Why not PEF?

The European Union Product Environmental Footprint (PEF) method “to measure and communicate the life cycle environmental performance of products” was updated in December last year and is now widely promoted by the Commission. More and more customers now ask us to have their Life Cycle Assessment (LCA) performed according to PEF. Yet, we still recommend our customers to settle for an ISO 14044 compliant LCA, while we wait for the PEF method to be further improved. This blog-post seeks to explain why.

A major problem with the current PEF method is the lack of consistency of the rules across different product categories. For each product category, there are separate rules, known as PEF Category Rules (PEFCRs), developed by Technical Secretariats that each represent industries that cover at least 51% of the European consumption market for the product category. The result is that we now have as many rules for how to make an LCA as there are product categories. This limits the relevance of the results to comparisons within each product category and creates potential conflicts with similar (but different) rules developed by industries outside Europe, and thereby with the WTO-rules. But in reality, most of the content of the PEFCRs regulate issues that could better be (or are already) addressed as general requirements for all products instead of having to be included in each and every single PEFCR. Examples include requirements on including capital goods and infrastructure, product lifetimes or standardized methods of computing lifetimes, the depreciation method to allocate the burden of capital goods over their service period, requirements for scenarios for use and End-of-Life, requirements on the use of primary data and the age of data, data sources, impact assessment methods and indicators, cut-off criteria, use of best estimates versus worst-case, requirements and assessment methods for data quality, and methods for handling multi-functionality. For each of these issues, the 2021 PEF methodology gives preferential status to the specific rules or the PEFCRs over the generic requirements of the PEF method, thereby furthering inconsistencies and hampering cross-category product comparability. We recommend instead to apply the same rules for all products, thus limiting the content of PECRs to only include issues for which it is meaningful that different product categories can have different requirements that are not more consistently regulated at the generic level of the PEF methodology for all products. This would not only allow consistent cross-category product comparisons, but also imply a considerable streamlining and significant cost-saving for maintaining the PEF system and for the construction of a harmonised and consistent database. The less prominent role of PEFCRs will also allow for better participation of the scientific and civil society, and reduce undue influence from industries within a specific product category.

Another major problem with the current PEF method is the lack of harmonisation with international standards, to ensure that calculations provide reliable and credible information for fair comparisons, preventing that claims based on the PEF method can be misleading and lead to litigation. In our view, the core understanding that all further rules should be based on, is that expressed in ISO 14040:2006, annex A.2:

“applications relate to decisions that aim for environmental improvements, which is also the overall focus of the ISO 14000 series. Therefore, the products and processes studied in an LCA are those affected by the decision that the LCA intends to support.”

An illustrative example of how the 2021 PEF revision is breaking away from this core understanding is the way that the earlier requirement to use the ISO 14044 allocation hierarchy is now immediately negated by an exception in the following sentence: “Specific allocation requirements in other sections of this method always prevail over the ones available in this section”. Other sections then require the use of special rules, e.g., for different agricultural, slaughterhouse, and recycling situations. We recommend to harmonise and improve the realism of the modelling of the multi-functionality of processes by adjusting the PEF ‘circular footprint formula’ as described in Schrijvers et al. (2021) and requiring its consistent use for all situations of multi-functionality and use of recycled materials and energy, while simplifying the description by providing separate requirements for each real-life situation and making clear references to the ISO 14044 allocation hierarchy. Furthermore, we recommend to harmonise and improve the realism of the linking of unit processes into product systems by adding precise requirements and guidance on system expansion and the identification of the upstream system for intermediate product inputs, based on the description provided in ISO 14044 Annex D.2.1.

While ISO 14044 only allows to leave out parts of the life cycle on grounds of insignificance, the PEF methodology document does not have such strict completeness requirements for the activities to be included. Although ‘completeness’ is one of the data quality requirements for PEF compliant data sets, this is in reality only a requirement to include all 16 PEF impact categories, while data gaps in the system model as such are allowed as long as ‘transparently reported’ and ‘validated by the verifier’. However, even when transparently reported and validated, data gaps will constitute a complication for comparability between studies. The 2021 PEF method explicitly recommends excluding capital goods (incl. indirect land use) unless a clear and extensive explanation is provided for why the inclusion of capital goods (including infrastructure) is relevant. Considering that comparable products can come from very different product systems that have very different use of capital equipment, this introduces a significant potential for biased comparisons across different products, as quantified by Font Vivanco (2020). Similarly, unbiased comparisons that involve products from forestry and agriculture cannot be done without including indirect land use as an important source of biomass production capacity, a necessary capital good for forestry and agriculture. Interestingly, the mandatory format guide for EF compliant data sets requires capital goods to be included unless clearly documented and checked by the reviewer. Of these conflicting requirements, the latter is obviously the more relevant. We recommend to follow to the ISO 14044 requirement that parts of the life cycle shall only be excluded on the grounds of insignificance, to remove the allowances for data gaps and exclusion of capital goods, and enforce the (existing) requirement that all non-elementary flows shall be modelled up to the level of elementary flows (flows either drawn from the environment without previous human transformation or released into the environment without subsequent human transformation), and finally to add a requirement to check the system completeness by applying mass balances (‘what goes in must come out’).

