Methodology & Sources
How the energy analysis is assembled
Data sources
| Source | What | License | Refresh |
|---|---|---|---|
| FRED | WTI, Brent, Henry Hub natural gas, US retail gasoline, US average electricity, US crude production / stocks / gasoline demand | Public domain | Daily |
| EIA Open Data API v2 | PADD-level weekly retail gas, STEO monthly forecasts, international retail prices | Public domain (US gov) | Daily |
Curated events (data/energy/events_energy.csv) |
Major energy-market events with editorial sentiment scores | Editorial | On demand |
All sources are non-commercial-friendly and public domain / editorial. Attribution appears on each page that uses the data.
Pipeline
data/economy/cache/ FRED series caches (incremental)
data/energy/cache/ EIA series caches (incremental)
data/energy/events_energy.csv Hand-curated events
data/energy/ Output CSVs consumed by the site
us_prices_daily.csv
us_gas_retail_weekly.csv
us_electricity_monthly.csv
us_supply_demand.csv
padd_gasoline_weekly.csv
padd_gas_current.csv
padd_gas_10y.csv
intl_prices.csv
steo_forecast.csv
energy_summary.csv
R orchestrators:
R/fetch_fred.R— common FRED fetcher (shared with other analyses)R/fetch_eia.R— incremental EIA v2 clientR/build_energy.R— joins caches, writes output CSVs
PADD regions (used throughout)
The Energy Information Administration organizes the US into five PADD regions (Petroleum Administration for Defense Districts):
| PADD | Region | States |
|---|---|---|
| 1 | East Coast | CT, DE, DC, FL, GA, ME, MD, MA, NH, NJ, NY, NC, PA, RI, SC, VT, VA, WV |
| 2 | Midwest | IL, IN, IA, KS, KY, MI, MN, MO, NE, ND, OH, OK, SD, TN, WI |
| 3 | Gulf Coast | AL, AR, LA, MS, NM, TX |
| 4 | Rocky Mountain | CO, ID, MT, UT, WY |
| 5 | West Coast | AK, AZ, CA, HI, NV, OR, WA |
Maps on the prices and change pages color every state by its PADD’s price / price-change. State-level weekly gasoline coverage exists for only ~10 states in EIA’s public data — PADD aggregation is the honest choice until broader coverage is available.
Forecast methodology
- Short-term (STEO): pulled directly from EIA’s Short-Term Energy Outlook monthly release. Horizon ~24 months. Updated monthly ~mid-month.
- Long-term (AEO): not published on this site. AEO is scenario-based, annual, and appropriate to consume directly from eia.gov/outlooks/aeo/.
Sentiment methodology
Energy-event sentiment scores in events_energy.csv are editorial single- author estimates of the market-and-industry impact of each event, on a −1..+1 scale:
- −1.0 to −0.6: major negative shock (hurricanes destroying refining capacity, wars cutting supply, COVID demand collapse)
- −0.6 to −0.3: notable negative (sanctions, price crashes that hurt producers, regulatory shocks)
- −0.3 to 0.0: mildly negative (uncertainty, minor disruption)
- 0.0 to 0.3: mildly positive (supply discipline, capacity additions)
- 0.3 to 1.0: significantly positive (major tech shift, coordinated supply stabilization)
Each row carries a notes column explaining the anchor. Future enhancement: derive event-period sentiment from GDELT average tone for the ±30-day window around each event, using the same fetcher already cached for the Sentiment analysis.
Caveats
- Retail prices carry large tax components, especially internationally (Europe >50%, US ~20%). Raw cross-country comparisons conflate market moves with tax policy.
- EIA series availability changes. Some international endpoints have patchy coverage; the fetcher degrades gracefully when a series is unavailable.
- STEO forecasts are institutional estimates, not guarantees. EIA explicitly notes STEO is a “best estimate” subject to revision.
- 10-year change windows use the closest-to-10y observation rather than exactly 10y. Weekly data means ± a few days of slippage.
- PADD is a supply-side construct, not a demand-side one — regions reflect where refining and pipelines go, not necessarily where demand is highest or pricing most representative.
Code license
MIT — see the LICENSE file.