IESET.
Hypotheses·energy·eu_gas_storage_winter_2022_2023_demand_destruction_vs_substitution

The EU's headline ~15% reduction in natural-gas demand over the August 2022 - March 2023 emergency window (vs the 2017-2021 average) was achieved through a measurable mix of (a) industrial demand destruction (rationing, plant mothballing, output cuts), (b) household + commercial conservation (thermostat reductions, voluntary cuts), (c) electricity- sector fuel-switching (coal restart, nuclear extension where available), and (d) unusually warm winter weather.

The decomposition is a contested claim because the policy-feasibility implications differ sharply across channels: weather-driven savings tell us little about policy capacity; industrial demand destruction is mostly capacity loss (output cost), not adjustment; conservation may be sustainable; coal switching is a climate-perverse temporary regression. This hypothesis pre-registers a quantitative decomposition.

INCONCLUSIVEengine/runs/eu_gas_storage_winter_2022_2023_demand_destruction_vs_substitution

INCONCLUSIVE_DATA_PENDING — treatment 'weather_attributable_savings_share' has no within-country variation under country fixed effects

confidence cueResult card produced; verdict unclassified.

policy briefCoverage too thin

In ordinary language

In plain terms, this asks whether post aug 2022 gas emergency dummy is actually linked to better or worse log total gas consumption from 2017 to 2024.

plain answer

This test cannot make a firm call yet. treatment 'weather_attributable_savings_share' has no within-country variation under country fixed effects

why it matters

This matters because energy claims should change belief only when they survive a pre-declared empirical test.

how the test works

It compares 13 country or place units from 2017 to 2024, using a panel fe decomposition design, with fixed effects for country and month of year.

what was measured
What changed
  • Post aug 2022 gas emergency dummy
Possible pathway
  • Weather attributable savings share
  • Industrial destruction share
What we checked
  • Log total gas consumption
  • Log industrial gas consumption
  • Log residential commercial gas consumption
what this does not prove

A single test is not the whole truth. It narrows the claim under a specific sample, time period, and method. Strong policy conclusions need the pattern to survive nearby tests, alternative data, and serious objections.

verification

No evidence packet has been generated yet.

Results

engine/runs/eu_gas_storage_winter_2022_2023_demand_destruction_vs_substitution
1007550250201720212024DEUFRAITAESPNLDBELPOL
illustrative sketch · run pending
No coefficients yet. When the model fires, this chart will show log_total_gas_consumption across 13 sampled countries over 2017–2024.
The shapes above are stylised — none of the lines are real data.
Placeholder for eu_gas_storage_winter_2022_2023_demand_destruction_vs_substitution. Published chart will be generated from engine/runs/eu_gas_storage_winter_2022_2023_demand_destruction_vs_substitution/chart_data.json.

Pre-registration

pre-registered
first-spec commit 098ce96 · 2026-04-30T12:57:33Z
run generated · 2026-06-29T17:52:06Z

The EU's headline ~15% reduction in natural-gas demand over the August 2022 - March 2023 emergency window (vs the 2017-2021 average) was achieved through a measurable mix of (a) industrial demand destruction (rationing, plant mothballing, output cuts), (b) household + commercial conservation (thermostat reductions, voluntary cuts), (c) electricity- sector fuel-switching (coal restart, nuclear extension where available), and (d) unusually warm winter weather. The decomposition is a contested claim because the policy-feasibility implications differ sharply across channels: weather-driven savings tell us little about policy capacity; industrial demand destruction is mostly capacity loss (output cost), not adjustment; conservation may be sustainable; coal switching is a climate-perverse temporary regression. This hypothesis pre-registers a quantitative decomposition.

Falsification criterion — what would disprove this

set before the run · honoured after

This hypothesis is considered falsified if:

Not supported if (a) the channel decomposition cannot bound any channel's share away from zero with 80% confidence (decomposition underidentified), OR (b) weather attribution exceeds 60% of the total cut (would imply the policy + behavioural response was minor), OR (c) industrial-destruction share is statistically zero (would refute the capacity-loss interpretation), OR (d) the second-winter cut is entirely weather-explained with no remaining policy/behavioural component (would suggest first-winter response was not durable).

formal test & threshold
test:      gas_demand_decomposition_panel
threshold: All four channels each contribute ≥ 5pp to the headline 15% cut at EU aggregate AND industrial-destruction share is statistically > 0 at p<0.10 AND second-winter cut retains > 5pp of non-weather component.

