitutions in foreground processes may be included in the definition of the functional unit, while substitutions in background processes are simply accounted for by including the affected processes and technologies when modelling the product systems. See also figure 1.5 and the explanatory text to this figure.



Click on the picture to see the html-version of: Figure 1.5

Explanations to figure 1.5: The substitution studied may be at the use stage (to use product A or Product B for the function P), at the production stage (to produce product A by route A¹ or A²), at the raw material stage (to use raw material R¹ or R²) or at the disposal stage. However, the choice of a specific product (say B) will typically imply a choice of production route and raw materials (R²) that is not put into question. It is only when the decision maker (in the case of the choice A or B, the user is the decision maker) has an influence on the choice of production and/or disposal route and/or raw materials use, that the other choices (e.g. A¹ or A² and R¹ or R²) can be included by the definition of the functional unit (e.g. specifying: "P produced using raw material R²", or the more conditional specification: "P produced with optimal raw material choice," which allows a comparative investigation of different raw materials). This is illustrated by the sphere of influence S². Usually, the influence of the decision-maker is more limited, typically to the choice between different products at the previous stage in the product chain (S¹). In this case, the functional unit is simply specified as "P" without indication of any specific conditions of production or disposal. Nevertheless, these choices will still be made by other decision-makers in the chain. So what will be included in the life cycle study is the expected result of these choices, i.e. the expected route of production and disposal as chosen by the decision-makers for these stages of the life cycle.

Figure 1.5. Product substitutions in relation to the sphere of influence of the decision-maker

Relating this to the application areas in figure 1.1, it can be seen that the conditions for a large area of influence (S¹ in figure 1.5) is limited to the upper left-hand corner of figure 1.1, as can also be seen in figure 1.2, namely for long-term, strategic applications involving relatively well-defined products from enterprises with a relatively large (expected) influence on the different actors in the life cycle.

For a thorough understanding of a specific product substitution, information is required on:

1. The extent of the studied substitution, where:

• small¹⁰, short-term substitutions affect only capacity utilisation, but not capacity itself,
• small, long-term substitutions affect also capital investment (installation of new machinery or phasing out of old machinery),
• large substitutions affect also the determining parameters for the overall technology development, i.e. the constraints on the possible technologies, the overall trends in the market volume, or the production costs of the involved technologies, so that the studied substitution in itself may bring new technologies into focus.

1. 2. The market segment affected, as determined by the obligatory product properties (i.e. properties that a product “must have” for a customer in that segment to accept the products as comparable and thus substitutable).
2. Product availability, i.e. whether the market situation actually allows a choice between the products to be made (markets and/or production technologies may be constrained by market failures, declining markets, regulations, or shortages in supply of raw materials or other necessary production factors).
3. The positioning properties of the products ("nice to have"), as well as price and information, which influences the degree to which a potential product substitution will actually be realised.

This is further elaborated in chapter 2 of this report.

1.4 Defining the functional unit

The functional unit plays several roles in a life cycle study:

• First, it serves as a reference unit, to which all other data in the study relates.
• Secondly, it reflects the amount of substitutions that the decision maker desires to influence, as outlined in section 1.3 (see especially figure 1.5),
• Thirdly, it is the basis of equivalence, when comparing different product alternatives in consequential studies.

For the latter role, the obligatory product properties must always be taken into account. To obtain a precise and unambiguous definition, it has proven useful to analyse in detail the actual obligatory product properties required by the relevant geographical markets and market segments.

A company-internal study comparing different options in the product development, may define additional properties as obligatory for their own brand, although they are only regarded as positioning properties on the general market (and would be determined as such in a more generic life cycle assessment comparing this brand with other brands).

Whether the other aspects of product substitution (availability, positioning product properties, price, and information) should also be taken into account depends on the time horizon of the study. In studies with a long time horizon (e.g. product development or strategic management), it may be reasonable to compare two products, for which substitution cannot be immediately realised, but where it is assumed that substitution will be realised under specific, future conditions of availability, price and product information. The shorter the time horizon of the study, the less relevant it is to include product alternatives, for which substitution is not likely to be realised under the present market conditions.

Two products may be compared even when they differ with respect to positioning properties. If these positioning properties can be determined to fulfil specific functions, equivalence between the products under comparison must be ensured by treating these functions as co-products (see section 1.7 and chapter 6).

1.5 Market-based system delimitation

¹¹As mentioned above, the processes to include in a consequential life cycle study - and the technologies of these processes - are the processes and technologies actually affected by the studied product substitution (as defined by the functional unit). To identify the processes affected, all four types of information on product substitution mentioned in section 1.3 are relevant. In chapter 4, we present a step-wise procedure for identifying the affected processes through a formalised treatment of the last three types of information.

Figure 1.6 can be used to illustrate the difference between such a consequential, market-based system delimitation, and the more traditional system description based on an attributional or accountancy approach, where material and energy flows are followed mechanically from process to process. In the figure, it is shown how a change in volume of one process (process 1 to the right) leads to a change in the demand for one of the raw materials to this process. However, many different technologies or processes can meet the specifications for this raw material supply. This is illustrated by the fully drawn processes to the left, which together make up the suppliers to the market. Now, the traditional system delimitation will either include an average of all these processes, weighted by their respective production volumes, or just include that specific process, which represents the current supplier to process 1, here illustrated by the fat box.

When applying an average, the result can be seriously affected by the delimitation of the market on which the average is taken. For example, it will make a large difference whether you regard the Nordic electricity market as one (relatively closed) market, so that Danish electricity consumption is calculated as an average of Danish, Finnish, Swedish and Norwegian electricity production, or whether it is assumed that Denmark is a market in itself (which is often seen in life cycle assessments). If we choose to look at the average for Denmark, which is not a closed market, it is decisive whether the average is calculated from the Danish production alone or whether you take into account the exchanges with the neighbouring markets, and how you take this into account, e.g. whether you calculate with Danish production plus import-mix (in periods with much available hydropower in Norway and Sweden), with Danish production plus import-mix minus export-mix (in periods with little hydropower available) or just Danish production plus net import/export (thus disregarding transit-trade). For Switzerland, having a large degree of transit-trade, Ménard et al. (1998) have shown how such different assumptions affect the average from 21 g CO₂ (Switzerland’s own production) over 140 g CO₂ (Switzerland plus import minus export) to 500 g CO₂ (UCPTE average, in that UCPTE can be regarded as a relatively isolated electricity market like the Nordic). The recommendation of Ménard et al. (1998) is to use the model that disregards transit-trade (48 g CO₂) with the argument that this best reflects the actual market conditions. It should be clear from this example that averages can be highly debatable, and possible arguments for preferring one average over the other is actually often market-based. This may in itself be regarded as a serious argument for taking the full consequence, and use a truly market-based system delimitation instead of the average approach.

