Transparency is a feature. This page explains what Component VaR measures and why we compute it — in plain terms. The precise formulas and numeric conventions are documented in our internal methodology; the summary here is deliberate, not an omission.
Component VaR answers a budgeting question: of the portfolio's total risk, how much does each holding contribute? It starts from a single portfolio-level Value-at-Risk — an estimate, under a Gaussian (delta-normal) assumption, of the loss magnitude the portfolio would not exceed at a chosen confidence level — and then splits that one number across the positions so the pieces add back up to the whole. The confidence level is user-selectable, with a conventional high-confidence default.
The split is done by each position's share of portfolio variance, which accounts not just for a holding's own volatility and weight but for how it co-moves with everything else it is held alongside. A large but well-diversified position can therefore carry a smaller risk share than its size alone would suggest. This allocation is exact: by construction, the components always sum to the portfolio's total VaR, whatever confidence level you pick.
Each component is reported as a positive risk magnitude — a risk budget for that holding — rather than a signed return. A position can show a negative component when it hedges the rest of the book (it moves opposite the portfolio), in which case it genuinely reduces total risk and the offset is reflected exactly. In a long-only portfolio with a well-behaved covariance input, every component is zero or positive.
