--- id: "VOLT-HOME-WP-073" title: "Why price and carbon intensity sometimes move in opposite directions" slug: "why-price-and-carbon-intensity-sometimes-move-in-opposite-directions" description: "An observational study of the association between daily mean price and low-carbon generation share." published: "2026-08-30" cluster: "Carbon-aware household flexibility" status: "measured" evidence: "/research-data/home-papers/why-price-and-carbon-intensity-sometimes-move-in-opposite-directions.json" figure: "/research-media/home-papers/why-price-and-carbon-intensity-sometimes-move-in-opposite-directions.webp" figure_alt: "Chart for Why price and carbon intensity sometimes move in opposite directions: price–low-carbon-share association, shown as price, low-carbon share." source_ids: - "iea-electricity-2026" - "acer-retail-2025" - "iea-demand-flexibility" - "ec-retail-flexibility-2026" - "volt-licensing" peer_reviewed: false --- ## Abstract Household energy advice often assumes that low prices are a reliable stand-in for low emissions. This working paper measures a simpler relationship in the frozen Voltcast corpus: the Pearson association between daily mean day-ahead price and daily low-carbon generation share. Across 44,437 paired observations, the reported correlation is -0.14380741909028127. The negative sign is consistent with higher operational low-carbon share tending to coincide with lower prices on average, but the relationship is weak enough that the two signals can frequently disagree. The title uses the familiar phrase “carbon intensity,” while the evidence does not measure interval carbon intensity or marginal emissions. It uses low-carbon generation share as an operational proxy. That proxy is not lifecycle marginal emissions, does not fully allocate imports, and cannot identify emissions caused by a household’s incremental consumption. The result is observational, contains no causal identification, and does not measure household emissions. It explains divergence between rankings, not the emissions consequence of acting on either ranking. ## Plain-language answer Price and the recorded low-carbon-share proxy moved in opposite directions on average, but only weakly. The Pearson r of -0.14380741909028127 means that higher low-carbon generation share was associated with lower daily mean prices in this pooled sample, while leaving substantial variation unexplained. A household controller should therefore expect many moments when the cheapest option and the highest-share option do not match. There is no contradiction in that disagreement. Prices summarize the balance of bids, demand, scarcity, constraints, expectations, and the available generation stack. Low-carbon share describes the proportion of represented operational output in selected categories. These are different objects. Neither is a complete measure of the other. The environmental field must not be overread. It is not lifecycle marginal emissions and is not a measurement of the emissions added by one household. The study cannot say that choosing a lower-priced day caused emissions to fall or rise. ## Research question The registered question asks why price and carbon intensity sometimes move in opposite directions. The measured question is whether daily mean price and daily low-carbon generation share exhibit a simple linear association in paired zone-day records. The primary metric is “price–low-carbon-share association.” The paper addresses “why” only at the level of measurement concepts. A market price and an average generation-share proxy are generated by overlapping but non-identical system conditions, so perfect alignment is not expected. The empirical statistic quantifies the direction and strength of their pooled association; it does not decompose mechanisms. No fuel-price, load, congestion, weather, storage, import, or unit-commitment control is used to assign causal contributions. The relevant decision question for a household is whether price can safely substitute for an environmental objective. A weak pooled association is evidence against treating the two rankings as equivalent. It is not evidence about the causal environmental impact of a particular schedule. ## Data and provenance The paper is bound to the public evidence JSON specified in the frontmatter. Its registered source contracts are `generation_mix`, `day_ahead_prices`, `capture_stats`, `res_accuracy`, and `zone_temp_weighted`. The primary result pairs finite low-carbon-share observations from the generation table with daily mean prices by zone-day. The evidence records daily-price coverage from 2021-01-01 to 2026-08-29, a detailed-interval window from 2025-10-01 to 2026-08-29, and a long-history range from 2015-01-01 to 2026-08-29. Its publication cutoff is 2026-08-30T00:00:00Z. The extraction transaction was read-only and carried a 180-second statement timeout. The frozen snapshot SHA-256 is `f77e3ae328f93916e53b1bab7516e1d0ac740a0dbf424cd2b73c81fee2559318`. The analysis code is bound to `57c57de79cdab2b5b6d6c54c485cb5162598c5ba0b0bfe995da40d75e6c52ba9`, the protocol to `adb36bf6b447af9f96339249b8becaefc20422499cca1977242866347a97bd4b`, the paper registry to `7bcb91d7476d0a69fe9fa75a5c7782f8117e0153f82f9112b7e1d307d3943717`, and the source registry to `07949550ac443ff673fda5c0209b99f137544f3ffecf6775f109bb9d09663bd6`. There is no household consumption, bill, device, or emissions ledger in the listed sources. The external references are the exact sources frozen in the registry and are included as context, not as containers for paper-specific findings. ## Method For each available generation record, the analysis obtains the matching daily-price record for the same zone-day and computes low-carbon generation share. Rows are retained when a daily price exists and the share is finite. From each retained row, the calculation uses daily mean price and low-carbon share. Pearson correlation is then calculated across the paired vectors. This method preserves the paired observation identity for the primary association, but it remains pooled. It does not weight zones by population, demand, generation, or household count. It does not condition on season or system regime. A zone with more paired data contributes more observations. The resulting r is therefore a corpus-level descriptive statistic, not a universal structural parameter for Europe. Pearson r captures linear co-movement. A negative result means higher values of one variable tend to accompany lower values of the other in the observed panel. It does not say that every row follows that pattern, and it cannot describe nonlinear or regime-specific relationships by itself. It also has no emissions unit. Most importantly, low-carbon generation share is only an operational proxy. It is not lifecycle marginal emissions, a consumption-based intensity, or a marginal dispatch estimate. The method cannot determine the generator responding to shifted household demand and cannot establish causality. ## Results The primary