A Case Study Of Top Four US Banks In The Energy Sector

Financed emissions targets reported by the top four US banks (Citibank, JP Morgan, Morgan Stanley, and Goldman Sachs) have a long way to go before they approach the comparability of numbers in financial statements.

An increasingly important area in carbon accounting relates to emissions indirectly financed by banks. In this piece, I look at how financed emissions are constructed and what they might mean.

For this exercise, I ignore banks’ own operational emissions as they are relatively trivial. Every financed sector (e.g., energy, auto, power, steel, cement) is massive and involves unique challenges related to mitigation, adaptation, measurement, and reporting. To keep the task manageable and restrict this piece to a readable length, I focus on the 2030 targets promised by the top four US banks for the energy sector. Financed emissions are thankfully subject to at least some kind of broad principles based standard set by the Partnership of Carbon Financials, or the PCAF. Despite the presence of PCAF, I found considerable dispersion in how individual banks report financed emissions.

Right off the bat, I was surprised that Citibank, and JP Morgan Chase report climate numbers for 2022 whereas Morgan Stanley and Goldman, surprisingly, have only put these out for 2021. Unless otherwise stated, I use the latest available reports for the analysis.

I considered heterogeneity in the following factors to compare these banks: (i) sub-sectors covered and the nature of the targets promised; (ii) methods underlying the measures used; (iii) rank ordering in terms of impact; (iv) dollars committed to the sector and exposure of such dollars to physical and transition risk; and (v) data quality.

1.0 Sub-sectors and targets promised

· Citi has promised to cut scope 1, 2 and 3 emissions for the energy sector by 29% from a 2020 baseline of 143 million metric tons of C02 equivalent. The 2030 target is to reduce emissions to 102 million metric tons of CO2 equivalent. Citi is the only bank in this set that promises a cut in absolute emissions. The others promise cuts in emissions intensity as seen next.

· Morgan Stanley (MS) promises to cut scope 1, 2 and 3 emissions for the energy sector by 29% from a baseline of 2019 of 1,839 tons of C02 equivalent per million dollars of lending commitment.

· JP Morgan (JPM) has different targets for scope 1 and 2 emissions relative to scope 3. JPM promises to cut scope 1 and 2 emissions by 35% from a 2019 baseline of 5.3 grams CO2 equivalent per mega joule of energy (g CO2e/MJ) and by 15% for scope 3 from a 2019 baseline of 66.5 grams CO2e/MJ.

· Goldman Sachs (GS) promises to cut scope 1, 2 and 3 emissions for the oil and gas sector by 17-22% from their 2019 baseline of 72 grams CO2 equivalent per mega joule of energy (g CO2e/MJ).

2.0 Methodology underlying these measures

· Industry coverage:

“Energy” in Citi appears to include “integrated oil & gas, oil & gas exploration & production, oil & gas storage & transportation, oil & gas refining & marketing, oil & gas equipment, services and drilling and commodity traders.” “Oil and gas,” as per page 48 of JPM’s report covers, “exploration, production, oilfield services and other oil and gas.” “Energy” as per MS net zero methodology report, covers “oil & gas drilling, oil & gas equipment & services, integrated oil & gas, oil & gas exploration & production, oil & gas refining & marketing and oil & gas storage & transportation.” It looks like MS does not include commodity traders in its definition. GS does not seem to define what sub-sectors are included in “oil and gas.”

· Net zero scenario used:

MS states that they use the absolute IEA (International Energy Agency) NZE (net zero emissions) 2050 emissions pathways for oil and gas to define the Energy sector’s 2030 interim target. To align with the IEA pathway, MS set a 2030 interim target to reduce their financed emissions lending intensity by 29% compared to the 2019 base year. Citi also uses the same IEA NZE 2050 pathway but has promised a similar 29% cut by 2030 but in absolute emissions.

GS states that their targets are based on the “Goldman Sachs Carbonomics 1.5°C net zero path which assumes a carbon budget for remaining net cumulative CO2 emissions from all sources from 2020 to be c.500 GtCO2, in line with the IPCC AR6 WGI Summary for Policymakers, and consistent with a 50% probability of limiting warming to 1.5°C by 2100.”

Their 2030 target is a CO2 equivalent per mega joule of energy of 56 – 60, which entails a 17 – 22% reduction on the 2019 baseline.