Click on the picture to see the html-version of: Figure 1.6

A market-based system delimitation will first determine the actual geographical and temporal market boundaries (see section 2.1), which in the electricity example will lead to the identification of the Nordic and the UCPTE markets as being the relevant electricity markets.

Within each such market, a market-based system delimitation will then - instead of considering averages - investigate whether any of the processes delivering to the market are constrained in their capacity to change as a result of a change in demand from process 1 (figure 1.6). These constrained processes are marked with C’s.

It should be noted, that also in a market-based system delimitation, the directly delivering process (the fat box) may well come into play. However, this requires that the change in demand overcome the constraints on the process, so that its production volume is actually affected. Thus, the change in demand must to some extent put the market forces out of play to ensure that a capacity adjustment is actually taking place in that specific process. This may especially be the case if the customer has a controlling influence on the supplier (possibly in the form of a monopoly position).

Another aspect of the market-based delimitation is that it investigates whether the change is so large that it gives room for new technologies (illustrated by the perforated box in the upper end of figure 1.6) or that it can affect one or more of the identified constraints, so that a C-marked technology can anyway come into play.

Now, if the technologies/processes in figure 1.6 are arranged in such a way that the most economical are at the top (this is often also the newest and most efficient ones, but this depends also on the cost structure, including the wage level) and the least economical at the bottom (often the older, less efficient), it will typically be either the upper or the lower unconstrained process that will be affected by a change in demand – depending on whether the market is expanding or shrinking. Contrary to the average, we are rather concerned with the extremes here.

If we now focus on the situation with an expanding market, where the possible (non-C-marked) processes are found in the upper part of figure 1.6 inside the perforated box, the final step in the market-based system delimitation is to look at the expected long-term marginal production costs of these technologies/processes (the figures in the boxes). With adequate respect for non-monetarised aspects (flexibility, quality, knowledge), the technology/process with the lowest expected long-term marginal production costs (marked with an arrow) can now be pointed out as the one that will be affected by the change studied.

The outlined procedure is explained in more detail and illustrated with numerous examples in chapter 4.

1.6 Handling co-production

When a process is related to more than one product, how should its exchanges be partitioned and distributed over the multiple products? This has been one of the most controversial issues in the development of the methodology for LCA, as it may significantly influence or even determine the result of the assessments.

The ISO standards on life cycle assessments requires a step-wise procedure to be applied. Besides the obvious solution of subdividing the unit process into separate processes each with only one product, whenever this is possible, the ISO procedure (ISO 14041, clause 6.5.3) consist of three consecutive steps:

• First, when possible, the system should be expanded “to include the additional functions related to the co-products”,
• Secondly, if the above is not possible, “the inputs and outputs of the system should be partitioned between its different products or functions in a way which reflects the underlying physical relationships between them; i.e. they shall reflect the way in which the inputs and outputs are changed by quantitative changes in the products or functions delivered by the system”. Clearly, this is a description of causal relationships, implying that the co-products can be independently varied (i.e. a situation of combined production).
• Finally, “where 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 which reflects other relationships between them. For example, input and output data might be allocated between co-products in proportion to the economic value of the products.” Although not stated explicitly, it can be seen from the parallel wording to the second step that the relationships referred to here should also be causal in nature, which is further emphasised by the only example provided, namely that of economic value of the products, which can be seen as the ultimate cause for the existence of the process. Economic value is so far the only causal relationship that has been found to fit this last step of the ISO procedure.

The two first steps of the ISO procedure are only relevant for consequential studies, since they rely on an analysis of relative changes in the output of the co-products and an adjustment of the systems to yield the same output (see also figure 1.6). This means that for attributional life cycle assessments, where such system adjustments are not possible, co-product allocation by economic relationships is the only option left.

In consequential, comparative studies where a co-product does not appear in similar quantity in all systems under study, it is necessary to expand the studied systems, so that they all yield comparable product outputs. The processes to include when making such system expansions must be those processes actually affected by an increase or decrease in output of the byproduct from the systems under study (see figure 1.6).



Click on the picture to see the html-version of: Figure 1.6

Explanations for figure 1.6: The two original systems to the left are producing product A either without by-products (system 1) or with the by-product B. System expansion (illustrated in the systems to the left) is performed with the following rationale: If system 2 substitutes system 1, more B will be produced for the same quantity of A. This additional amount of B will substitute another existing production of B, which must then be added to system 1 to take this effect into account. Here, the difficult task is to identify which existing production of B will be substituted. If system 2 is substituted by system 1, less B will be produced, thus requiring a new substitute production to be added to system 1. Here, the difficult task is to identify which production of B will be the substitute.

Thus, to identify the processes for a system expansion, one may apply the procedure mentioned in section 1.5 for identifying the processes and technologies actually affected by a product substitution. In chapter 5 it is demonstrated that when applying this formal procedure, system expansion is always possible, i.e. it is always possible to identify those processes that will be affected by a shift between the studied systems. Obviously, the identification can be made with more or less precision, but even an uncertain identification of the affected processes gives a more useful result than an arbitrary allocation according to e.g. economic relationships between the co-products.

From the observation that system expansion is always possible for consequential studies, and never for attributional studies (leaving only the option of economic allocation for such studies), we obtain a much simpler description of the procedure for co-product handling than the description in ISO 14041, although leading to the same result as when following the ISO procedure.

Also other suggestions for allocation procedures, such as the recycling allocation procedure using material grades (Wenzel 1998, Werner & Richter 2000) and the so-called 50/50 procedure for recycling allocation (Ekvall 1994), can be shown to be simple procedures for system expansion relevant in situations of limited information.

1.7 Forecasting processes

Obviously, forecasting is only relevant for prospective life cycle assessments, where the description of the product systems should reflect the relevant time horizon. It is relevant to forecast:

• the future market conditions determining which future product substitutions will take place,
• the geographical and technological conditions of the future processes, and
• the future environmental exchanges of these processes.

