An end-to-end Python data pipeline — FRED API extraction, pandas ETL, statistical trend detection, recession early-warning signals, and AI-generated macro briefings. Scroll to explore the data.
Click an indicator. Move your cursor over the chart.
A 0–100 composite gauge synthesising all 7 weighted indicators — inflation, payrolls, and GDP scored on their year-over-year rate of change rather than their raw level, normalised against 10 years of historical range, direction-adjusted, and colour-banded into four health states. One number that captures the macro picture.
6-month ARIMA(1,1,1) predictions across every indicator, with 80% and 95% confidence interval bands. Fitted via statsmodels with AIC reported. Turns a retrospective view into a forward-looking one.
Claudia Sahm's recession early-warning indicator: it triggers when the 3-month average unemployment rate rises 0.5pp or more above its low over the previous 12 months. It has signalled every US recession since 1970, and on today's revised data it also crossed the line twice with no recession following, in mid-2003 and mid-2024.
On each data refresh, the Claude API generates a 3-sentence plain-English macro briefing citing the actual numbers — so the dashboard explains itself. Optionally enabled via an Anthropic API key.
Linear regression over a rolling 6-month window — flagged only when statistically significant (p < 0.10, R² ≥ 0.30). Suppresses noise so signals actually mean something. YoY expressed as pp-change for rates, % for levels.
Python 3.11+. A free FRED API key from the St. Louis Fed. Anthropic key is optional — it enables the AI briefing panel.
# 1. Clone the repo git clone https://github.com/tarang-tj/economic-pulse-dashboard.git cd economic-pulse-dashboard # 2. Install (statsmodels + anthropic included) pip install -r requirements.txt # 3. Get free FRED key at fred.stlouisfed.org # 4. Add your keys cp .env.example .env # FRED_API_KEY=... (required) # ANTHROPIC_API_KEY=... (optional) # 5. Run python pipeline.py && python app.py