document that it is the country-specific attributes and not the systematic risks that matter.
Suspecting that the limited success of the previous result in relating excess currency
returns to systematic risks may be due to the choice of a single-factor model, Francis et al.
(2002) employ a three-factor Fama-French model (Fama and French, 1993) for nine emerging
markets for the period 1980-2000. They incorporate size and value factors and note that
these factors may reflect financial risks and future growth opportunities for an economy,
as suggested by Liew and Vassalou (2000). Francis et al. document that excess currency
returns over the UIP-implied level are significantly driven by systematic risks. Using a
multivariate GARCH framework, they report that excess currency returns for emerging
markets can be attributable to time-varying risk premia.28
Tai (2003) employs an international CAPM model assuming away PPP for four East
Asian countries during January 1986-July 1998. The results reported by Tai (2003) are in
line with those of Francis et al. (2002), i.e. excess currency returns are driven by systematic
risk factors. In particular, Tai explores whether excess returns from forward exchange rate
contracts are due to time-varying risk premia. The results show that systematic risk factors
are significant in explaining excess currency returns for the countries in the sample. Hence,
Tai argues that deviations from the UIP condition are attributable to the existence of
time-varying risk premia especially in the form of currency risks.
To sum up, there exists significant risk premia for emerging market assets. Moreover,
through decomposing the risk premia into currency, political and default risks, empirical
studies cited in this section report that the latter risks, which are generally negligible for
developed economies, are significant for emerging market assets. Similarly, studies that
employ variants of the CAPM suggest that risk premia are indeed offered by these country
assets. Accordingly, ignoring the existence of these risks while estimating the UIP condition
through equations (3a) or (3b) induces an omitted variable bias.
24