When geopolitical crises hit, oil markets move fast. This project tracks 30 years of major shocks to answer one question: how bad does it get, and how quickly?
Built during the active 2026 US-Israel conflict with Iran and the first-ever closure of the Strait of Hormuz, this project combines Python-based event study analysis with Tableau dashboards to compare historical price responses, model forward-looking scenarios, map which countries face the greatest exposure, and track what it all means at the pump for American consumers.
The full data lifecycle was covered: sourcing and cleaning raw EIA, World Bank and FRED data across five Jupyter notebooks, event study methodology, scenario modeling calibrated from historical events, inflation adjustment, lag correlation analysis, and interactive Tableau dashboards published to Tableau Public.
Daily Brent and WTI spot prices were sourced from the US Energy Information Administration, with TTF natural gas futures from Investing.com and country inflation data from the World Bank. Data was cleaned and merged using Pandas, with weekend gaps forward-filled to create a continuous daily series.
An event study framework normalized prices to 100 at each event date, enabling direct comparison of shock magnitude and speed across different price environments. Metrics including peak percentage, days to peak and days to recovery were computed for all 8 events.
Scenario projections for the Iran 2026 conflict were modeled using exponential decay curves calibrated from historical events rather than assumed speeds: Scenario A from 9/11 (3 days), Scenario B from Libya 2011 (74 days), Scenario C from Ukraine 2022 (12 days). Actual price data was added alongside the projections as it became available.
Country vulnerability scores were calculated as composite indices of oil exposure, LNG exposure and availability of pipeline alternatives. Inflation correlation analysis using World Bank CPI data identified which countries most directly pass oil price shocks to consumers.
A fifth notebook analyzed US retail gasoline prices in depth, adjusting 30 years of weekly pump prices to February 2026 dollars using monthly CPI data from FRED. Lag correlation analysis confirmed that Brent price changes take approximately one week to show up at the pump. Per-president panel charts break the full history into presidential terms for lay audience readability.
All analysis was built in Jupyter Notebooks and visualized in Tableau Public.
Building this project during an active geopolitical event added a dimension that tutorial projects rarely have. The data kept updating and the analysis had to keep pace. That pressure sharpened both the methodology and the storytelling.
The scenario modeling approach, calibrating speeds from historical analogues rather than assuming parameters, is something I would apply to other forecasting problems. The key insight is that the historical record already contains the answer to "how fast does this kind of thing move." You just have to look.
Next steps would include automating the data refresh pipeline via Google Sheets so Tableau Public can update on a schedule, expanding the country vulnerability model to include forex reserve buffers and subsidy capacity, and adding a correlation view showing which scenario the current price trajectory most closely resembles in real time.