Data Stops at Capital: Mapping the Climate Investment Evidence Gap
The argument in Dig Up Stupid is that hard-to-abate industrial sectors — steel, cement, aluminum, shipping, aviation — absorb a fraction of climate capital relative to what they emit. That claim needed a number. Here’s how I tried to get one, and where the data ran out.
The emissions side
Climate TRACE publishes asset-level emissions by subsector, licensed CC 4.0, updated to 2024. The API (v6) returns global annual totals by subsector in a single call — no ZIP download required for aggregates. Steel came in at 2.8 Gt CO₂e for 2023. Cement 2.3 Gt. Aviation and shipping combined: 1.9 Gt. Against a fossil/industrial baseline of ~50.7 Gt (excluding land-use), those six sectors together sit at roughly 13.5% of global emissions. Add oil & gas and road transport and you get to the ~41% the article cites from WEF — consistent with Climate TRACE.
One flag: the brief specified EDGAR (IEA-EDGAR CO₂, 2024) as a cross-reference. The XLSX credits the CO₂ component as CC BY-NC-ND 4.0 — no derivatives — not the CC BY 4.0 the brief listed. Climate TRACE covers the same sectors, so nothing was lost, but worth knowing if you cite EDGAR elsewhere.
The investment side
This is where the data runs out. BloombergNEF’s sectoral investment breakdown is paywalled. The public summary PDF gives a single aggregate: “clean industry” received ~$155B in 2024 against $2.1T in total clean energy investment — a 7.4% share for all hard-to-abate sectors combined. There is no public download that breaks this figure out by steel, cement, or shipping individually. The chart therefore shows emissions shares as bars, with a single reference line at 7.4% marked in red. The visual argument still holds — steel alone (4.7%) exceeds the entire clean industry investment allocation.
The CFC chart and the WSA wall
For the Montreal Protocol chart, the UNEP Ozone Secretariat table is query-only — no bulk download, manual extraction per substance per country. The same underlying data is served via Our World in Data (CC BY 4.0), which is what the CFC consumption chart uses. One gap: 1987 and 1988 are absent from the UNEP country-reporting dataset, so the reference line marking the Protocol’s signing falls between two data points rather than adjacent to one.
The steel carbon intensity analysis hit a similar wall. The World Steel Association publishes BOF/EAF production volumes by country — exactly what you need to compute tCO₂e per tonne of steel — but not as a bulk download. So the steel chart shows absolute emissions rather than intensity. China peaks near 1,800 Mt in 2020 and plateaus; India rises continuously from 2015 through 2023. The divergence is visible without normalisation, though the intensity story is richer.
The constraint worth noting: the investment side of a climate investment argument is harder to verify than the emissions side. Emissions data is increasingly open, asset-level, and machine-readable. Capital allocation data is not. That asymmetry shapes what any data journalist can actually show.
