- 15 -
(2005) however find that the same cannot be said when data of 3 month horizon is used.
Even if one accepts that higher estimated betas imply that UIP held better in the 1990’s
one if still left with explaining why the same should be the case for the 1970’s. Given the
evidence reviewed above, the case for estimates from linear equations being biased
because of non-linearities introduced by frictions is strong, and an adequate answer to the
question of whether or not deviations from UIP have reduced over time needs to take
these into account while testing for UIP in samples with different time periods. I am not
aware of any such study, even for the industrialized countries.
Unbiasedness in Emerging markets
Chinn and Frankel (1994) was an early study on UIP in emerging markets and
used survey data to test the condition for East Asian economies.
Flood and Rose (2002) use daily data on interest differentials for 10 emerging markets for
the 1990’s and find that for four out of the ten, the coefficients are non-negative. Their
study however, focuses exclusively on the 1990’s and on a few emerging markets that
suffered a crises in that period, and the question they are trying to answer is whether
during crises the UIP holds better than otherwise. Both Flood and Rose (2002) and
Bansal and Dahlquist (2000) find that unbiasedness holds better in times of high inflation
and inflation volatility, perhaps because the exchange rate depreciations are more easily
forecast in such periods. The latter also find that it holds better for lower income
countries6, a finding not corroborated in Flood and Rose (2002). Frankel and Poonawala
6 Why this should be so, is a question left largely unanswered.