The role of EEIO frameworks in Green Public Procurement monitoring

Motivation

The public sector avails of significant purchasing power when contracting with the private sector for goods, works and services. This provides it with meaningful leverage across a good fraction of economic activity. Significant efforts are underway worldwide to utilize this leverage towards achieving sustainability goals. Yet implementing sustainable procurement at scale remains a task of enormous complexity that requires significant investment: in developing suitable policies, establishing appropriate information flows and methodologies and, not least, embedding these new practices and behaviors across swathes of economic activity.

In this blog post we provide a brief overview of how an important class of economic analysis tools termed EEIO, or environmentally extended input-output frameworks, can be utilized to support the implementation of key public procurement sustainability objectives. We sketch the nature of these tools, how they can be used to provide relevant information, and the pros and cons of such an approach.

The critical role of GPP

In numbers, public procurement may account for anywhere between 12 and 30 percent of a country’s economic product (GDP). Roads, public buildings, railways, public transport and energy are but a few examples of procured goods and services. Public procurement may thus entail significant environmental impact. Ergo, central and local government policies prioritizing forms of environmentally and socially responsible purchasing can over time exert significant influence on markets in the direction of sustainability. Such efforts go collectively under the name Sustainable Public Procurement (SPP). As a subset of such initiatives, the environmentally focused pillar of SPP is termed Green Public Procurement (GPP).

GPP implies that public authorities procure goods, services and works with a reduced environmental impact when compared to baseline goods, services and works with the same primary function that would be procured in the absence of GPP policies. By now a number of countries have introduced and practice some form of GPP. GPP programs vary in the numbers of types of products and services covered and the nature of policies introduced. An important and common focus area is to address anthropogenic climate change through the reduction of Greenhouse Gas emissions (GHG), and the promotion of renewable sources of energy and energy efficiency. The protection of soil, water, biodiversity and material use efficiency (circular economy) are other examples of important environmental sustainability objectives.

For definiteness, we will focus the discussion on GHG emissions but the overall capabilities of EEIO frameworks are in-principle applicable in more contexts. A number of other environmental and social impacts can be approached in a similar fashion.

The need for effective GPP Monitoring

GPP Monitoring is the set of processes of tracking progress in the implementation of GPP programs. It is recognized that such monitoring is an important factor for their success. The wide scope of GPP, which ultimately reflects the diversity of public procurement needs, implies several distinct types of GPP monitoring. Monitoring activities have been conceptually categorized (see e.g., SEAD guide for monitoring and evaluating green public procurement programs, 2013) as follows:

1. monitoring the institutionalization of GPP policies (the pace and quality of the introduction of new processes in the public sector and relevant supply markets)
2. monitoring the utilization of GPP in actual procurement activities (the degree to which policies and tools are embedded in day-to-day practices)
3. monitoring the objective impact of GPP adoption in reduced environmental footprint (quantifying the degree to which environmental sustainability objectives are being met)
4. monitoring market developments (the degree to which the leverage effect of public procurement is reshaping market structure)

Within this wide range of monitoring requirements the EEIO toolkit is primarily relevant for the third class of monitoring activities (impact) and to some extent also the fourth class. Here we will focus on its role on impact monitoring as the main requirement.

The challenge of GPP Impact Monitoring

Monitoring the impact of GPP policies on achieving environmental objectives serves multiple critical functions:

• establishes a baseline (status quo) from which to assess policy effectiveness
• via the substantiated communication of progress it demonstrates commitment (which also further incentivises the private sector)
• it improves the accountability of contracting authorities and enhances the transparency and communication quality
• it may improve the GPP implementation effectiveness by identifying hotspots, bottlenecks and other challenges

As mentioned, a defining aspect of GPP impact monitoring is the variety of economic activity impacts that must be reliably integrated. Consequently, this challenge led to the development of a variety of methodologies and approaches. These all come with their own specific strengths and weaknesses, and it is important to articulate in each case what these are. Irrespective of the specific environmental impact type, the conceptual work of the GHG Protocol provides useful guidance about the criteria and principles that underpin high-quality impact monitoring.

The GHG accounting principles we enumerate next formally apply (with some variations) to city-wide, corporate or project based GHG emissions accounting and reporting. They are also very relevant in the context of public procurement. Note that monitoring, measurement and accounting are effectively synonyms at this level of the discussion.