As illustrated in figure 1.7, short and medium term (1-5 years) forecasts for specific product systems may be based on simple extrapolation of trends and historical data. For long-term (5-25 years) forecasts, and forecasts for decisions on less specific systems (e.g. the general disposal system of society), it becomes increasingly relevant to use modelling methods, such as trend impact analysis, which adjusts the extrapolations with the expected impact of mechanisms analogous to those determining past events. For generic studies, aimed at influencing many stakeholders (e.g. ecolabelling), it may be relevant to use participatory methods incorporating the insight and opinions of experts and stakeholders. Scenario methods, incorporating several parallel forecasts, are most relevant for systems used in long-term, strategic studies for both societal decisions and product development. The product development process may also benefit from the systematic creativity in exploratory methods, which combine analytic techniques dividing a broad topic or development into increasingly smaller subtopics or consequences, and imaginative techniques aimed at filling all gaps in the analytical structure. For long-term, strategic applications, involving relatively well-defined products from enterprises where the decision maker is expected to have a large degree of control over the future and the different stakeholders involved, it may be relevant to apply normative forecasting, which investigates how we want the future to be and how to obtain this goal.



Click on the picture to see the html-version of: Figure 1.7

____________________________________________________________
⁵ An early version of this introduction was presented to the 3rd International Conference on Ecobalance, Tsukuba 1998.11.25-27 (see Weidema 1998b).
⁶ Also known as status-quo or descriptive LCAs as opposed to the consequential LCAs, which are also known as change-oriented, effect-oriented or comparative (Ekvall 1999). In principle, attributional LCAs may also be performed in an estimated future situation, and consequential LCAs may describe the consequences of a historical decision. We therefore generally use the terms attributional and consequential rather than terms that signal a temporal context.
⁷Although Ekvall (2000) and Ekvall et al. (2001a, b) argue that choices could also be based on other premises than environmental consequences, in which case an attributional LCA based on these premises could be relevant (see also the further text in this section).
⁸ The issue of product declarations is dealt with in more detail in section 4.7, since the potential ambiguity in the purpose of this application make it useful as a touchstone for methodological debates.
⁹ It is worth noticing that in the following methodological explanations, we have not relied on these terms but only used them in brackets to show the places where these terms can be used. Our point in doing this is to show that, even with our more precise definition, the terms are not necessary, and since they are often used without a precise definition, they may be more misleading than guiding. We therefore suggest that these terms should not be used in general for systems descriptions.
¹⁰ In earlier presentations of the procedure to identify the processes or technologies affected by a substitution (e.g. Weidema et al. 1999), the term “marginal” was used extensively to signify small changes and the processes they affect. In this report, as well as in the guidelines, we now generally avoid the term, as it is in everyday-language used in many different meanings and may therefore give rise to confusion. We suggest to use it only to distinguish between small (marginal) substitutions, where an increase and a decrease will affect the same process, and large substitutions where this may not be the case.
¹¹An early version of this section was published in Guinée (1999, pp. 33-46).

2 Product substitution

We define a product substitution as a replacement of one product or group of products with another product or group of products, fulfilling the same needs of the customer. We define a product also in terms of its production process, which implies that a product substitution will always imply one or more process substitutions (understood as process changes or complete replacements), and that a process substitution can also be seen as a product substitution, even when the product itself is unchanged (e.g. in terms of its physical properties).

Product substitutions may occur anywhere in the life cycle, from raw material substitutions, over substitutions in the production and use stages, to substitutions between alternative waste handling options. In a consequential, comparative life cycle assessment, the object of study is the environmental impacts of a potential product substitution. This product substitution, as delimited by the functional unit, implies a change in demand as the customer replace one product in favour of another: More is bought of the one product, less of the others. This change in demand is transferred all the way backwards through the life cycle stages of the products involved in the substitution (and sometimes also forward, if the substituted products are not completely identical). At the other stages of the life cycle, further substitutions occur, as the suppliers scale their production up or down according to the change in demand.

Thus, product substitution is a core concept to consequential, comparative life cycle assessment. In spite of this, a proper methodology has been lacking for including the available knowledge about product substitution into life cycle assessments. This implies that life cycle assessments have often based their functional unit and system delimitation on intuitive or arbitrary choices, rather than on analytical grounds. This arbitrariness is unnecessary, since knowledge about product substitution is available, although requiring information from sources not traditionally used for life cycle assessments.

The objective of this chapter is to describe the general aspects of product substitution, including the issue of data availability (section 2.5), as well as those procedural steps that are common to the more specific elements of the life cycle assessment method, covered by the procedures presented in the following chapters of this report.

Knowledge on product substitution is applied in the following elements of life cycle assessment:

• When defining the functional unit and which alternative products can be or should be compared (chapter 3),
• When identifying the individual processes to be included in the system under study (chapter 4),
• When identifying the processes to be included in a system expansion to accommodate differences in the functions provided by the compared systems (chapter 5),
• When identifying the processes affected on future markets (chapter 6).

The following sections are structured according to the necessary conditions for a product substitution to take place, namely that:

• the products are substitutable, i.e. that the products have the obligatory properties ("must have" properties) required by the customer in the market segment in question (section 2.1),
  
• the products are available to the customer, i.e. that their supply is not constrained by market failures, declining markets, regulations, or shortages in supply of raw materials or other necessary production factors (section 2.2),
  
• a decision is made so that the potential product substitution is actually realised (section 2.3 and 2.4).

This division is in accordance with Sheth’s theory on buying behaviour that distinguish three main elements in the buying process: product requirements, supplier accessibility and customers ideal and actual choice (Sheth 1973, 1981).

2.1 Product properties and market segments

A product substitution is ultimately a decision of the customer. For a product to be considered relevant for a potential product substitution, the customer must see it as fulfilling the same need. This can be expressed in terms of the obligatory properties of the product. What is regarded as obligatory product properties change across market segments, and may thus be identified by analysing the requirements on the market in which the product is to be sold. However, in life cycle assessment, it is not uncommon to first describe the product in terms of it properties, and then to identify and describe the market on which it is to be sold. Thus, it is a bit of a “hen and the egg” situation, where the information on obligatory product properties and market segmentation is mutually dependent.

Product properties may be divided in three groups depending on their importance:

• Obligatory properties that the product must have in order to be at all considered as a relevant alternative. Example: A beverage container must not leak.
• Positioning properties that are considered nice to have by the customer and which may therefore position the product more favourably with the customer, relative to other products with the same obligatory properties. Example: A beverage container may be more or less easy to handle.
• Market-irrelevant properties that do not play a role for the customer’s preferences. Example: A (refillable) beverage container may be more or less easy to clean.

