The exchange rate is influenced by multiple national and international macroeconomic factors, which generates high levels of uncertainty. The objective of this research is the construction of ARIMA-GARCH and ARIMAX-GARCH models as a tool for the forecast of the exchange rate in Colombia from the daily returns of the closing prices USD/COP and its analysis of dynamic correlation with some of the most explicative variables. The results suggest that the incorporation of significant exogenous variables within the ARIMAX-GARCH model with persistent correlation according to the DCC (Dinamic Conditional Correlation) model to the USD/COP pair generates out-of-sample forecasts with better performance than the ARIMA-GARCH univariate models. © 2019, Universidad Pablo de Olavide.