In this paper we study exchange rate effects due to shifts in the portfolio composition of the Colombian financial sector during 2003-2014. We first provide a theoretical understanding of the channel's transmission mechanism by modeling how the banking sector optimally allocates its portfolio composition. This allows us to characterize departures from the uncovered interest rate parity condition (UIP) in terms of foreign and domestic assets. In the empirical application, we control for a potential simultaneity bias by using a novel instrument for portfolio compositions: the use of sovereign credit ratings and outlook changes made by Moody's, Standard and Poor's and Fitch Ratings. Our findings indicate that shifts in portfolio balances affect only the long term (5-year) risk premium in up to five months before the effects subside. Additionally, we find stronger and more persistent portfolio effects in cases in which US ratings increased relative to Colombian ratings.