The obligatory properties determine substitutability and are related to market segmentation. Positioning properties may influence the extent to which a potential substitution is actually realised (see section 2.3) and may - together with the market-irrelevant properties - determine the amount of substituted product or the interaction with other product systems. For example, the ease of handling and cleaning a beverage container (positioning and non-market relevant properties, respectively) can influence the amount of car-driving on behalf of the consumer and the type and amount of cleaning agent, respectively.

The same product property may be placed in different groups on different markets (see below).

For a product substitution to be possible, the obligatory properties must be present. Only when these demands are met, the positioning properties can influence the willingness of the customer to switch from one product to another.

Product properties may be related to:

• Functionality, related to the main function of the product
• Technical quality, such as stability, durability, ease of maintenance
• Additional services rendered during use and disposal
• Aesthetics, such as appearance and design
• Image(of the product or the producer)
• Costs related to purchase, use and disposal
• Specific environmental properties

Functionality, aesthetics, and image characterise the primary services provided to the user.

Technical quality and additional services ensure the primary services during the expected duration of these.

Environmental properties may be included among the properties included in the functional unit. However, since the very purpose of a life cycle assessment is to study the environmental impacts of the products, it is not meaningful to state in advance that the studied products should have such general properties as ”environment-friendly” or ”non-toxic.” If environmental properties are included as obligatory, they must be expressed as specific properties, like ”the barley must be from ecological farms”, so that it is possible to judge - prior to the life cycle study - whether a product has the required property.

Of the above-mentioned properties, price is the only one that can be put into well-defined terms. Technical quality and functionality can be described a little less well defined, but still quantitatively. Other properties, such as aesthetics and image, cannot be measured directly, but must be described qualitatively. Some of these properties can seem very irrational, since they are not present in the product, but in the buyer’s perception of it. These properties can be greatly influenced by commercial activities of the supplier.

Markets are typically differentiated

• geographically,
• temporally, and
• in customer segments,

which each have their own uniform set of preferences and demands for product properties.

The geographical segmentation of markets may be determined by differences in:

• natural geography (climate, landscape, transport distances etc.),
• regulation or administration (regulation of competition and market transparency, legislative product requirements, product standards, taxes, subsidies),
• consumer culture.

Temporal segmentation of markets is common for service products (e.g. peak hours and night hours in electricity consumption, rush hours in traffic and telecommunication, seasons in the tourist industry). For physical goods, markets are generally only segmented temporally when adequate supply or storage capacity is missing, either due to the nature of the product (e.g. food products), or due to immature or unstable markets, as has been seen for some recycled materials.

This temporal segmentation should be distinguished from the fact that markets generally develop in time, e.g. governed by developments in fashion and technology, and that both geographical and temporal segmentation and customer segmentation therefore may change over time. In general, there is a tendency for positioning properties to become obligatory with time and for markets to become more transparent and geographically homogenous, but at the same time more segmented with regard to quality requirements.

Each geographical market is typically divided into a number of customer segments. Customer segments are generally defined in terms of clearly distinct function-based requirements, i.e. based on the needs fulfilled by the products rather than based on the physical products themselves. Very similar products may serve different needs and hence serve different markets. And very different products may serve the same need, thus being in competition on the same market. Differences in customer requirements may be based on differences in the purchase situation, the use situation, customer scale, age, sex, education, status, “culture”, attitudes etc.

To have a practical relevance, market segments must be (Lancaster & Massingham 1998):
of a size that can provide adequate revenue to support a separate product line. clearly distinct and with a minimum of overlap, so that all products targeted for a segment are considered substitutable by the customers of this segment, while there should be low probability that a product targeted for another segment would be substitutable, implying that product substitution from segment to segment can be neglected. For example, the market for office chairs is divided into at least three well distinguished customer segments exit, based on three different working situations: The labourer‘s chair, intended for the labourer who is sitting on the chair at intervals only and not for many hours at the time, the computer workstation chair, intended for the worker who is primarily sitting, and who is working behind a visual display unit at least two hours a day, and the manager‘s chair, intended for the design-oriented person, not working much on computer or desk, but rather reading, talking on the telephone and the like. The latter chair could typically be for the employer or senior employee, to whom design, aesthetics, and image/representativity to customers are important issues. There is only very little overlap between these groups of customers. The probability that a chair targeted for one segment should sell to a customer in one of the other segments is small, so that the product substitutability from segment to segment can be neglected. This implies that life cycle studies of office chairs should consider each of the market segments separately and not allow for comparisons between them.

2.2 Product availability and constraints in supply

Even when products have the same obligatory properties, they can only take part in a product substitution if they are available to the customer, i.e. that supply is not constrained.

There can be many reasons that a potentially substitutable product is not available to the consumer, notably market failures, declining markets, regulation, and shortages in supply of raw materials or other necessary production factors.

In a market with only one supplier of the specific product (a monopoly), product substitution is per definition not possible. However, few markets are monopolies. Even the so-called natural monopolies such as the railroads, telephone and electricity markets, which were long divided into regional monopolies, are now being opened up to competition. Still, patents and product standards may limit market entry of new suppliers, and transaction costs may be prohibitive for some potential substitutions to take place in practise.

In a declining market, the penetration of modern technology is constrained, since new capacity is not being installed, limiting competition to those suppliers already present.

Regulatory constraints typically take the form of minimum or maximum quotas on the process (like the Danish minimum quotas on the use of biofuels for heat and electricity generation) or any of its exchanges, e.g. product quotas (like the EU milk quotas) or emission quotas (like the Danish SO₂ and NO_x quotas for electricity generation, which limits the use of coal based technology). The forced phasing out of specific polluting technologies may also render these unavailable to substitution as a result of changes in demand. Taxes and subsidies may also constitute virtual constraints on production. An example is the negligible import of cereal grains to the EU, because of a very high import tax. Similarly, the farmer’s choice of crops is strongly dependent upon the level of subsidies given for different crops, virtually imposing a constraint on crops less subsidised.

The necessary production factors, notably raw materials, may not be locally available or may only be available in limited quantity (for example, the availability of fresh, untreated drinking water may be limited in areas with limited rainfall, water for hydropower likewise, and on an expanding market for a material, the availability of recycled material will be constrained). For products that do not store easily and products and semi-manufactured materials with a low price to weight ratio (such as biomass for energy and paper pulp), transport distances and infrastructure can impose a constraint on products and materials not produced locally. Waste treatment capacity may be a constraint on processes with specific hazardous wastes.

