[bennet21]

Fixes historical cement trade that was too high (e.g. 1880–1900) in regions containing countries without cement production.

Previously, the backcasting reference mixed units (tonnes of production, USD of GDP for countries without production, tonnes of trade for the US). When aggregated, GDP dominated the regional reference (e.g. ~99.7 % in EUR and NEU, coming from small territories like MLT or GIB). This affected both the backcast and the 2024–2025 forecast.

Changes:

• GDP of countries without production is converted to cement tonnes using the yearly cement intensity of GDP, excluding China so that its construction boom does not dominate.
• The reference is scaled to tonnes of trade (global BACI trade / reference over BACI years), so that the USGS data for the US fit in consistently.
• The US backcast uses the full reference (incl. shipping cost) instead of production only.
• Added a note that for clinker the US reference contains clinker + cement trade (overweighting the US in custom aggregations that combine it with other countries; irrelevant for ISO, EU21 and H12).

Effect: 1995–2023 unchanged. H12 trade before 1995 is lower in regions with countries without production; 2024–2025 forecasts change by up to ±15 % (e.g. EUR imports). At ISO level, mainly small countries without production are affected.