Compound Annual Growth Rate (CAGR)
How we annualize price changes across 1, 3, and 5 year windows.
Formula
CAGR is the geometric mean annualized return. Given a start median V_start at time t_start and an end median V_end at time t_end, CAGR = (V_end / V_start)^(1 / years) − 1, where years is the actual time span between the two anchor dates.
cagr = (V_end / V_start) ** (1 / actual_years) - 1
Anchor windows
We look for an aggregation centered approximately 1, 3, or 5 years prior to the latest period_end, within a ±180 day search window. If no aggregation exists in that window, we fall back to the oldest aggregation we have and compute against the actual elapsed years — this is how the 5y figure can use only 3.4 years of data on a younger variant.
Why geometric mean, not arithmetic
Arithmetic average of yearly returns overstates long-run returns when there is volatility. CAGR uses the geometric mean, which is the only annualization consistent with compounding. A bag that goes up 50% then down 50% has an arithmetic average of 0% but a CAGR of −13.4% — the latter is the truth.
Confidence
CAGR confidence is the lower of the two anchor aggregations' confidence labels. If either is low, the CAGR is low confidence too.
Other topics
- Median, percentiles and IQR · Why we publish median and percentiles instead of mean prices, and how to read th…
- Volume weighting and index construction · How components contribute to a Bagonomics index: square-root weights, caps, and …
- Premium to retail · Definition of premium-to-retail, the country-of-retail convention, and how to in…
- Confidence labels · What 'low', 'medium', 'high' confidence on an aggregation means and when to use …