The PEF method requires the use of 16 specific impact categories and a single overall score based on weighting factors that are not based on a science-based elicitation procedure. Reporting additional impact assessment results as additional information is prohibited. While this strict enforcement of a specific impact assessment method does increase comparability between products, it would be preferable if this was based on the latest scientific evidence. We therefore recommend to replace the current PEF normalisation and weighting factors with those recommended by the international UNEP-GLAM project, and to allow the results from scientifically based cause-effect models, such as for example Global Temperature Potentials, to be reported under ‘Additional environmental information’.

The 2021 PEF method includes a data quality scheme that requires a specific software to provide an overall data quality score without any empirical basis and has no relation to the resulting additional uncertainty that stems from lacking data quality. Without the ability to convert data quality into uncertainty, the PEF study results lack an important feature for decision support. We recommend instead to apply empirically confirmed average uncertainty values for each data quality score, based on the method of Ciroth et al. (2016), so that uncertainty from low quality data can be added to the basic uncertainty of each flow and thus included in the overall uncertainty propagation when calculating LCA results for each product.

In addition to the above issues, the 2021 PEF method includes a large number of unnecessary and/or burdensome requirements that effectively hampers its practical application. In fact, we have yet to see an LCA study that have fulfilled all the PEF requirements. What most practitioners do today, is to say: “We have followed the PEF standard as far as possible”, which is not really a way to ensure consistency and comparability. To make the PEF method operational in practice and reduce unnecessary costs, we therefore recommend to remove or reword these PEF requirements:

  • The requirement to subdivide the product life cycle into ‘life cycle stages’ and to provide extensive decomposition analyses for each overall weighted result of a PEF study. The intention can more easily and flexibly be provided by giving public access to the detailed results in an electronic database format.
  • The requirement that all dataset used in a PEF study shall follow the guide for EF compliant data sets. Constructing, modifying, linking, and reviewing datasets in the ILCD data format are tasks for specially trained personnel. The 2021 PEF method describes what to do when an EF compliant data set is not available, including the option to exclude the activity and note it as a data gap. Thereby, the data format requirement becomes a source of incompleteness that could be avoided by a less strict format requirement, without compromising the data quality.
  • The validation and verification requirements for background data: That validation shall include coverage, precision, completeness, representativeness, consistency, reproducibility, sources, uncertainty, plausibility, quality, and accuracy of the LCA-based data, the correct conversion of measurement units, and quality and accuracy of additional environmental, technical, and supporting information, including parameters and datasets used to model the circular footprint formula. That company-specific data are required to be validated through a visit to the production site. That the verifier(s) shall check at least 80% (in number) of the most-relevant processes and for at least 30% (in number) of the remaining processes, whether they are correctly identified and that all related activity data and the datasets used to model these processes shall be validated, including their correct implementation in the software for at least 50% (in number) of the most-relevant processes and for at least 10% (in number) of the remaining processes. A validation report and validation statement are mandatory, and the latter shall always be provided as an annex to the PEF report. A validation report and validation statement shall refer to one specific PEF study only and is valid for a maximum of 3 years. If during these 3 years, the verifier(s) shall evaluate whether the results of one of the impact categories or the aggregated score have worsened by more than 10% or 5%, respectively, in which case the PEF study and report shall be updated. I think the reader will agree that these validation requirements can at best be described as excessive. On the other hand, validation by mass balance, one of the most simple and solid validation procedures for system completeness, is made impossible by the required mixture of market-based substitutions and economic and physical allocation factors. Considering that typically more than 10’000 processes are included in an LCA, it is obvious that the described validation cannot be done manually. If, on the other hand, a single, complete, and consistent database is used for all background data, the entire validation of this database can be done automatically and continuously as new data are entered, so that for a specific LCA, only the additional foreground data needs to be manually validated (Weidema et al. 2013). This would require some rewording of the current verification and validation requirements.

We have raised these issues and made the above recommendations in a recent report to the European Parliament, and will continue to gather support for a corresponding revision of the PEF method to make it reliable, fair, and practicable.