Background
Issue #61 covers the ex-post Tracking Error for Portfolio — the realized TE computed from the historical time series of portfolio vs benchmark returns.
This issue is about the ex-ante (forward-looking, planned) Tracking Error, as defined in Hwang & Satchell, Tracking Error: Ex-Ante versus Ex-Post Measures (2001), eq. (1):
TE_ex-ante = sqrt((a − b)' Σ (a − b))
where:
a — portfolio weights
b — benchmark weights
Σ — covariance matrix of asset returns
Scope
- Unlike the ex-post measure, ex-ante TE requires benchmark weights, not just benchmark returns. The benchmark must be expressed in the same asset universe as the portfolio (e.g. another
Portfolio object over a shared set of symbols).
- First step: estimate
Σ as the sample covariance matrix of historical monthly returns (annualized). Advanced estimators (shrinkage, EWMA) are out of scope.
- Per Theorem 1 of the paper, ex-ante TE systematically underestimates ex-post TE because portfolio weights are stochastic ex-post (they drift between rebalancings). This should be documented in the docstring.
Related
Background
Issue #61 covers the ex-post Tracking Error for
Portfolio— the realized TE computed from the historical time series of portfolio vs benchmark returns.This issue is about the ex-ante (forward-looking, planned) Tracking Error, as defined in Hwang & Satchell, Tracking Error: Ex-Ante versus Ex-Post Measures (2001), eq. (1):
where:
a— portfolio weightsb— benchmark weightsΣ— covariance matrix of asset returnsScope
Portfolioobject over a shared set of symbols).Σas the sample covariance matrix of historical monthly returns (annualized). Advanced estimators (shrinkage, EWMA) are out of scope.Related
Portfolio