thus putting it in the error term. Basically, RNEMH requires that no other variable have
any explanatory power in (4).
Note that (4) can be transformed into:
' ' = α∣ + b∣( f ' ) + ηt (5)
st st
This is equivalent to (4) under the assumptions a1=b1=0, i.e. under H1. This
specification tests for existence of predictable excess returns in the forward market.
Note also that the interest differential may be expressed as:
it -it* ≡ [it -it* - (ft -st)]+ (ft - ste+l )+ (ste+l -st) (6)
where the first term on the right is the covered interest differential, the second
term is the excess of forward rate over expected future rate and the last term is the
expected depreciation. Covered interest differential may exist because of transactions
costs, existence of capital controls and risks of future capital controls, thinness of markets
and other reasons discussed above. The second term may be non-zero if the agents are
risk-averse. Taking the last term to the left hand side gives us the uncovered interest
differential, which by definition then, is the sum of the covered interest differential and
what is known as the currency or exchange risk premium, i.e.
it-it*-(ste+l-st)≡[it-it*-(ft-st)]+(ft-ste+l) (7)
When the covered interest parity doesn’t hold, then the uncovered parity will not
hold, unless the exchange risk premium for some reason is negative and exactly equal to
the covered differential, a highly unlikely scenario. This observation is particularly
relevant when testing for the parities in emerging markets.