Survey of Literature on Covered and Uncovered Interest Parities

- 12 -
transactions costs can induce relatively large hysteresis bands, i.e. a range of deviations
within which speculative activities will not occur, so that the usual linear equation tests
are misleading. Tests of CIP seem to have incorporated these transactions costs more
readily than tests of unbiasedness. Baldwin (1990) also shows that with transaction costs,
hysteresis bands would be wider for daily trading horizons than with annual trading, so
that long-horizon parity would hold better in linear tests than short-horizon, a result seen
in the literature, as discussed below. Nonlinearity based literature also explains better the
changes in betas over time and across countries [Baillie and Bollerslev, 2000 and Figure
4 below]

Unbiasedness at very short and at long horizons

Lyons and Rose (1995) and Chaboud and Wright (2005) find support for
Unbiasedness with very high frequency data. Both the studies exploit the fact that interest
is only paid on positions open overnight. Chaboud and Wright (2005) regress exchange
rate change between end of day t and beginning of day t+1 [1630 NYT and 2100 NYT]
on overnight interest differential and find that for all but yen-dollar trade, estimated
coefficients are positive and insignificantly different from 1. Lyons and Rose (1995) use
data on intra-day positions for currencies under attack (so that a depreciation is expected
and interest differentials are large) and argue that such positions don’t earn interest and so
need to be rewarded by appreciation, conditional on the expectations of depreciation not
having been realized. Moreover, they show that the larger the expected depreciation, and

<<   10   11   12   13   14   15   16   >>

More intriguing information