Method

Template
panel_fe_decomposition
Fixed effects
country, month_of_year
Clustering
country
Sample
13 countries · 2017 – 2024
Evidence type
causal

Primary specification: weather-normalise consumption (regress on HDD at country-month level), then decompose residual gas-demand change Aug 2022 - Mar 2023 (and the second winter Aug 2023 - Mar 2024) into industrial / power / residential channels. Power-sector decomposition uses ENTSO-E generation mix to identify coal-substitution share. Industrial-destruction decomposition: gas use change minus output- elasticity-implied change (calibrated from 2017-2021 panel) = structural cut. Cross-country: estimate channel shares per country, test whether higher-gas-exposure countries (DEU, ITA, NLD) show larger industrial destruction shares than lower-exposure countries (FRA, ESP). Second-winter test (Aug 2023 - Mar 2024) distinguishes weather (the 2022-2023 winter was warm; 2023-2024 also mild) from other channels. If the second-winter cut is similar in size and channel mix, weather is doing less than headline; if it shrinks substantially, weather was a large share. Known limitations: (1) Weather normalisation is sensitive to HDD specification (linear vs threshold-based); report sensitivity. (2) Industrial-destruction vs efficiency split requires output data (IPI / sectoral VA) on the same monthly panel; OECD STAN does not ship monthly. Use proxy IPI for monthly, validate against annual STAN. (3) Hidden-flow problem: gas imports into EU may have been embedded in non-gas imports (ammonia, hot-rolled steel, aluminium); the true demand-destruction figure is overstated by the volume of imported energy-intensive intermediates. (4) AGSI+ storage data has lag and revisions for some smaller member states.

Data

VariableSourceTransform
log_total_gas_consumption
outcome
eurostat:nrg_cb_gasmtier 1
log
log_industrial_gas_consumption
outcome
eurostat:nrg_cb_gasmtier 1
log
log_residential_commercial_gas_consumption
outcome
eurostat:nrg_cb_gasmtier 1
log
log_power_sector_gas_consumption
outcome
eurostat:nrg_cb_gasmtier 1
log
heating_degree_days
outcome
eurostat:nrg_chdd_mtier 1
level
log_coal_generation
outcome
constructed:ENTSO-E aggregated thermal generation by fuel — coal/lignite. entsoe fetcher pending; Eurostat nrg_cb_em as fallback.tier 5
log
log_industrial_production_index
outcome
oecd:STSIND4_PRODMANtier 2
log
post_aug_2022_gas_emergency_dummy
treatment
constructed:indicator = 1 from August 2022 (EU 15% gas-demand reduction regulation effective) through March 2023; 0 otherwise. Secontier 5
indicator
weather_attributable_savings_share
channel
constructed:counterfactual gas consumption holding HDD at 2017-2021 monthly mean; difference = weather component.tier 5
level
industrial_destruction_share
channel
constructed:industrial gas use change net of industrial-output-weighted change (proxy: IPI or sectoral VA). Residual after output-adtier 5
level
power_fuel_switching_share
channel
constructed:change in power-sector gas use vs. counterfactual at constant power-mix; residual after coal/nuclear/renewables substitutier 5
level
residential_conservation_share
channel
constructed:change in residential gas use after weather-normalisation (HDD-adjusted).tier 5
level
log_ttf_gas_price
control
constructed:Dutch TTF spot, monthly. eex fetcher or manual-drop pending.tier 5
log
log_eu_ets_price
control
eea:eu_ets_verified_emissionstier 2
log
gas_storage_filling_level
control
constructed:AGSI+ EU gas-storage filling level by country, monthly. Manual-drop pending.tier 5
level

● ready  · ● pending  · ● reconstruct-needed

Detailed result card

Result card — eu_gas_storage_winter_2022_2023_demand_destruction_vs_substitution

Verdict: INCONCLUSIVE_DATA_PENDING — treatment 'weather_attributable_savings_share' has no within-country variation under country fixed effects

