Reconstructing grid emission intensity for Poland and testing its response to EU ETS carbon prices: A replicable...

Extended LMDI decomposition incorporating Scope 2 (indirect) emissions requires a reliable time series of grid emission intensity (EFᵍʳⁱᵈ). In Poland, official values are available only from 2014 (KOBiZE), creating a nine-year data gap. This method article documents (i) a fully replicable two-source triangulation protocol reconstructing EFᵍʳⁱᵈ for 2005-2013 from open Eurostat and UNFCCC data, and (ii) a formal test of the relationship between EFᵍʳⁱᵈ and EU ETS allowance (EUA) prices over the single-instrument Phase II-IV window (2008-2022). All inputs derive from open sources with cell-level provenance; every result is reproducible from the deposited data and scripts. EFᵍʳⁱᵈ reconstructed (2005-2013) via two independent paths - a Eurostat fuel-input backcast (R̅ = 0.876; CV = 0.86%) and a UNFCCC CRT allocation (α = 0.856) - which agree to within 9.5 g CO₂/kWh in every year; leave-one-out max error ±13 g CO₂/kWh (1.4%); series spans 928→685 g CO₂/kWh (2005-2022), a 26% decline. Over Phase II-IV (2008-2022), EUA prices Granger-cause grid emission intensity (bootstrap p = 0.005 at lag 1, N = 5000): a 10% EUA increase is associated with a 0.28% decline in EFᵍʳⁱᵈ the following year (R² = 0.56), consistent with merit-order fuel switching. The result is robust across three inference methods, leave-one-out year exclusion, and all reconstruction variants. No reverse causality is detected (CO₂,ind ↛ EUA, p > 0.62) and transmission to industrial emissions is weak (p = 0.078). EFᵍʳⁱᵈ should therefore be treated as predetermined but not strictly exogenous in LMDI and panel applications; the decomposed grid effect partially embeds a carbon-price response, which we quantify.