For multi-product processes, supply of a co-product may be constrained if it does not have a value that can sustain the production alone. In general, this will be the case if the studied product has a low value compared to the other co-products, so that the studied co-product cannot in itself provide an economic revenue that is adequate reason for changing the production volume (like animal manure versus milk and meat, or rape seed cakes versus rape seed oil), or if the market trend for the studied co-product is low compared to the market trend for the other co-products. See also section 5.4.

2.3 Realising substitution

Even when products have the same obligatory properties, and an unconstrained supply, substitution is only realised when active decisions are made by the customer.

In the first part of the life cycle (e.g. in relation to raw material substitution), price tends to play a larger role in purchase decisions and product quality is often less complex, more easy to define precisely, more dominated by technical aspects, and more stable over time than later in the life cycle (consumer products) where complexity increases, preferences change more quickly, and qualitative aspects and irrational behaviour may have larger influence.

It is possible to further subdivide market segments into market niches. A market niche is a further sub-category of a market segment, where a part of the customers consider only niche products substitutable, although the majority of the customers allow substitution between products from the niche and other products in the segment. Thus, the difference between a segment and a niche is that between segments substitution is negligible, while a large part of the customers in a segment will allow substitution between niche products. Niche products are aimed at a smaller group of consumers within a segment, for whom specific product properties are obligatory, while the same properties were only positioning properties in the broader market segment.

Office chair example: Market niches
The substitutability between chairs depends on customer preferences. And these preferences vary from customer to customer. From the 18 years old sporty person, who maybe just wants a chair and gives no thoughts what-so-ever to ergonomic properties, to the older secretary with back troubles. And from the small 10-employee private company to the public institution buying through the public purchase service. Especially within the market segment for computer workstation chairs (see Weidema et al. 2003a), there might be well-distinguished niches between which, the product substitutability is only limited. Examples are:

• The niche of occupational therapist prescribed purchase, within which high level ergonomic properties become obligatory, such as synchronic and weight adjusted movements of seat and back rest,
• In Denmark, public institutions of the state, counties, and municipalities can buy through the National Procurement institution (SKI - Statens og Kommunernes Indkøbsservice) that has bulk sales agreements with suppliers. SKI may then in turn specify requirements to be fulfilled by suppliers in order to deliver through SKI, which may include both functional and environmental requirements. For office chairs, these requirements include e.g. ergonomic properties, tests for stability, durability, and strength. Because public purchase is a very large share of the market for office chairs, SKI plays an important role, also in raising the general level of customer requirements.

Market segments and niches are typically identified by dividing the market according to a number of customer characteristics, such as customer scale, age, sex, education, status, “culture”, attitudes etc. in such a way that demand for product properties is homogenous within segments/niches and heterogeneous between each segment/niche (Lancaster & Massingham 1998). There is also some evidence that market segments and niches may better be modelled by purchase or use context, rather than by customer characteristics, thus allowing the same customer to have different preferences in different situations, and different customers to behave similarly in the same situation (Moss & Edmonds 1997). Edmonds et al. (1997) provide a model algorithm that enables quantitative market segmentation also in situations where domain experts lack confidence in their own judgements or where their initial segmentation is found not to be in accordance with available sales figures. Bech-Larsen & Skytte (1998) provide an example of using conjoint analysis for segmentation of the vegetable oil market into four niches: one with strong preferences for neutral colours and a low content of transfatty acids, one with strong focus on supplier characteristics in terms of quick delivery and ISO certification, one with a strong preference for a high oxidative resistance and antipathy for rape seed oil, and one with strong preferences for neutral taste and a high nutritional value.

2.4 Supply elasticity as a measure of actual substitution

The supply elasticity is a formal measure of the substitution realised (i.e. the change in supply) as a result of a change in an influencing factor, e.g. the demand. If a change in demand leads to a similar change in supply, that supply is said to be fully elastic. If a change in demand does not lead to any change in the supply, that supply is said to be fully inelastic.

On competitive, unconstrained markets (i.e. where there are no market imperfections and no absolute shortages or obligations with respect to supply of production factors, so that production factors are fully elastic in the long term), individual suppliers are price-takers (which means that they cannot influence the market price) and the long-term market prices will be determined by the long-term marginal production costs (which implies that long-term market prices, as opposed to short-term prices, are not affected by demand). In this situation, the long-term supply will be fully elastic. In most life cycle inventory models, this is applied as a default assumption: For each process in the life cycle, the demand for 1 unit of product is assumed to lead to the supply of 1 unit of product, and other customers/applications of the product are assumed not to be affected.

Individual suppliers or technologies may be constrained in the long and/or short term and therefore have an inelastic supply. In this situation, the demand will shift to an alternative supplier/technology that is not constrained.

If all suppliers to a specific market segment are constrained, or if one or more production factors are not fully elastic, a change in demand will lead to a change in market price and a consequent adjustment in demand (i.e. a behavioural change). This adjustment will be accommodated by the customer(s)/application(s) most sensitive to changes in price, measured in terms of their demand elasticity (i.e. their relative change in demand in response to a change in price).

A special class of constraints are those related to co-production. If the co-producing process is otherwise unconstrained, it is reasonable to apply the default assumption above, that the long-term supply elasticity is fully elastic, also for a determining co-product (see chapter 4 for a definition of determining co-products and a detailed description of our procedure for handling co-products). Thus, the demand for 1 unit of a determining co-product is assumed to lead to the supply of 1 unit of the determining co-product along with the corresponding amount of the dependent co-product(s). Depending on the market situation for each dependent co-product, this additional supply of dependent co-products will go to waste (when the dependent co-product is already only partially utilised), lead to a displacement of the most sensitive alternative supply (when the dependent co-product is already utilised fully and alternative suppliers are not constrained), or lead to an increase in consumption (when the dependent co-product is already utilised fully and all alternative suppliers are constrained). In the situation where displacement occurs, the default assumption implies that the suppliers are price-takers and the co-producing process cannot influence the market price. The market price for the dependent co-product will therefore be determined by the long-term marginal production costs of the displaced supply.

