The solution to the forward-bias puzzle

On the (in)feasibility of covered interest parity as a solution to the forward bias puzzle

“…For x r , ı 1 = 0.932, which is very close to one despite the fact that y and x r are entirely independent of each other. For x − , we see that ı 1 = 0.748, which is now slightly smaller but again fairly close to one and comparable to Pippenger's (2011) average coefficient of 0.845. Note, however, that this positive result arises despite the fact that the underlying relationship between y and x − is actually negative.…”

“…(5) reveals that the essence of the CIP-based approach to solving the forward bias puzzle, abstracting from transaction costs (which are quite miniscule in practice), is to regress s t+1 on (f t − s t ) and [ s t+1 − (f t − s t )]. Thus, it is no wonder that the latter term explains the variation not explained by the former and that, overall, Pippenger's (2011) reported R 2 coefficients are all extremely close to one. Moreover, since Var t s t+1 ≈ Var t ( s t+1 + e t+1 ), the claim that the CIP-based approach also solves the closely related variance puzzle is also unfounded.…”

“…In this section, we extend Pippenger's (2011) logic to show that the CIP-based approach is tautological in nature and that corresponding empirical results follow mechanically from this fact.…”

“…More evidence from simulations is provided in Section 4. 8 As a reminder, and not to belabor the point, the term [(y + ε) − x] is analogous to the term [( st+1 + et+1) − (ft − st)], which equals (ft+1 − ft ) − (it+1 − i * t+1 ), which Pippenger (2011) in turn treats as two separate terms. Thus, our general specification, which has one less term than Pippenger's, is nevertheless entirely applicable.…”

A comment on: “The solution to the forward-bias puzzle”

“…In a recent article in this journal, Pippenger (2011) presents a solution to the longstanding forwardbias puzzle. This puzzle relates to the prediction that, in efficient financial markets with rational agents, the current forward premium (the difference between the natural logs of the forward, f t , and spot, s t , exchange rates) should be an unbiased predictor of the actual future change in the spot rate:…”