Establishing a framework

In order to construct a coherent investment framework, one needs to identify (1) the underlying investment universe and (2) the key issues within the environmental, social and governance pillars (which are often referred to as a ‘materiality map’). Bloomberg is currently launching corporate ESG scores on a rolling basis. These scores are based on a comprehensive database of publicly disclosed company-level data. As part of this effort, materiality maps are created per sector (as defined by Bloomberg’s industry classifications BICS), following which company level sector scores are constructed based on the relevant metrics. For commodities-related ESG research, we expect to follow the Bloomberg corporate score rollout since it both ensures data quality checks and identification of the relevant issues (via vetting of the materiality maps). A more detailed understanding of the Bloomberg ESG methodology can be found on BESG . The ‘scores’ section contains detailed documentation on the construction of company level scores to date.

The Bloomberg Commodities Index (terminal ticker is BCOM Index) is a leading commodities benchmark. As of September 2020, it contained 23 commodities covering the five major sectors. Benchmark replication is carried out using US dollar denominated futures. Many investors allocate to either the flagship BCOM index or the sector indices; with the aim being to diversify portfolio returns, hedge inflation risk and, increasingly, to provide exposure to alternative risk premia.

Given its prominence both in sustainability policy discourse and amongst investors, we begin with GHG emissions. In subsequent publications, we will extend this analysis to match other issues found in the materiality map. Since

the BICS framework allows for an easy mapping between metals and mining producers and the metals included in the BCOM benchmark, we start our analysis by focusing on the industrial and precious metals sectors.

Given the investible instruments are futures contracts, we need to measure GHG exposure per contract or per US dollar. In-line with standard asset allocation decisions, we use the US dollar as the unit of measurement. Calculating the GHG per US dollar for each metal is a two-step process: identify a suitable data source and transform the raw data to a US dollar metric.

An aim of this publication is to explore possible avenues by which existing commodities investors can incorporate ESG objectives into their allocations. Presently, the dominant route appears to involve allocating to a commodities benchmark followed by an offset trade. (E.g. carbon offsets). Alternatively, some investors have abstained from certain commodities or sectors (E.g. BCOM excluding Agriculture & Livestock). We examine an alternative approach whereby reweighting commodities within a portfolio can reflect investors’ ESG preferences and thresholds. We also hope this framework will help re-engage those investors reluctant to allocate to commodities because they believe it is incompatible with ESG investing.

Literature Survey

A review of the literature on the metals’ extraction processes highlights several factors which complicate GHG estimation. The main findings are:

1. Metals are jointly extracted (and/or are by-products);
2. Mining equipment varies by company;
3. Significant geographical variations in soil-type and deposit access;
4. The energy sources for extraction vary by country; and
5. Recycled production of metals is less GHG emissions intensive than primary extraction.

Academic and practitioner studies on estimating GHG scores for metals take one of two approaches, which can be summarized as follows:

1. Macro-based: Combine GHG estimates from supranational agencies with production estimates from industry bodies and
2. Micro-based: Use the data available in public companies’ annual and sustainability reports.

The first approach has the benefit of aggregating different production methods and geographical variations, but suffers from a high degree of opacity with respect to estimation methodology and the underlying data sources. The second method is typically based on a limited number of companies and can be biased by region or the companies selected for the study. We use a regression-based method that combines the two approaches above; it aggregates a wide universe of company level data to form a macro-level estimate.