How should weather enter a natural-gas demand comparison?

What are heating degree-days measuring?

Heating degree-days summarize how far a day's average temperature falls below a chosen base. Cooling degree-days summarize the opposite side. EIA's explainer uses a 65-degree Fahrenheit base in its examples. State the base and temperature scale explicitly: changing either changes the statistic. A degree-day is a temperature accumulation, not a gas volume or a measure of price pressure.

For an invented two-day example, daily averages of 45 and 55 degrees Fahrenheit generate 20 and ten heating degree-days against a 65-degree base, totaling 30. A warmer day above the base contributes zero heating degree-days under this convention, not a negative amount that cancels previous cold. Preserve the daily aggregation rule before comparing weekly or monthly totals.

Match the weather geography to the demand question

A national temperature average can be a poor description of heating conditions where gas-consuming households are concentrated. Population weighting and gas-customer weighting answer different questions, and both can change through time. Record the source geography, weighting vintage and observation period. A matched country label alone does not establish that the weather index represents the relevant demand footprint.

Power-sector gas use also depends on electricity demand, generation availability and competing fuels. Cooling conditions can matter without explaining every movement in gas burn. Industrial demand, exports and pipeline supply constraints require separate evidence. Avoid treating a degree-day adjustment as a complete reconciliation of total gas demand simply because temperature is important for one component.

Inspect a deliberately assumed weekly adjustment

Consider two synthetic seven-day weeks with heating degree-day totals of 100 and 140. Assume, solely for arithmetic illustration, non-weather demand of 210 Bcf per week and an additional 0.1 Bcf for every heating degree-day. The resulting demand proxies are 220 and 224 Bcf. The four-Bcf difference follows from an assumed coefficient, not an estimated historical relationship.

Daily averages are approximately 31.428571 and 32 Bcf, respectively. Do not multiply the weekly degree-day total by seven again: the accumulation already covers the week. These figures are not actual consumption observations, fitted parameters or predictions. The exercise makes the unit chain inspectable and shows why an apparently modest scaling mistake can overwhelm the weather contribution being discussed.

Do not confuse weather normalization with fitting a model

A weather-normalized comparison asks what demand would be under a declared weather reference and relationship. Establishing that relationship requires evidence, specification choices and validation beyond this guide. An assumed sensitivity is suitable for a transparent scenario but cannot be relabeled a calibrated adjustment. The EIA weather-sensitivity supplement provides context for these issues, not a transferable coefficient for every location or era.

Specify whether the reference is a particular historical weather period, a published normal or an invented scenario. Normal weather does not mean guaranteed future weather. Also distinguish a revised historical weather series from the weather observations or forecasts available at an old decision date. A current download cannot establish that an earlier researcher knew the same weekly values.

Connect demand context to storage carefully

A larger withdrawal during a colder week can be consistent with higher heating demand, but storage also responds to production, imports, exports and other consumption. Reconcile the complete supply-demand boundary before assigning the withdrawal difference entirely to weather. Regional storage deliverability and transport access can matter even when the national volume appears adequate.

Compare storage levels and seasonal deviations separately from the weather adjustment. A low absolute stock level and a positive deviation from a chosen baseline can coexist. Neither state provides a direct price forecast. Keep the weather evidence, storage observations and balance assumptions in distinct columns so a reader can challenge one component without losing the others.

Which questions make a weather comparison useful?

Ask which geography and customer mix the index represents, what temperature base is used, and whether the period contains the same number of days. Ask whether the proposed coefficient is estimated, externally documented or simply assumed. Then identify non-weather changes that could explain the residual. The answer should preserve uncertainty rather than assigning every unexplained quantity to structural demand.

Failure modes include mixing Celsius with Fahrenheit degree-days, applying weekly totals as daily values, double-counting weather in a seasonal baseline and treating a forecast weather path as observed data. A useful note retains all units, the weighting method and the adjustment's exact status. No demand model is fitted here and no actual instrument research, price recommendation or production process is introduced.

Keep in the decision record

• Degree-days measure temperature accumulation, not physical gas demand.
• Keep geographic weighting and period length with the statistic.
• An assumed sensitivity is a scenario, not a fitted weather-normalization model.

Source notes

1. EIA - Degree-days, units and calculation examples; public educational page reviewed
2. EIA - Weather Sensitivity in Natural Gas Markets, March 2014 STEO supplement; public full PDF reviewed

Source dates and scope matter. Historical inventory observations are retained snapshots; forecasts and synthetic curve assumptions are labelled separately. No investment outcome or product coverage is implied.

Continue the analysis