The science behind spend-based emission factors

‍Spend-based emission factors have a bad reputation. But by keeping the limitations in mind, they can be an easy and powerful starting point to measure Scope 3 emissions.

Spend-based emission factors are often considered a second-best option, especially compared to activity-based measurements. Critics point out that measuring emissions by spend can come with shortcomings: 

• Firstly, they can be too generic, ignoring differences between products, services, suppliers, industries, countries, and geographies. 
• Secondly, spend is a bad proxy: When your vendor offers you a discount, your emissions decrease, though in reality nothing has changed.  

However, spend-based factors are better than their reputation suggests—provided they are used carefully and with an awareness of their limitations, which we will explain in detail below. If users keep these limitations and their potential for errors in mind, spend-based factors offer companies an accessible starting point for estimating their emissions—especially for Scope 3, which otherwise can be a laborious task. 

‍

The power of spend-based factors

The spend-based method’s biggest advantage lies in its pragmatic nature. While the alternative to spend-based emission factors—activity-based emission estimates—are more accurate, the collection and cleaning of all data points is difficult and cumbersome. Due to its complexity, most companies are unable to follow the activity-based approach. As a result, a thorough spend-based approach makes for a powerful alternative.

Here’s why we at Climatiq value the spend-based approach: 

• Companies are generally organized around cost and value. Accounting or procurement departments usually have a solid understanding of which goods or services are purchased across their supply chains and how much money is spent.
• This spend data, when combined with spend-based emission factors, gives companies a solid starting point for estimating their Scope 3 carbon footprint.
• Spend-based emission factors are easy to use. They estimate average emissions per unit of currency through broad, high-level modeling. In such a scenario, the modeling assesses the total money spent in a given sector of the economy, the total emissions generated by this sector, and the impact of the trade of goods into and out of the economic region in question.

With this in mind, the spend-based approach can be a powerful tool for measuring emissions, provided the underlying "science" is carefully considered and applied.

Let’s take a closer look at how spend-based emission factors work. In doing so, we can determine ways for companies to mitigate the limitations that come with the spend-based approach.

‍

How are spend-based emission factors modeled? 

Spend-based emission factors are estimated using 1. Input-Output Models that are 2. applied internationally and then 3. Environmentally Extended (“EE MRIO”). What exactly does this mean?

• Input-Output Analysis (IOA) is an analytical framework used to describe, understand, and analyze the interdependencies between industries within an economy. It was originally developed in 1930 by Wassily Leontief, who was awarded a Nobel Prize for economics in 1973 for his work on econometrics. This method underpins many economic models and frameworks and is based on national accounting matrices known as input-output tables (IOTs) (see Miller and Blair, 2009).

• Multi-Regional (MR) refers to models that contain IOTs corresponding to many countries, often covering a global scale. These global models can describe entire global supply chains. 

• Environmentally Extended (EE) refers to economic models complemented with variables that quantify the environmental impacts of economic activities, i.e. the impact on water consumption, energy use, air pollution, and greenhouse gas (GHG) emissions. The quantified impact on GHG emissions are used to derive spend-based emission factors. 

In a broad sense, EE MRIO models represent a snapshot of the global economy in a given year. For more details on how an IOT is constructed, please refer to Miller and Blair, 2009. 

‍

Let’s make this more concrete

Figure 1 provides a graphical representation of an IOT, which illustrates the inter-industry monetary flows between the 30 economic sectors in China, developed by Mi et al., (2018).

• The rows correspond to the output of a specific sector.
• The columns are associated inputs to generate the sector’s output.
• Orange corresponds to the largest monetary flow and blue to the lowest.

‍

How to read this table: the sector in column nine, for example, refers to the “Wood and Furnishings” sector. It has the strongest economic ties to sector one, Agriculture and Furnishings, as well as itself, meaning a lot of its products are used by other industries within the same sector. 

‍

How are spend-based emission factors calculated?

In an environmentally extended input-output table, emissions data is added to the economic output data for each sector. By combining monetary flows with emissions data, it becomes possible to calculate spend-based emission factors.

We now know how much monetary flow was generated in sector one and sector nine, how much sector one has contributed to sector nine, and how many emissions were generated in sector one and sector nine. As a result, we can derive spend-based emission factors by allocating the GHGs according to monetary flows. These emission factors are generally called “multipliers.” 

In scientific terms, these multipliers are represented by the square matrix ĝ in the following equation:

^G = gY

• The matrix ĝ contains emission factors that are multiplied to final demand, represented by matrix “Y” in order to obtain a matrix of total GHG emissions by sector which is represented by “G.” 
• Matrix ĝ is obtained by multiplying direct emissions intensities (i.e., emissions per unit of output, like euro) to the so-called “Leontief” matrix. 
• For other basic MRIO calculations, look at PyMrio documentation. For more details, refer to Miller and Blair (2009).

