Abstract: |
Scope 3 emissions constitute the majority of a company’s greenhouse gas emissions but are notoriously difficult to measure due to data quality and coverage issues. To address these challenges, we develop a methodology that uses Scope 1 and 2 emissions, as well as mathematical properties of complex networks, to estimate a company’s upstream and downstream emissions. Our model is constructed under the reasoning that, in a closed economy, a company’s indirect emissions result from the direct emissions of clients, partners, and suppliers in the supply chain. As a result, calculating indirect emissions can be reframed as a question of attribution of direct emissions in a client-supplier network. This fact informs the two assumptions needed for constructing our model: First, we assume that we have complete knowledge of business-to-business (B2B) transactions among firms in a client-supplier network. Second, we assume that we observe all Scope 1 and 2 emissions of the companies in this network.
This initial derivation of our model represents the ideal case where transaction network data and environmental data are fully available. For this reason, we refer to this as the Full Information scenario. Unfortunately, the conditions for this scenario are seldom achieved in practice. The breakdown of our assumptions motivates a second, less strict version of the model. In this second version, we relax the requirement of knowing the values of all B2B transactions between companies. Instead, this requirement is simplified to a binary Yes or No indicator of whether two companies are engaged in a commercial transaction.
Two main implications arise from this change of definition. In the Full Information scenario, the B2B transaction data was used to assign the weight to each client (supplier) represented in a company’s downstream (upstream) emissions. Since the information on B2B data is replaced by a binary indicator, this will result in all clients (suppliers) receiving an equal weight in our calculations. Because of this equal weighting, we refer to this case as the Naïve scenario. On the other hand, since client and supplier disclosure data is much more ubiquitous than B2B transaction values, this change results in a considerable expansion in the number of companies in our sample of production network data.
In the final sections of our paper, we present results from an empirical application of our methodology on two datasets: FactSet Supply Chain Relationships for production network data and Trucost for data on Scope 1, 2, and 3 emissions. Our sample includes company data from 2010 until 2021 and produces two main results. First, we find a very high correlation when comparing the results of the Full Information and Naïve scenarios. This result has important implications since it indicates that the more flexible, applicable, Naïve scenario provides a reasonable estimate for the more robust Full Information scenario. Second, we apply the Naïve model to our complete sample. When comparing these results to the reported Scope 3 figures in Trucost, our calculations generate larger estimates in more than 60% of instances. Because of the high correlation between both scenarios, this 60% figure may indicate that the more accurate Full Information scenario might produce a similar result. |