For operational emissions of the oil and gas sector, JPM states it relies on the “IEA SDS (Sustainable Development Scenario) with methane added based on supplemental IEA data consistent with SDS.” For end use emissions of oil and gas sector, JPM relies on “IEA SDS with adjustments for nonenergy oil and natural gas demand.” This leads to a 35% intensity reduction from 2019 baseline by 2030 and a 15% intensity reduction from 2019 baseline for end use reduction.”

Are there substantive differences in these three scenarios: IEA NZE 2050, IPCC (Intergovernmental Panel on Climate Change) AR6 WGI (working group I) and IEA SDS with methane added? I had a hard time making up mind on which of these was conservative. For instance, Greenpeace asks financial institutions to use the IEA NZE 2050 as the minimum standard and avoid IEA SDS. But Greenpeace is not referring to IEA SDS with methane added.

· Methods used:

Citi states that it calculates absolute financed emissions metrics using committed funds (the capital available to a client for a certain use). This number is presumably divided by the sum of equity and debt of the firm and the quotient thus derived is multiplied by the scope 1, 2 and 3 emissions of the energy firm, as per the PCAF standard. Their 2020 TCFD report’s appendix does not seem to clarify how they would deal with situations when committed financing is much larger than actual debt used in the firm’s enterprise value calculations.

Citi’s definition of the “energy” industry is the most expansive. In its 2020 TCFD report, Citi clarifies that to minimize double counting in the value chain, for Scope 3, they focused on the extractive and refining sectors, but they did not include Scope 3 emissions for the transportation and storage, retail and marketing, and other energy sub sectors. The other banks don’t seem to exclude such emissions, but the definition of “oil and gas” for GS and JPM may not cover transportation and storage, retail, marketing sub-sectors in the first place.

MS starts with firm’s annual GHG emissions and multiplies that by the ratio of lending commitment to a firm divided by its enterprise value (market value of equity plus book value of debt including cash). That ratio gives us number of tons of carbon financed per firm by MS. These lending commitments for firms within a sector are then aggregated across sectors by weighting them by the ratio of MS’s lending commitment to a specific sector relative to all sectors. Much like the comment for Citi, this works well if lending commitments are not orders of magnitude larger than actual debt used in the enterprise value of the firm.

JPM states that they compute Operational Carbon Intensity as (scope 1 + 2 emissions – credits (g CO2e))/(embedded energy in oil + gas + bioenergy (MJ of mega joule)) whereas End Use Carbon Intensity is computed as (scope 3 emissions – credits (g CO2))/(embedded energy in oil + gas + bioenergy + other renewables (MJ)). JPM’s Carbon Compass methodology document says that JPM intends to monitor such physical intensity by sector and for the portfolio. I could not find clear definitions of how these emission intensity measures are weighted for total dollars committed or withdrawn by the sector, relative to such dollars across all sectors.

GS defines Oil Gas Client Intensity = scope 1 + scope 2 + scope 3 end use emissions – Carbon offsets (gCO2e)/embedded energy produced (MJ). On page 43, GS clarifies that they also multiply the emissions per mega joule measure by the ratio of client financing to sector portfolio financing. Financing includes “corporate lending commitments, debt and equity capital markets financing; and other on-balance sheet debt and equity investments.”

Note that JPM and GS appear to allow for carbon offsets and credits whereas Citi does not. JPM scales by embedded energy in oil, gas, bioenergy, and renewables whereas GS simply states, that its divisor above is “embedded energy produced.”

3.0 Can we rank which bank is doing more to decarbonize in the energy sector?

Citi is exemplary at reporting several versions of their emissions measure, which, in turn, enables some degree of cross-bank comparability. For instance, Citi states that their 2021 intensity per million dollars of lending commitment is 2,260 metric tons whereas its physical intensity measure is 81.8 grams of CO2 equivalent per mega joule of energy. Recall MS’s 2019 baseline intensity measure was 1,839 metric tons per million dollars committed. As mentioned, both JPM and GS seem to report around 72 grams of CO2 equivalent per mega joule of energy.

On the surface, Citi appears to have the highest carbon exposure of all the four banks. But that is almost mechanically true given their more extensive coverage of sub-sectors under “energy” and hence the larger number of firms represented. And how does a user compare MS with its lending commitment intensity measure with GS and JPM’s physical intensity measures?

4.0 Progress related to impact

Even if cross-bank comparisons are hard, one can at least compare time-series or historical trends for the same bank.