‍

How are spend-based emission factors represented?

Spend-based emission factors can be expressed in either basic prices (reflecting the cost of production at the factory gate) or purchaser prices (the final cost to the buyer, including taxes, transport, and trade margins). This means they include upstream emissions and account for the impacts of global trade, but they do not cover the use or end-of-life phases of a product.

Input-output analysis (IOA) is useful to compare consumption and production-based emissions, for example in the work of Wood et al. (2019) who examined the structure of the carbon footprint and carbon exports of Europe. Peters (2008) is another great example of this scenario.

‍

‍

EE MRIOTs usually produce emission factors—also known as multipliers—at the industry level because countries commonly report production and emissions data aggregated by industry. Afterwards, this data may be disaggregated into the products that each industry produces. Depending on the use case, companies decide which set of emission factors to use: for calculation of emissions from investments (e.g. PCAF methodology) companies have to use industry emission factors; for purchased goods and services, it is advised to stick to product-based emission factors.  

‍

The different types of EE MRIOTs 

To account for differences across regions and economic sectors, there are several different IOTs available, such as WIOD (Timmer et al. 2015), EXIOBASE3 (Stadler et al. 2018), GTAP (Andrew and Peters, 2013), EORA (Lenzen at al. 2013) and others.

Climatiq’s API offers a selection of spend-based emission factors derived from the following IOTs:

‍

Let’s talk about limitations and uncertainty

EE MRIOTs are powerful tools to understand the structure of global trade, as well as the environmental impact across industries and countries. Spend-based carbon footprinting approaches and IOTs are increasingly gaining industry interest because they are simple to use, rely on freely available datasets, and offer detailed data broken down by region and industry. 

However, IOTs come with several limitations. Here are some of the major sources of errors:

• Homogenization: IOTs are synthetically constructed from several underlying data sources. Different methods of homogenization can introduce uncertainty into final outputs (see Timmer et al., 2015)
• Unreliability of input data: Emissions data for the environmental extensions of the IOTs are usually obtained from UNFCCC National Inventory Reports or verified open databases like EDGAR (Crippa et al. 2022). Macroeconomic data is usually obtained from national statistics reports for each country. The reliability of such data can vary from region to region, which directly impacts the calculation of EFs. 
• Heterogeneity of data quality: Some countries, because of the limited data availability, are aggregated into “Rest of the World regions” that sometimes cover entire continents. For example, in EXIOBASE, Africa (other than Egypt and South Africa) is considered a single region. This affects the granularity of emission factors. 
• Industry/product averages: Spend-based emission factors are based on the average emissions intensity of all products in a given category. In reality, of course, emissions vary by product and supplier. 

In the case of EXIOBASE, which contains more than 8000 entries for each year, assessing the quality of each data point can be complex. At Climatiq, we compare data points within and across categories and geographies in order to detect entries that should be used with caution. We add quality flags to such data. You can read more about data quality flags and the method behind our outlier detection here. 

Due to their imperfection, several studies have looked at the uncertainty of IOTs: 

• Moran and Wood (2014) found that carbon footprint results for most major economies disagreed by less than 10% between different MRIOs, including EORA, EXIOBASE, and WIOD. 
• Steubing et al. (2022) compared LCA-based carbon footprints for various products and industries using ecoinvent data with results from EXIOBASE. They found good agreement between the two for sectors like agriculture, energy (excluding renewables), and manufacturing, but poor agreement for waste management and mining. Compared to EXIOBASE, carbon footprints calculated with ecoinvent were:

• -- lower for the agricultural sector (-9%) and the manufacturing sectors (-12%), 
• -- higher for the energy sector (+16%) and even higher for the mining (+43%) and water supply/waste (+90%) sectors.

• In general, they found that carbon footprints calculated with EXIOBASE can differ greatly from those calculated with ecoinvent. Despite being a positive match for some sectors, more than half of the product footprints calculated differed by a factor of 2, with carbon footprints calculated from ecoinvent being higher in more than 50% of cases. For more details, refer to the original work of Steubing et al. (2022). 

‍

How to apply spend-based emission factors 

Now, let’s clarify the ultimate question of how to apply spend-based factors. 

Input-output analysis is a powerful tool to understand supply chain emissions and can be used by companies to conduct a first approximation of their Scope 3 emissions. Frameworks such as the GHG Protocol, PCAF, TCFD, and others encourage the usage of spend-based methods where activity data is absent or difficult to obtain. They are ideal as a first assessment to pinpoint emissions hotspots, which should then be assessed with activity-based emission factors where possible. 

It's important to remember that spend-based emission factors are drawn from a global representation of emissions. They provide a solid, high-level assessment but cannot differentiate or depict nuances between products or suppliers at the local level. 