result is a Pearson correlation of -0.14380741909028127 over 44,437 paired observations. The negative sign indicates an inverse average association: as the low-carbon-share proxy rises, daily mean prices tend to be lower. The magnitude is limited, so price remains an unreliable one-dimensional substitute for the proxy. This result helps explain apparent opposition between price and environmental signals. A weak inverse aggregate relationship permits many individual rows, days, and zones to depart from the average. The statistic does not report how often the variables literally move in opposite directions from one interval to the next; it summarizes cross-observation linear association. It also does not quantify the size of a price change for a change in share. The frozen evidence interpretation states: “A weak or positive relationship explains why cheapest and cleanest schedules may disagree.” In this snapshot the measured coefficient itself is weak and negative, not positive; the interpretation’s general reference to weak or positive relationships does not change that observed sign. The value should be read directly from the evidence rather than translated into a causal story. It provides no measured household-emissions result, and the proxy is not lifecycle marginal emissions. The figure labels are “price” and “low-carbon share,” with values 135.2379098544006 and 59.38346189343282, rendered as 135 and 59.4. These bars show input summaries on different scales, not the -0.14380741909028127 correlation. The evidence reports a null bootstrap interval and says the interval method is “not reported for this estimand.” ## Robustness and placebo checks The preregistered family states that inferential claims are subject to within-family Holm control and that blocked intervals should preserve serial dependence. The paper-level evidence classifies this output as descriptive and explicitly makes no unadjusted significance claim. No secondary results or paper-specific placebo outcomes are recorded. The correlation is therefore reported without a claim of statistical significance. The evidence `assumptions` array is empty. That means no paper-specific assumptions were encoded in that field; it does not make the observational design assumption-free. Potential robustness analyses would include zone-specific estimates, seasonal strata, time-block comparisons, demand weighting, rank correlation, and nonlinear fits. A placebo could break the zone-day pairing or compare dates that should not share system conditions. None of those outputs appears in the frozen evidence, so this paper does not imply that they were run or passed. The completed integrity checks are the declared cutoff, read-only extraction, and cryptographic bindings for the snapshot, code, protocol, and registry. They protect against silent data-vintage drift. They do not protect against omitted-variable bias or turn the descriptive association into a causal estimate. ## Limitations Terminology is a major limitation. The title says carbon intensity, while the empirical field is low-carbon generation share. Operational share is not lifecycle marginal emissions. It does not allocate all imports, include every lifecycle boundary, or identify marginal response. Readers should not treat those concepts as synonyms. Daily aggregation hides intraday co-movement. A day with a low mean price may contain expensive intervals; a day with a high average low-carbon share may contain household-availability windows with a different mix. The calculation is not a charging or battery schedule and has no device constraints. Pooling can mask heterogeneity by zone, season, technology mix, and market regime. The result contains no controls that would distinguish demand, fuel prices, weather, network constraints, or policy conditions. Correlation cannot establish direction, mechanism, or causality. The study includes no retail tariff, tax, network charge, supplier margin, or actual customer consumption. It is not a household bill study. It reports no measured household emissions and offers no trading strategy or trading advice. ## Practical implication Price-only automation should be labelled price optimization. A user should not be told that it is automatically carbon optimization. If a product presents both objectives, it should expose the source, resolution, and accounting boundary of the environmental signal and explain when objectives disagree. The result can motivate a multi-objective schedule, but it cannot choose the weight between objectives. That is a user preference and method-design question. Nor can the result justify claims about caused or avoided emissions. Such claims would require an interval-appropriate emissions model and an explicit counterfactual for incremental load. Low-carbon share may still be a useful operational context signal. Its proper label is essential: it is an average generation proxy, not lifecycle marginal emissions. Household constraints and retail costs would need a separate study before any bill-facing recommendation. ## Reproducibility Retrieve the evidence JSON and verify identifier `VOLT-HOME-WP-073`, slug, status, metric, source contracts, publication cutoff, and all five provenance hashes. Rebuild the finite paired zone-day panel from daily mean prices and low-carbon generation share using the frozen snapshot and code identity. Pearson correlation across those paired vectors should reproduce -0.14380741909028127 with 44,437 observations. Changing to interval data, a rank statistic, a weighted estimator, a per-zone model, or a lifecycle or marginal-emissions field would answer a different question. Such a variant should be published as new evidence and should not be substituted into this result. The public figure visualizes price and low-carbon share at the aggregate level. Licensing and redistribution of underlying records remain governed by the registered Voltcast licensing source. ## Disclosure Analysis and drafting were model-assisted. This working paper is not peer reviewed, is not a household bill study, makes no causal claim, and is not trading advice. Volt has no live traders or live capital. It does not report measured household emissions. Low-carbon generation share is an operational proxy, not lifecycle marginal emissions. ## References - [International Energy Agency — Electricity 2026](https://www.iea.org/reports/electricity-2026) - [ACER and CEER — Rewarding flexibility: How retail contract choice can help unlock consumer flexibility](https://www.ceer.eu/wp-content/uploads/2025/11/ACER-CEER-2025-Retail-monitoring.pdf) - [International Energy Agency — Scaling Up Demand Flexibility](https://www.iea.org/reports/scaling-up-demand-flexibility) - [European Commission — Communication from the Commission on the Citizens Energy Package](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A52026DC0115) - [Voltcast — Voltcast data licensing and redistribution](https://github.com/ossedk/voltcast/blob/main/docs/voltcast/LICENSING.md)