Short-term traffic forecasting is playing an increasing role in modern transport management. Although many short-term traffic forecasting methods have been explored, the spatiotemporal dependency of traffic flow, an important characteristic of traffic dynamics that can benefit the forecasting of traffic changes, is often neglected in short-term traffic forecasting. This paper first investigates the spatiotemporal dependency of traffic flow using cross-correlation analysis and then discusses its implications in terms of traffic forecastability and real-time data effectiveness. This can help us to understand traffic flow, and hence improve the performance of forecasting models.