There is a significant historical link between maritime trade and economic activity. With the rapid growth of containerised trade over the past 60 years, a significant share of global trade is transported by container nowadays. This suggests there is a connection between container flows and economic activity. There is also a time delay between containers with goods being imported and the announcement of quarterly Gross Domestic Product numbers. With US container imports and exports being publicly available information, this leads to the research question of this thesis: Can loaded and/or empty container flows be used to predict economic activity?” If so, this could prove to be valuable information for economists, policy makers and, as this thesis highlights, professional traders working at investment banks or hedge funds. This thesis focusses on analysing transpacific containerised trade as a potential so-called forward indicator of US GDP. To measure these transpacific flows, an aggregate of US West coast port data is produced and de-seasonalised. The container flows are separated into Loaded In, Empty Out and Loaded Out flows. It is important to understand the dynamics of container trade and the variables influencing it on this route. Variables are identified that could possibly affect the supply-side of container shipping and therefore the transpacific container flows (the demand-side is seen as an expression of economic activity). After analysing the Cross Correlation Functions (CCF) of the identified, data based influencing variables, five different GDP growth prediction models are constructed. This is done by performing OLS single and multivariate regressions of the individual container flows (together with the influencing variables) on historical GDP growth data from 2000 to 2017. The resulting models are tested against an existing, commonly used forward indicator: the Purchasing Managers Index (PMI). The results show that three models, all using a form of loaded containers, outperform the PMI when predicting the direction of US GDP growth 3-months ahead over the 17-year time period of the dataset used. The probability of large prediction errors with these models are also smaller than the PMI benchmark model.