Citi’s numbers seem to suggest that they have already hit their 2030 target in 2022. Page 58 of Citi’s 2022 TCFD report states their 2021 Financed Exposure number is 100.3 million metric tons of CO2 equivalent relative to 143 million metric tons in 2020. A footnote suggests that adjusting for volatility in the enterprise value number leads to a higher 113 million metric tons number. Regardless, the decline is impressive, but it is not obvious how such a decline was achieved. In its 2022 report, JPM reports no change in operational emissions and a 1% increase in scope 3 emissions for oil and gas.

Assessing progress for MS and GS will have to wait till they release their 2022 reports.

5.0 Dollars committed to the sector

Citi in their 2022 TCFD report state the 2021 year-end committed exposure to energy sector was $44.4 billion. Their funded exposure is $13.485 billion. Note the vast difference between committed and funded exposure. Also recall that Citi has used committed exposure to compute emissions suggesting that their share of firms’ emissions is arguably overstated and hence conservative.

In their 2022 TCFD report, JPM reports “total credit exposure” for their “oil and gas” sector at $42.6 billion as of December 31, 2021. Total credit exposure is defined to include retained loans, lending-related commitments, and derivative receivables. Hence, Citi and MS’ exposures to oil and gas are similar in dollar magnitudes, subject, of course, to the earlier question of industry coverage. The 2021 climate reports for MS and GS do not appear to report dollars committed to or drawn by the sector.

6.0 What about physical risk and transition risk exposure?

Citi reports total dollars committed and drawn by sub-sectors in the energy area, and the associated severity of physical and transition risk for each sub-sector. Citibank ranks the energy sector as low on physical risk (ranked 2) relative to transition risk (ranked 4), except for commodity traders whose transition risk is rated at 3. Physical and transition risks are reported to be considered over the medium to long term. I am not sure how long is the long term. GS clearly defines long term as a period beyond seven years.

Among the others, JPM is the only bank that discloses dollars committed to sub-sectors of oil and gas industry and the relative degree of climate risk. For instance, JPM rates the physical risk associated with the two subsectors it considers (exploration & production and oilfield services and other oil and gas) as “high” and carbon intensity as “very high.” I could not find their transition risk scores for these two sub-sectors in their climate report.

I did not see detailed disclosure of the dollars committed to sub-sectors and the severity of physical and transition risks related to such exposure in GS and MS reports.

My understanding is that these physical and transition risks apply to the sector as a whole and do not necessarily adjust for whether such risk will come to pass during the period the loan is outstanding with the bank. That consideration is presumably accounted for in the climate stress test that Citi and the other banks run.

As an aside, I find these stress tests to be highly opaque because dozens of assumptions and estimates go into the black box. Will any bank be willing to leave an excel model with its assumptions on its website so that the user can kick the tires and understand what is really behind these stress tests?

7.0 Data quality underlying these estimates

Citi is also transparent with the quality of data underlying these estimates. In the energy sector, for 2021, they report that 65% of their portfolio reports scope 1 and 2 emissions. Remarkably only 3% appear to report scope 3 emissions and that percentage has fallen from 18% last year on account of technical reasons. The PCAF scope 1 and 2 data quality score (out of 5) is 2.9. The PCAF scope 3 data quality was 3.6. I especially love Citi’s quantitative emphasis on data quality.

MS chairs the data quality group at the PCAF. One of the statements in the MS report really resonated with me, given my prior work on data quality of emissions: “two financial institutions that provide similar levels of financing to the same company may report different financed emissions figures based on which data vendor they source emissions data from.”

MS reports that the data quality score for GHG scope 1 and 2 emissions, in aggregate, was between 3 and 4. For Scope 3, their data quality score was between 3 and 4.11. MS did not break down the data quality score by sector financed, unlike Citi. I did not see similar detailed quantitative discussions about data quality in GS and JPM’s reports.

In sum, these comparisons are somewhat insightful but have a long way to go before they approach the relative comparability of numbers in financial statements. It continues to be difficult to know whether the measurement of financed emissions is decision relevant for the bank and the investor, both in terms of economic impact to the bank’s bottom lines and for society in terms of environmental impact.

Of course, this strategic ambiguity may be deliberate and arguably even good for society. If we squeeze the brown sectors too much by denying them credit, they will simply find less transparent ways to raise capital. Is that good for anyone?