In this way, our treatment of co-production is simply a consistent application of the default assumptions generally applied in life cycle inventory modelling. It therefore appears inconsistent to dismiss our substitution procedure as “fairly unrealistic”, “rather unrealistic”, and to view it “not as part of inventory modeling, but as a type of allocation” (Guinée et al. 2001, part 3, page 125-6, 129). In contrast, Guinée et al. (2001) refrain from modelling the effects of co-production and recommend instead an allocation procedure based on the co-products’ shares of the total revenue. This allocation procedure is equivalent to assuming (see also section 5.10):

• that for any requested co-product a co-producing process will react to a change in demand with an increase in production volume in proportion to the co-product’s share in the total revenue, implying that the remaining part of the demand will be covered by an alternative supply and/or a reduction in consumption elsewhere,
• that the additional supply of other co-products, caused by the increased production volume of the co-producing process, will always be utilized fully and lead to an equivalent increase in consumption (since there is no additional increase in waste handling from the co-producing process, and no displacement of alternative supply) implying that the demand elasticities for these co-products are infinite (even for “near-to-waste” co-products that are partly disposed of as waste),
• that the environmental impacts of the above alternative supply and/or changes in consumption are insignificant (since the system is not expanded to include this alternative supply and/or changed consumption and related processes), and
• that there will be no displacement of alternative supply (i.e. that supply elasticity is 0, even when the supply is not constrained and the same supply elsewhere in the same study may be modelled to respond to a change in demand with the default fully elastic supply).

It is difficult to see how these assumptions can be regarded as more realistic than extending the default assumptions used in the remaining inventory modeling to cover also the situation of co-production.

In fairness, it shall be noted that Guinée et al. (2001), immediately having passed the above controversial and somewhat harsh judgment on our procedure, proceed to recommend the application of our procedure as a sensitivity analysis, “to gain an indication of the possible effects of substitution,” although this may be stretching the concept of sensitivity analysis beyond its original meaning. Rather, the intention is to suggest the inclusion of our procedure as a separate scenario, as an extension “for improving the quality of detailed LCA in these respects where shortcomings are most obvious. A key example is the absence of economic mechanisms in the model, an unfortunate feature in cases where there are extreme inelasticities of supply and demand” (Guinée et al. 2001, part 3, p.59, last bullet). Also in their research recommendations, they suggest: “By incorporating certain economic mechanisms in the inventory model, particularly in cases involving extremely high or low elasticities, inventory modeling might be made more realistic and some of the principal defects of LCA redressed” (op.cit., p. 133). Furthermore, it appears that their judgement of our method has been based on an insufficient understanding: “the … method of Weidema, still difficult to understand, …” (op. cit., p. 128), which is also confirmed by their extensive list of research recommendations (op.cit., p. 133-4).

It should be noted that applying the revenue-based allocation procedure in consequential studies, as recommended by Guinée et al (2001), leads to further inconsistencies when seen in combination with the general recommendation of Guinée et al. (2001) to identify the affected processes in terms of market averages. For example, if we assume that a market is supplied with 10% from a single-product process and 90% from a co-producing process, but this product only contributes with 10% of the revenue, then only a tenth of the co-producing process will be included in the system, implying either an unrealistic decrease in demand elsewhere or that the remaining 90% will be supplied from the single-product process, which is however not consistent with our knowledge that it supplies only 10%.

Further, applying the revenue-based allocation procedure in consequential studies is also inconsistent with defining the functional unit in terms of a single function from a multi-functional process (e.g. the isolated cleaning function of an anti-dandruff-shampoo). When the functional unit comes from a multi-functional process (the hair washing providing joint cleaning and anti-dandruff functions), the demand for the functional unit should – according to the allocation procedure – only affect a part of the analysed product systems equivalent to the share of the functional unit (the price that can be attributed to the isolated cleaning function) out of the total revenue. Nevertheless, when analysing this isolated function, Guinée et al. (2001, part 3, page 78) just mention this allocation procedure as one option, suggesting that the other functions may equally justifiable be either neglected or dealt with through system expansion (adding the anti-dandruff function to the functional unit).

As can be seen from this analysis, all three above elements of life cycle inventory modelling (the method for defining the functional unit, the method for identifying the processes to be included in the system, and the method for dealing with co-production) are interwoven and relate to the same issue, namely that of product substitution as outlined in this chapter. Only when applying the same fundamental method and assumptions for all three elements, as in the following three chapters, a consistent result will be obtained.

2.5 Availability of market data

For the study of product substitutions, and thus for consequential life cycle assessments, the availability of market data is essential. We have therefore investigated the current availability of market data, and have come to the conclusion that availability is a minor problem compared to the availability of technical data on environmental exchanges (the more well-known data availability problem in LCA), although access is still not straightforward.

On the basis of the description of product substitution in the previous sections, five types of market data can be distinguished, the availability of which are discussed separately (illustrated by milk as a specific product) below. Further examples of specific data are provided in chapters 3 and 4.

1. Obligatory and positioning product properties in different market segments and geographical markets. Information can be obtained from the marketing departments of the enterprises supplying products to the market segment. If such direct information is not available, the same information may be obtained from retailers, industrial organisations, industrial research institutions and industry consultants, regulating authorities and standardisation bodies (issues regulated in national and international legislation and standards will typically be obligatory properties), marketing and consumer research institutions, or trade statistics (the latter especially to document geographical market boundaries). Examples of publicly available information are the analyses of industrial sectors or “resource areas” provided by the Danish Agency for Industry and Trade (Erhvervsfremmestyrelsen 1993a,b,1994a,b, 2001). Weidema et al. (2003a) provide an example of identifying market segmentation and product properties of office chairs, based on a small survey of the Danish market.
   Milk example: Statistical publications have information on market share of various sales channels. Published nutritional surveys of food consumption per population, sex and age group (based on questionnaires) can be used to assess whether and to what extent specific products are consumed in sex- and age groups. Besides such public sources, marketing departments of dairies have a good understanding of the market segments, which they use for planning the marketing their products. The obligatory product properties of milk in each segment: temperature, age after milking, keeping ability, packaging properties etc., are also well known by the marketing departments of the dairies.
2. Data on constraints in production and supply. Regulatory and political constraints, typically in the form of minimum or maximum quotas, are obviously well known and public (examples: Political decisions not to build any more hydropower or nuclear power plants in Europe, Danish minimum quotas on the use of biofuels for heat and electricity generation, EU milk quotas, Danish SO₂ and NO_x quotas for electricity generation, which limits the use of coal based technology). Constraints in the availability of raw materials, waste treatment capacity, or other production factors are typically well known in the industry and not regarded as confidential. Constraints due to co-production can be determined from their share in the economic revenue combined with their relative market trends, cf. the procedure outlined in section 5.4. In case of missing information on constraints, it should be assumed that there are none. Unjustified exclusion of suppliers is thereby avoided.
   Milk example: Data on milk quotas, subsidies and restitution in the EU are publicly available. The same is true for agricultural fodder crops. Often, it is rather obvious what supplies are constrained, e.g. fodder by-products from the food industry, which depends on the demand for food products, not on changes in demand for fodder.
3. Data on market trends. This information is typically a combination of statistical data showing the past and current development of the market and different forecasts and scenarios. Trade and production statistics are typically publicly available, either from the national statistics or from product specific industrial organisations. Sector forecasts are typically available from national and supranational authorities, while more product specific forecasts are available from industrial organisations.
   Milk example: Statistical data are published yearly, e.g. sector profiles of the dairy sector with information on spending on milk and milk-products per capita, market share of milk and milk-products and the development in total use of milk and milk-products in kilograms per capita. Sector studies are available in which buying behaviour and scenarios for the future are described. These studies are mostly qualitative.
4. Data on the parameters that influence decisions on realised substitution, e.g. prices of different technologies and the effect of information on buying behaviour and investment decisions. Data on production costs for individual plants, countries, or technologies are obtained from the industry in question, from industry consultants, or from research organisations. Examples are Doms (1993) for data on the U.S. manufacturing energy market, World Steel Dynamics (2000) for data on steel (process-by-process costs and world cost curves replicating key process costs for 284 steel plants in 49 countries), Dernecon (2000) for newsprint, and SRIC (1999) for chemicals. If data cannot be obtained, it may be assumed that modern technology is the most competitive and the oldest applied technology is the least competitive. With respect to geographical location, it can be assumed that competitiveness is determined by the cost structure of the most important production factor (labour costs for labour intensive products, else energy and raw material costs). When comparing labour costs, local differences in productivity and labour skills should be taken into account.
   Milk example: Data on the costs of different technologies may be obtained from production engineers and suppliers of machinery. Public sources are not common. Prices of different fodder crops are published and can be used to calculate the changes in fodder composition as a result of changes in production. The most difficult decisions to model are those of the farmers, e.g. how the choice of crops are made. Models can be based on information on the marginal revenue of different crops and the relation between the costs of different inputs (fertiliser, pest control), the influence on the yield, and the price of the agricultural products.
5. Data on the scale of change that may influence what technologies and processes to include. These data regard boundary conditions of some of the above-described data on market sizes and constraints, market trends, and production costs. Thus, the sources and availability of these data are similar to the above.

