Skip to content
Time-Series
Econometrics
Ian Helfrich PhD ↗

Lecture 12 summary · April 9, 2027

Predictable risk with an unpredictable return

Complete chapter · Laboratory ch12 · Practice · All summaries

The economic question

A zero predictable mean can coexist with predictable risk. A variance recursion does not establish mean-return predictability, and a tail quantile needs a distribution.

Central relation

ht+1=ω+αut2+βht,h_{t+1}=\omega+\alpha u_t^2+\beta h_t,
h‾=ω1−α−β\overline h=\frac{\omega}{1-\alpha-\beta}

Individual problem

Omega is 0.15, alpha 0.1, beta 0.75, current variance two, and latest surprise negative three. Find next and long-run variance, then convert percentage-point returns to decimals.

Worked reasoning

Next variance is 2.55 and long-run variance one. Dividing returns by one hundred divides variances and omega by ten thousand. Alpha and beta are dimensionless and unchanged.

A consequential failure

Variance predictability is called mean predictability, or return units change without the variance intercept.

Repair

Separate mean and variance, enforce the domain, check exact unit invariance, and declare the innovation distribution for a risk quantile.