| Resumo: | Predicting the future value of an exchange rate has been a long-standing challenge in economics. There is still no evidence of any model or technique that has consistently been proven to beat the random walk model. The current objective of this thesis is to check if there is a liquidity channel tied to banking funding that allows us to explain some part of the performance of currency returns. The present analysis focuses on the paper “Risk Appetite and Exchange Rates” by Adrian et al. (2015) where it is claimed that there is a statistically significant relationship between banks’ funding capacities and changes in exchange rates. This relation seems to be more prominent for currencies of more developed countries. In my analysis, the liquidity aggregates (Commercial Paper and Repo) also display some explanatory power, though less than in Adrian et al. Importantly, however, I show that using linear time de-trending as the authors do presents stationarity problems for both liquidity aggregates, especially for Repo volume. The statistical inference of the OLS results is therefore limited. Moreover, in the fitted models, adding a dummy variable and a dummy variable with interactions with the two liquidity aggregates, as in Adrian et al. (2015), reduces the individual significance of the coefficients’ estimates for the liquidity variables. Overall, my analysis casts doubt on the results obtained in Adrian et al. (2015).
|
|---|