• The Relevance Criterion. The scope of the measured impact should reflect the legal responsibilities, organizational structure and business models of the entities engaging in procurement activities (both buyers and sellers). An important consideration here is the concept of both direct and indirect impact and accountability (e.g., via upstream or downstream supply chains, or Scope 1, 2 and 3 in the GHG protocol nomenclature).
• The Completeness Criterion. The scope of the measured impact should be comprehensive, reflecting all material impacts that are attributable to a specific procurement contract but also covering the entire portfolio of procurement activities.
• The Consistency Criterion. Measured impacts should be readily comparable across all contracting activities, even when these concern very different sectoral and regional domains.
• The Transparency Criterion. Measured impacts should be based on traceable, verifiable (auditable) and reproducible methodologies and data.
• The Accuracy Criterion. The quantified impacts should be reasonably certain (correct within a certain range). Any uncertainties should be reported and managed (e.g., by applying a Conservatism principle).

The above criteria are effectively common-sense expectations. They, by-and-large, reflect the long body of practice that has developed around financial accounting principles. They are necessary ingredients if the rolling out of GPP policies is to survive the stresses and tensions of real economies involving a wide range of economic actors. The elevation of GPP impact monitoring to a desired level of rigor will be a long process. Much of the required infrastructure and methodology are still to be developed, let alone be validated and show to work under various conditions. The role of EEIO frameworks in this context is indeed interesting because they naturally fulfill some of the above criteria that are otherwise difficult to satisfy.

Using EEIO for GPP Monitoring: How To

The core concept behind Input-Output frameworks

A few words first about early Input-Output (IO) frameworks that were developed for the comprehensive analysis of the structure of economic activity. IO frameworks constitute arguably the most complete macro overview of global economic interdependencies. As tools of holistic economic analysis and policy they are hardly new: the first published IO type analysis by Leontief dates back to 1936. Their primary input data (the underlying IO databases) provide a comprehensive accounting matrix of who engages in economic activity (transacts), with whom, and for what purpose (what good or service is actually exchanged). This accounting exercise can be performed both the monetary level (tracking goods and services in terms of flows of money) and the physical level (tracking the actual quantities of goods being provided).

It is useful to distinguish between this fundamental IO data collection and classification activity, typically done by national statistical agencies, and the subsequent IO economic models built from that database using additional assumptions about economic behavior. The essence of such IO economic models is to postulate that the technologies used by different sectors is changing only slowly over time (following cycles of capital investment etc.) whereas the actual volume of activities is more rapidly variable, e.g., subject to changes in supply and demand, various shocks etc. Such IO models provide rudimentary forecasting capability as to what might happen under various scenarios. While scenario / forecasting ability may certainly be useful in GPP context, the primary and immediate role of IO frameworks for our purposes is as a statistical accounting database.

The emergence of EEIO extensions

Since the seventies there have been significant methodological enhancements that expand the scope of traditional IO analysis to provide additionally a diverse set of environmental and social impact indicators. More concretely, Environmentally Extended Input-Output (EEIO) databases introduce additional datasets (such as stressor or environmental impact intensities) that aim to link the volume of monetary (or physical unit) exchanges to environmental impacts. A fundamental assumption of such extensions is that impact scales linearly with economic activity. Thus, if $A$ is a measure of activity (e.g., oil and gas consumption in millions of Euros) and $f$ is the GHG emissions intensity of that energy mix, the impact is simply:

$$ I = f A $$

The potential of EEIO systems to support sustainability-related accounting has been already recognized and utilized in the broader context of GHG Accounting . It is for example an available approach for attributing emissions to individual lending contracts in sustainable finance methodologies such as PCAF . An important and rather unique capability of this framework derivers from the fact that the economic activity measure $A$ can be suitably generalized to capture not just the direct activity of a supplying sector but also indirect activities in that sector’s supply chain.

Currently, there is a wide range of official initiatives that further enhance and refine the capabilities of EEIO tools. Nevertheless, they do not constitute a magical solution that ticks all the desirable boxes we enumerated above. Before we can take stock though, we need to dig a bit deeper into the nature of the EEIO data and how they can be used for GPP monitoring purposes.

The economic side of EEIO databases (the $A$’s) are compiled by national statistical agencies, principally from business survey data. The environmental impact intensities (the $f$’s) are sourced from the corresponding competent authorities (IEA, EEA, IPCC etc.). In terms of geographical coverage, in modern interconnected economies the environmental impacts from economic activity in one region may be manifesting in a completely different world region. Therefore, it is frequently necessary to use larger, so-called MR-IO databases, that provide multi-regional scope. Costs considerations and commercial privacy issues limit the granularity and frequency by which economic relationships can be established.

EEIO databases are evolving slowly through the collection of more data and the refinement of methodologies. Historically a five-year schedule of updates has been typical. This long timescale reflected the observed timescale of changes in underlying technologies and the cost-benefit tradeoffs in the compilation of the datasets. With increased digitisation and expanded use, there is a general trend for larger, more granular and more frequently updated databases.

