This paper concerns the joint modelling of wind power and hydro inflow for long-term power system scheduling. We propose a vector autoregressive model applied to deseasonalized series to describe the joint generating mechanism of wind and inflow. The model was applied to daily and weekly bivariate time series comprising wind and inflow from seven regions in Norway. We found evidence of both lagged and contemporaneous dependencies between wind and inflow, in particular, our results indicate that wind is useful in forecasting inflow, but not the other way around. The forecasting performance of the proposed VAR models was compared to that of independent AR models, as well as the persistence forecasts. Our results show that the VAR model was able to provide better forecasts than the AR models and the persistence forecast, for both the daily and weekly time series.