A specific problem in data collection for consequential LCAs is that the product substitutions often involve processes that do not belong to the immediate supply chain. This means that data will be required from companies that may not see the immediate relevance of their participation, thus affecting their willingness to supply data. However, practical experience rather suggests that willingness to participate is more a question of the general company culture towards professional secrecy than a question of closeness of business relations.

Just as for technical data on environmental exchanges, market data can be provided in terms of best estimates, best-case, and worst-case data, the latter being suited to provoke data providers to supply data of improved quality.

3 Market-based definition of the functional unit

The functional unit is a central term in LCA, as it signifies the common basis for a product substitution or comparison. In practice, the functional unit is only one specific aspect of the larger task to:

• determine the object of study thereby making a first delimitation of the product systems to be studied. Example: Artificial outdoor-lighting with daylight-spectrum for existing European fittings.
• provide a quantified reference unit for all other data in the study (this is the functional unit). Example: Lighting 10 square metres with 3000 lux for 50000 hours with daylight spectrum at 5600 K. The functional unit describes and quantifies those properties of the product, which must be present for the studied substitution to take place. These obligatory properties (the functionality, appearance, stability, durability, ease of maintenance etc.) are in turn determined by the requirements on the market on which the product is to be sold, as already outlined in section 2.1.
• determine the reference flows that provide equivalence between the alternative product systems in a comparative study. Example: 15 daylight bulbs of 10000 lumen with a lifetime of 10000 hours compared to 6 daylight bulbs of 10000 lumen with a lifetime of 25000 hours. The reference flows translate the abstract functional unit into specific product flows for each of the compared systems, so that product alternatives are compared on an equivalent basis, reflecting the actual consequences of the potential product substitution. The reference flows are the starting points for building the necessary models of the product systems.

For a systematic treatment of these elements of a product life cycle study, we have developed a 5-step procedure:

Step 1: Describe the product by its properties.
Step 2: Determine the relevant market segment. Step 3: Determine the relevant product alternatives.
Step 4: Define and quantify the functional unit, in terms of the obligatory product properties required by the relevant market segment.
Step 5: Determine the reference flow for each of the product systems.

Table 3.1 gives an overview of the relations between the five steps in the procedure and the above three bullets, which reflect the purposes of the procedure. Figure 3.1 gives a graphical summary of the information flow between the 5 steps.



Click on the picture to see the html-version of: Table 3.1



Click on the picture to see the html-version of: Figure 3.1

[Information til layout‘er: Ovenstående figur er et udkast. Den endelige figur skal tages fra A5-vejledningen (redigerings-vejledning til figuren findes i udkastet til A5-vejledningen).]

Although the procedure is described in five consecutive steps, it should be noted that it may often be relevant to perform the procedure in an iterative or concurrent way: The product properties described in step 1 may be determined at the same time as, or even from, information on the market segmentation (step 2). The product or the product alternatives (step 3) may be given in advance, and thus contribute to the definition of the relevant product properties (step 1). And the functional unit can be defined (step 4) without having first determined the relevant product or the product alternatives (step 3).

The two first steps of the procedure, description of product properties and determination of market segments, are closely related, as already described in section 2.1.

In developing environmentally more preferable products, it is important to understand the relationship between the individual properties and the environmental impact. If the environmental impacts are particularly linked to specific properties, it is especially important to consider whether these ”environmentally costly” properties are obligatory properties that the product “must have” or only positioning properties that it is “nice to have”, and whether it is possible to influence the trade-off made by the customer between the properties in question and the environmental properties of the product, e.g. by environmental information to the customer.

Here, it may also be relevant to consider the concept of market niches (see section 2.3), since the distribution of product properties over the categories obligatory, positioning, and market-irrelevant, may be different for a product aimed at a specific niche than for a product aimed at the general market segment. As an example, an analysis of the US market for manufacturing energy (Doms 1993) show how some operations specifically require gaseous fuels that have the capacity of reaching high, precisely controlled flame temperatures, while other operations have less specific requirements and therefore may substitute between a diversity of sources. In targeting products for different niches, suppliers may utilise such differences in obligatory properties between niches.

Environmentally conscious consumers may give rise to new market niches. This challenge may be met with three strategies for product changes:
a) reducing environmental impact by reducing functionality,
b) reducing environmental impact while maintaining or moderately improving functionality,
c) maintaining or reducing the environmental impact per unit of function while improving functionality.
This is illustrated in figure 3.2.