Some Preliminary Work

The above EEIO toolkit must slot into a conceptual model of the various processes that constitute GPP and in particular GPP impact monitoring. In previous work EU Datathon 2022 we provided a proof-of-principle demonstration of how EEIO environmental impact data can be connected to public procurement contract data to estimate GHG emissions of existing procurement portfolios.

• In Part 1 we motivated and defined the scope of the study that explored the possibilities using open Public Procurement data from Europe (TED).
• In Part 2 we reconstructed an economic representation of the public procurement process on the basis of the available contract data.
• In Part 3 we associated and inferred through the sectoral profile of the procurement contract (the main CPV category) the amount of CO2 emissions that can be attributed to these activities. This was done via a mapping of CPV sectors to NACE sectors, which then enables the use of Eurostat EEIO data for GHG emissions per sector.
• In Part 4 we discussed further how the public procurement oriented data framework we have developed within the open source Equinox platform can be mapped into the canonical set of portfolio management concepts and analyses.

Our proof-of-concept utilized readily available open data sets and demonstrated that (subject to contract data quality) one can already identify and rank the environmental impact of different procurement activities (per product and region) at the macro level.

After that brief overview of the modalities of utilizing EEIO frameworks in GPP monitoring context it is worth reviewing EEIO against the desired criteria:

Using EEIP in GPP Monitoring: Pros and Cons

Relevance

The relevance of EEIO data in GPP context is a critical aspect for market acceptance. As a starting point, IO transaction data are directly reflecting payments from buyers to sellers (on annualized basis) which establishes a strong relation. In particular the monetary cost as inferred from awarded public procurement contracts is represented directly in IO context as payments from the public sector to various industries. In IO nomenclature public procurement is part of the so-called final demand, alongside with household demand.

Yet the aggregate (statistical) approach underlying EEIO frameworks somewhat muddies the above picture. There is no direct association of impact for any concrete contract / supplier. EEIO works on the basis of assumed homogeneous pools of similar contracts/agents (grouped by sectors and regions). In other words, idiosyncratic aspects are by construction ignored (averaged out). This is as with any statistical model.

In practice there are mechanisms that can be introduced to mitigate this shortcoming (when that is necessitated by the use case). Such mitigating procedures are widely used in, e.g., financial risk management, and are in effect expert-based adjustments or overrides of the standardized result. This is analogous to the situation where statistical credit rating methodologies are developed on the basis of a pool of homogeneous entities but can be adjusted when considering a new contract on an individual basis to accommodate special circumstances. While this mechanism obviously opens the door for subjectivity, if well managed it can contribute valuable information towards iterative improvement of methodologies and policies.

Completeness

The macroeconomic approach of EEIO means that all significant economic sectors and regions are (in-principle) covered. This is an important advantage of EEIO, as it eliminates potential blind spots that might originate in more partial data collection.

A direct benefit of this “macro-completeness” is the potential of reconciliation with national GDP statistics, which is relevant e.g. in the context of monitoring national commitments such as NDC’s.

The ability of EEIO to trace supply chain dependencies relies also on this completeness of coverage: If a high-impact sector is not a direct supplier to the public sector but does provide goods or services via the supply chain of the actual providers, their impact can still be accounted for properly (if desired).

Consistency

Consistency of methodology is another important attribute of EEIO. In principle, the same conceptual data collection and integration approach is applied throughout the economy. This provides some comfort that the relative ranking of impacts by different sectors or regions, the identification of hotspots etc. more closely reflects actual ground truth and is less affected by methodological differences.

Nevertheless, in practice the low-level data collection and processing may have some idiosyncrasies due to sectoral peculiarities.

Transparency

EEIO data sets are generally open data and the compilation methodologies are published by statistical agencies and academic consortia. This provides a significant head start towards transparency and reproducibility of results, which can be further enhanced with implementing open source GPP monitoring platforms.

Accuracy

Alas, there is no free lunch. The distinct advantages of EEIO frameworks deriving from their macro / sectoral aggregation means that the methodology is also potentially less accurate compared to more direct measurement approaches. To make this more precise it is useful to recount the standard “ladder of accuracy” as articulated and used in GHG accounting context. Measurements can be classified according to their nature as:

• directly metered impacts (e.g., measurements from installed sensors on a supplier’s production facility)
• inferred impacts from physical production activities of the supplier (e.g. tonnes of cement)
• inferred impacts from economic activities of the supplier (based on monetary value flows)

Clearly, directly metered impacts are the most accurate but also the most costly and less readily available. They make sense for high impact cases and can be used for validating other approaches but are unlikely to provide the completeness level required across the public procurement universe.

