The Northwest Atlantic continental shelf is a large ecosystem undergoing rapid environmental changes, which are expected to modify the productivity of natural marine resources. Current management of most fished species assumes stationary production relationships or time-invariant recruitment rates. With linear state-space models, we examined the evidence of dynamic productivity for 25 stocks of the Northeast US shelf. We expanded the suite of options available within the state-space approach to produce robust estimates. Fifteen of the stocks exhibited time-varying productivity or changes in their maximum reproductive rate. Few productivity time series are related across the whole region, though adjacent stocks of the same species exhibited similar trends. Some links to region-wide environmental variables were observed. We demonstrate that fish recruitment can often be better predicted over a short-term horizon by accounting for dynamic productivity, which could be valuable for fisheries management. Improving predictions by incorporating environmental covariates or covariance among the stocks must be considered case by case and with caution, as their relationships may change over time.