IOTs are generally not updated every single year, and data may be outdated by a few years. However, due to inflation, prices increase every year. This means that if companies use expenditure data from recent years in their emissions calculations, they must apply inflation correction in order to compare to older data from IOTs. They should also pay attention to match the spend-based factors with their specific country, as the estimated emissions for a product can differ hugely across the globe depending on where that product was purchased. 

As was pointed out above, users should check precisely which type of spend data they should use depending on source. For example, USEEIO emission factors from EPA are already in purchaser prices, including all margins and taxes—the price paid for a product should simply be multiplied by an emission factor. On the other hand, EXIOBASE provides data in basic prices, so companies have to make sure they use spend data without transport, tax, and trade margins, otherwise they may get an overestimated carbon footprint. 

If you don’t know your margins, Climatiq’s Procurement feature helps you to navigate this complexity, obtaining the basic price and providing industry-specific inflation correction (for the EU—in other countries, a nationwide inflation rate is applied). At Climatiq, we ensure that all data provided has passed our quality controls to ensure that, where activity data is not available, spend-based emission factors can be applied with confidence.

‍

—

Acknowledgement

We thank Dr. Marco Sakai at the University of York for reviewing and contributing to this article. 

‍

‍

References

de Bortoli, Anne and Agez, Maxime, (2023) Environmentally-extended input-output analyses efficiently sketch large-scale environmental transition plans: Illustration by Canada's road industry, Journal of Cleaner Production, Volume 388, 2023, 136039, ISSN 0959-6526, https://doi.org/10.1016/j.jclepro.2023.136039 or https://www.sciencedirect.com/science/article/pii/S095965262300197X

Miller, R.E. and Blair, P.D. (2009) Input-Output Analysis: Foundations and Extensions. 2nd Edition, Cambridge University Press, Cambridge.

Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R. and de Vries, G. J. (2015), "An Illustrated User Guide to the World Input–Output Database: the Case of Global Automotive Production" , Review of International Economics., 23: 575–605

Daniel Moran & Richard Wood (2014) CONVERGENCE BETWEEN THE EORA, WIOD, EXIOBASE, AND OPENEU'S CONSUMPTION-BASED CARBON ACCOUNTS, Economic Systems Research, 26:3, 245-261, DOI: 10.1080/09535314.2014.935298

Mi, Z., Meng, J., Zheng, H. et al. A multi-regional input-output table mapping China's economic outputs and interdependencies in 2012. Sci Data 5, 180155 (2018). https://doi.org/10.1038/sdata.2018.155 or https://www.nature.com/articles/sdata2018155.pdf 

Richard Wood, Karsten Neuhoff, Dan Moran, Moana Simas, Michael Grubb & Konstantin Stadler (2019): The structure, drivers and policy implications of the European carbon footprint, Climate Policy, DOI: 10.1080/14693062.2019.1639489

Peters, G.P. From production-based to consumption-based national emission inventories. Ecol. Econ. 2008, 65, 13–23.

Lenzen, M., Moran, D., Kanemoto, K. and Geschke, A., 2013. Building Eora: a global multi-region input–output database at high country and sector resolution. Economic Systems Research, 25(1), pp.20-49.

Robbie M. Andrew & Glen P. Peters (2013) A MULTI-REGION INPUT–OUTPUT TABLE BASED ON THE GLOBAL TRADE ANALYSIS PROJECT DATABASE (GTAP-MRIO), Economic Systems Research, 25:1, 99-121, DOI: 10.1080/09535314.2012.761953

Ingwersen, W. AND M. Li. Supply Chain Greenhouse Gas Emission Factors for US Industries and Commodities. U.S. Environmental Protection Agency, Washington, DC, EPA/600/R-20/001, 2020.

Crippa, Monica; Guizzardi, Diego; Muntean, Marilena; Schaaf, Edwin; Monforti-Ferrario, Fabio; Banja, Manjola; Pagani, Federico; Solazzo, Efisio (2022): EDGAR v6.1 Global Air Pollutant Emissions. European Commission, Joint Research Centre (JRC) [Dataset] PID: http://data.europa.eu/89h/df521e05-6a3b-461c-965a-b703fb62313e

Steubing, B., de Koning, A., Merciai, S. and Tukker, A., 2022. How do carbon footprints from LCA and EEIOA databases compare?: A comparison of ecoinvent and EXIOBASE. Journal of Industrial Ecology.

Stadler, K., Wood, R., Bulavskaya, T., Södersten, C., Simas, M., Schmidt, S., Usubiaga, A., Acosta-Fernández, J., Kuenen, J., Bruckner, M., Giljum, S., Lutter, S., Merciai, S., Schmidt, J. H., Theurl, M. C., Plutzar, C., Kastner, T., Eisenmenger, N., Erb, K., Koning, A., Tukker, A., 2018, EXIOBASE 3: Developing a Time Series of Detailed Environmentally Extended Multi-Regional Input-Output Tables. Journal of Industrial Ecology.