Inferred impacts must make use of estimated regional and sectoral / technology based impact factors. This is where the important role of resolution or granularity comes into play. For sectors that utilize a mix of technologies (with differentiated impact) it is important for the purposes of GPP monitoring to differentiate between them.

The canonical approach to estimating impact of a new proposed procurement contract is described concisely by the GHG Project Protocol. The requirement is to estimate the Business-as-Usual emissions (baseline impact, using currently existing technologies, without GPP policies) alongside the reduced emissions from adopting an environmentally friendlier alternative (i.e., applying a GPP policy).

Clearly such a tailored approach is more data intensive as it requires the disaggregation of sectoral or regional impact profiles into at least two (and potentially more) profiles, each one reflecting a concrete set of alternative technologies. Such disaggregation is standard procedure in the EEIO literature. They can be done at two levels: the easier level introduces additional impact factors and market shares of different technologies within each sector or region, without changing the overall structure of the economy. The more profound disaggregation methodology attempts to generate a more granular transaction matrix (replacing sectors or regions with subsectors / subregions). This exercise is more computationally expensive as the number of relationships that need to be established scales with the square of number of subsectors.

Importantly, the disaggregation approach remains statistical in nature. Any specific procurement contract (and its associated GPP clauses and criteria) must be mapped into a limited menu of options. In the end, the amount of detail incorporated in EEIO systems will always express a compromise between cost and utility for the intended uses. For use cases that require highly accurate and specific information it would be natural to augment the default EEIO framework with additional measurements (i.e., hybrid approaches).

Towards Sustainable Procurement Portfolio Management

In the above analysis and review we incidentally sketched a range of practices that emerge with the advanced institutionalization of GPP processes. It is useful to anticipate how a more fully fledged GPP-based procurement lifecycle would look like when it reaches a mature, fit-for-purpose, stage and in particular whether an EEIO based impact attribution framework is well integrated with other considerations.

Envisaging the complete set of information flow requirements (data and methodologies) can help the planning, integration and prioritisation of investments in data infrastructures, training etc. Useful guidance in this respect is offered by domains where qualitatively similar requirements have been addressed in the past. In a dedicated white paper we framed the overall task of Sustainable Public Procurement as an instance of Sustainable Portfolio Management, and we outlined three distinct information processing pillars that are relevant in this context. Financial portfolio management frameworks used by lenders, investors etc., around the world are one important such reference point. Sustainable portfolio management can be seen as an extension of pure financial portfolio management, in the sense that new considerations are added to existing requirements. In GPP context central portfolio management processes (1. attribution, 2. origination, 3. allocation) create their own new requirements. Here we provide a brief summary of these ideas.

1. Attribution Processes

Monitoring the ESG impact of an existing portfolio of procurement contracts maps in to the accounting and attribution methodologies of financial portfolio management. It is the equivalent of extracting relevant attributes of financial contracts from term sheets (amounts, durations, conditions, cash flow schedules etc.) and thereby establishing a monetary value (using further reference information such as market rates, valuation models etc.). In similar vain, environmental impact attribution associates with each existing contract a range of environmental indicators. Attribution is done either at the individual contract level or on a pool basis - if there are large numbers of effectively identical contracts.

The role of EEIO in the attribution context was the central topic of this blog post: how to provide direct and indirect impact monitoring on a portfolio-wide basis. An important ultimate requirement in this direction is to be able to provide reliable, convincing explain functionality, namely interpret and report the period-on-period developments of portfolio-wide environmental impact on the basis of sectoral and regional developments. Such explanation must identify both varying amount of activities and potentially the changing nature of activity (e.g., decarbonization).

2. Origination Processes

Engaging in procurement activity requires among others the preparation of tenders, market search activities eventual evaluation and selection, post-award monitoring etc. In this “origination” context the EEIO-based monitoring framework plays a smaller but important role: it provides a standardized indicator of anticipated impact (direct and indirect). In the most basic form this is a lookup table that multiplies impact intensities with attributes such as total contract value, total product volume etc. Yet in more material procurement the evaluation of a new proposal may involve significant additional analysis.

3. Resource Allocation Processes

The important third pillar that complements the previous two concerns the forward-looking budgeting and planning of resources. This is a process that allocates (reserves) impact “budgets” to the various procurement portfolio segments. It is the analog of allocating financial capital or human resources towards investments and projects. This activity does not evaluate individual contracts but sets sectoral or regional limits, fixes GPP criteria, develops long-term strategies etc. over a planning period that may be spanning from one year to multiple decades. This task requires developing and using realistic scenarios for the availability of alternative technologies. Ultimately it is the performance against those targets that measures the success of GPP policies. EEIO models can be used in this direction in an auxiliary manner, for example assessing the economic impact of changing public sector demand.