Click on the picture to see the html-version of: Figure 3.2

The first strategy will appeal only to a narrow niche of very environmentally concerned customers, with high requirements for environmental properties and low requirements for functional properties. In this niche, the type and number of obligatory properties are reduced compared to the general requirements of the whole market segment. Thus, the functional unit for products aimed at this niche is different from the functional unit for products aimed at the whole market segment. The functionality that is reduced relate to properties that are not obligatory in this niche and therefore not part of the functional unit. For the “normal” market, reductions in functionality of obligatory properties are not allowed. Here, environmental improvements must be sought that does not compromise functionality. This type of solution may appeal to both the environmentally neutral customer and the environmental conscious customer.

The last strategy, where improvements in functionality are paired with a reduction in environmental impacts per functional unit, may at first sight be seen as an ideal solution implying no trade-offs. However, this strategy may meet some resistance among the more environmentally conscious, since the overall environmental impact may actually increase, if the increase in functionality leads to an increase in the demand for the function, which is a rather common phenomenon also known as the rebound effect. The rebound effect may in fact occur for any type of change, even when the functionality decreases, if the decrease in functionality leads to compensatory actions. Similarly, also the opposite of the rebound effect, a reduction in consumption, may occur as a result of improved functionality (“rather have one piece of high quality than two mediocre”), a parallel to the acceptance of less functionality in the “environmentalist niche.” These secondary effects must be taken into account either when defining the functional unit (by using a broader perspective including the behavioural responses, e.g. rather “average work-related personal transport behaviour during one year” than “30000 person-km”) or when determining the reference flows (there including the additional processes affected), as further explained below.

The identified market segment or niche may further delimit the products that may be involved in a product substitution, thus laying the ground for the further specification in step 3 of the product alternatives of what products shall be included in the study, depending on the goal of the study. For example, an enterprise internal study may be performed for a very specific purpose, which gives a large degree of freedom to define what is regarded as relevant alternatives, while public applications typically aim at influencing a predetermined market and therefore must relate to the products that are (expected to be) available on this market.

What remains in step 4 is mainly the quantification, which should as far as possible relate to the functions of the product rather than to the physical product. For example, rather “seating support for one person working at a computer for one year” than “one computer workstation chair”, rather “freezing capacity of 200 dm³ at -18° C” than “one 200 dm³ refrigerator”, rather “annual lighting of a work area of 10 square metres with 30 lux” than “bulbs providing 30000 lumen for one year”. In this way, it is ensured that all obligatory properties - as well as the duration of the product performance - are addressed.

As a reference unit, the size of the functional unit is - in principle - arbitrary. In general, it does not matter whether the office-chair study is normalised to seating support for 0.28 persons, 1 person, 1000 persons or 1.4 million persons. However, two concerns may be relevant when deciding on the size of the functional unit:

• the scale of the studied product substitution,
• the ease of comparison of the outcome of the study to other known quantities.

The studied product substitution may be small or large. A large substitution is defined as one, which affects the determining parameters for the overall technology development. Thereby, the studied substitution may in itself lead to new technologies being brought into focus. It can be a change so large that it affects the general trend in the market volume, e.g. from decreasing to increasing, whereby a new technology comes into play. It may also be a change so large that it overcomes a constraint which otherwise prevents the use of a specific technology. Further, a change may be so large that it affects the production costs of the involved technologies, e.g. through economies of scale. For such instances, it may be misleading if the functional unit is chosen independently of the actual scale of the studied substitution. When studying substitutions involving the entire market of a major product or process, e.g. studies dealing with the entire waste handling system of a region or studies dealing with legislation or standards for an entire sector, it is relevant to choose a functional unit of the same size as the affected market. In evaluating the size of the affected market, it may be relevant to take into account the existence of market niches that react differently to the studied product alternatives, with the aim of quantifying the importance of these niches. While this may affect the chosen size for the functional unit, it should not affect the nature of the functional unit (i.e. as defined by the obligatory properties common to the entire market segment studied). Only when studying substitutions in a specific niche, the nature of the functional unit will be affected compared to the functional unit for the entire segment.

Most often, however, life cycle studies deal with small substitutions, which do not affect the overall trends in market volumes, nor the constraints on and production costs of the involved technologies. Therefore, the consequences of the substitution can be assumed linearly related to the size of the substitution so that the precise size of the functional unit will have no importance for the interpretation of the results.

For such small substitutions, another concern may be relevant: When presenting the outcome of the study, it should be as easy as possible to compare the outcome to something well-known to the reader. For this reason, the environmental exchanges are typically normalised to the annual exchanges from a region, from an average person living in this region (person-equivalents as in the EDIP-method), or from the average monetary expenditure in this region. To ease this normalisation, and to present the results in an easily comprehensible way, it may be an advantage to set the size of the functional unit equal or close to the annual per capita consumption of the studied product on the studied market segment.

In some instances, two products may be so closely linked that the separation of some of the processes in their life cycle may lead to an increase in uncertainty. If all the analysed product systems provide the same amount of such linked products, this additional uncertainty may be avoided by including both products in the functional unit.

The final fifth step in the procedure is to determine the reference flow for each of the product systems. The reference flow is a quantified amount of the product(s), including product parts, necessary for a specific product system to deliver the performance described by the functional unit. For a composite product, the reference flow will typically be identical to the parts list of the product, multiplied by a factor to scale it to the functional unit. The purpose of the reference flows is to translate the abstract functional unit into specific product flows for each of the compared systems, so that product alternatives are compared on an equivalent basis, reflecting the actual consequences of the potential product substitution.

As noted in section 2.1, it is not just the obligatory product properties that determine the amount of substituted product or the interaction with other product systems. To ensure the equivalence between compared products, it is therefore necessary to analyse systematically all product properties and judge for each one whether it leads to differences in the amount of substituted product or in the interaction with other product systems. If several such additional properties can be identified, it is important to investigate whether one of the properties can be identified as the one determining the difference in performance.

Examples of determining12 properties¹²:
In comparing different alternatives for hand drying, the technical properties of the tissue paper such as mass, absorption-power and tensile strength, may all influence the number of tissue papers used. However, these properties may all turn out to be irrelevant if in practice it is the dispenser design that determines the amount of paper used. Similarly, technical specifications of electrical hand driers, such as the volume of air and its temperature, may be irrelevant for comparing relative performance, if the actual operating time and energy consumption of the devices are fixed by other factors, e.g. built-in timers that give a fixed time per hand-drying event to be multiplied with the effect of the device (kW/minute).