to lie within the bounds implied by the existing transactions costs, thus validating the
covered interest parity relationship. How much the latter differential can be in an efficient
market has been the subject of several asset pricing models (e.g. Lucas, 1982; Frenkel
and Razin, 1980; Svensson, 1985 and Stulz, 1984), portfolio balance models (Branson
and Henderson, 1985; Frankel, 1983, 1984) and currency crises models (Arias, 2001).
The presence of an expected variable, unobserved to the econometrician, makes it
hard to test for UIP directly. It has therefore been tested empirically by assuming rational
expectations and/or CIP. Taking logs of (2) and imposing rational expectations, we get:
Et(st+1 -st)≈it -it*
where the small case letters denote logs. The above relation can be expressed as
the null H0 : α = 0, β = 1 in the equation:
st+1 -st =α+β(it -it*)+εt (3)
If one additionally assumes that CIP holds, then a test of H1 : δ = 0, ρ = 1 in the
following equation is equivalent to the test in (3)3:
st+1 -st =δ+ρ(ft -st)+ξt (4)
Both the above specifications have been used to test the validity of UIP
assumption, although it must be remembered that neither is a direct test, and (4) involves,
in addition to the RE assumption, the assumption of no risk premium in forward rate.
H1 : δ = 0, ρ = 1 is therefore referred to as the Risk Neutral Efficient Market
Hypothesis [RNEMH]. A weaker version is the unbiasedness hypothesis, which allows a
risk premium in (4) but constrains it to be uncorrelated with information set at time t.,
3 Taking logs of (1) yields ft - st ≈ it - it*