- 16 -
(2004) examine the unbiasedness hypothesis using data on forward rates for the period
1996-2004, for 14 emerging markets and find eight positive coefficients, although these
are imprecisely estimated. However, the specification in equation (4) is a test of
unbiasedness only on the assumption that covered interest parity holds. This assumption
is very likely to be violated for emerging markets, for reasons discussed in Section II.
Francis, Hasan and Hunter (2002) investigate the impact that emerging market
liberalization has on the time-varying risk premium demanded by US investors. They try
to link each liberalization episode with subsequent movements in excess returns in
currency deposits and find that much of the excess returns are due to and compensation
for bearing systematic (non-diversifiable) risk, a claim disputed in Bansal and Dahlquist
(2000). . Their data is averaged monthly deposit rates from the IFS. Since much of the
arbitrage in emerging markets is undertaken by financial institutions, not individual
investors, one needs to look not at deposit rates (which, as given in IFS are the average
monthly rates on deposits offered to resident customers) but at rates at which financial
institutions (including non-resident financial institutions) can lend and borrow between
dates A and B and the depreciation between those two dates.
When discussing unbiasedness in emerging markets, consideration has to be taken
of the fact that most of these have had fixed exchange rates over at least part of the last
20 year period. If the fixed rate is perfectly credible, the expected exchange rate would
equal the actual rate and the home and foreign rates would be equal. Arias(2001)
summarizes how the UIP relationship is modified in models where the fixed rate is less
than credible, hence susceptible to an attack (most of these modifications are in the spirit
of the portfolio balance models discussed above). The interest rate in this case would