Pre-registration

  • Claim: The EU's headline ~15% reduction in natural-gas demand over the August 2022 - March 2023 emergency window (vs the 2017-2021 average) was achieved through a measurable mix of (a) industrial demand destruction (rationing, plant mothballing, output cuts), (b) household + commercial conservation (thermostat reductions, voluntary cuts), (c) electricity- sector fuel-switching (coal restart, nuclear extension where available), and (d) unusually warm winter weather. The decomposition is a contested claim because the policy-feasibility implications differ sharply across channels: weather-driven savings tell us little about policy capacity; industrial demand destruction is mostly capacity loss (output cost), not adjustment; conservation may be sustainable; coal switching is a climate-perverse temporary regression. This hypothesis pre-registers a quantitative decomposition.
  • Falsification rule: Not supported if (a) the channel decomposition cannot bound any channel's share away from zero with 80% confidence (decomposition underidentified), OR (b) weather attribution exceeds 60% of the total cut (would imply the policy + behavioural response was minor), OR (c) industrial-destruction share is statistically zero (would refute the capacity-loss interpretation), OR (d) the second-winter cut is entirely weather-explained with no remaining policy/behavioural component (would suggest first-winter response was not durable).
  • Falsification test: gas_demand_decomposition_panel

Estimate

  • Error: treatment 'weather_attributable_savings_share' has no within-country variation under country fixed effects

Variables resolved

  • eurostat:nrg_cb_gasm → log_total_gas_consumption (outcome, publisher=eurostat, n=652)
  • eurostat:nrg_cb_gasm → log_industrial_gas_consumption (outcome, publisher=eurostat, n=652)
  • eurostat:nrg_cb_gasm → log_residential_commercial_gas_consumption (outcome, publisher=eurostat, n=652)
  • eurostat:nrg_cb_gasm → log_power_sector_gas_consumption (outcome, publisher=eurostat, n=652)
  • eurostat:nrg_chdd_m → heating_degree_days (outcome, publisher=eurostat, n=1426)
  • constructed: counterfactual gas consumption holding HDD at 2017-2021 monthly mean; difference = weather component. → weather_attributable_savings_share (decomposition_channels, publisher=constructed, n=104)
  • constructed: change in power-sector gas use vs. counterfactual at constant power-mix; residual after coal/nuclear/renewables substitution accounting. → power_fuel_switching_share (decomposition_channels, publisher=constructed, n=104)

Variables missing data

  • constructed: ENTSO-E aggregated thermal generation by fuel — coal/lignite. entsoe fetcher pending; Eurostat nrg_cb_em as fallback. (outcome, name=log_coal_generation) — vintage not on disk
  • oecd:STSIND4_PRODMAN (outcome, name=log_industrial_production_index) — vintage not on disk
  • constructed: indicator = 1 from August 2022 (EU 15% gas-demand reduction regulation effective) through March 2023; 0 otherwise. Secondary: 1 through March 2024 for second-winter test. (treatment, name=post_aug_2022_gas_emergency_dummy) — vintage not on disk
  • constructed: industrial gas use change net of industrial-output-weighted change (proxy: IPI or sectoral VA). Residual after output-adjusted gas decline = efficiency/destruction component split. (decomposition_channels, name=industrial_destruction_share) — vintage not on disk
  • constructed: change in residential gas use after weather-normalisation (HDD-adjusted). (decomposition_channels, name=residential_conservation_share) — vintage not on disk
  • constructed: Dutch TTF spot, monthly. eex fetcher or manual-drop pending. (controls, name=log_ttf_gas_price) — vintage not on disk
  • eea:eu_ets_verified_emissions (controls, name=log_eu_ets_price) — vintage not on disk
  • constructed: AGSI+ EU gas-storage filling level by country, monthly. Manual-drop pending. (controls, name=gas_storage_filling_level) — vintage not on disk

Generated by scripts/run_panel_fe.py at 2026-06-29T17:52:06+00:00

Strongest opposing argument

Every hypothesis ships with its charitable opposing argument. The framework earns credibility by handling objections at their strongest, not weakest.

Notes

Data readiness: - Eurostat nrg_cb_gasm + nrg_chdd_m: ready - OECD industrial production: ready - ENTSO-E generation mix: entsoe fetcher pending; Eurostat nrg_cb_em as fallback - TTF gas price: eex fetcher pending or manual-drop - EU ETS price: eea ready - AGSI+ storage levels: manual-drop pending under data/manual/derived/ Run when ENTSO-E fetcher + TTF + AGSI+ manual-drops complete.

Authored framework. Read the transparency note.