volatility effects. For this reason, we defer making any conclusions to the next section, where the analysis is done at a monthly frequency.
5
Predictability of Carry Trade Profits
VX _{j,t−3 }


0.0003 (0.363)
R^{2 }within Observations
0.065 846
0.164 846
0.165 846
If currencies eventually return to their fundamental equilibrium exchange rate and if volatility in foreign exchange markets portends carry trade unwinds, can these variables be used to construct a predictive model that produces positive carry trade returns, with lower risk and zero or even positive skewness (i.e. no peso problems) relative to a naïve carry trade investment strategy that solely focuses on arbitraging the interest rate differential between two countries? This section focuses on answering this question with formal outofsample predictive ability tests based on forecast loss functions that replicate conventional investment performance measures.
The basis of our analysis is expression (5), where changes in bilateral exchange rates are expressed as a linear projection of past values of exchange rates; the inflation differential; the longrun PPP equilibrium condition; the interest rate differential; and
VECMX
0.264 (0.000)

2.133
(0.008)
0.580 (0.039)

0.136
(0.000)
Notes: Sample: December 2000 to March 2009, monthly.
VAR
VECM
0.233 (0.000)
0.271 (0.000)
1.991 (0.001)
2.569 (0.000)
0.731 (0.021)
0.533 (0.061)

0.134 (0.000)
(q _{j,t−3 }
− q)
Table 3 – Estimates of
)
)
the Three Linear Models
(i
* t−3
(π
* t−3
Regressors
(e _{j }_{,t −3 }
− e_{j,t−6 }
− i_{j,t−3 }
)
−π _{j,t−3 }
18