Additional stuff

Question 1

Theil U statistic

Answer 1

value < 1 means model is superior to benchmark (problem: if benchmark is same as real values -> U is infinite)

Question 2

decision makers loss functions and underlying decision problems

Answer 2

TIGHT BUT FAR FROM UNIQUE

Question 3

EMH and predictability

Answer 3

The EMH does not imply that stock returns should not be
predictable, but any predictability ought to be justified by
exposure to systematic risk factors

Question 4

R^2 upper bound?

Answer 4

yes, gamma^2sigma^2r_mkt

Question 5

Diffusion indices

Answer 5

latent factors with potential predictors (summarize the true variables), estimated by principal components

Question 6

CER

Answer 6

certainty equivalent return, utility gain from using predictive regression in place of hsitorical average foreacast (fair ptf mgmt fee)

Question 7

Campbell Thompson OOS R^2

Answer 7

measures reduction in MSPE relative to historical avg forecast. negative -> worse than BM