The reliability of long-period ground motion from digital strong motion data is first explored using records from co-located broadband and digital strong motion instruments. It is shown that displacement waveforms obtained by double integration of the accelerogram need not be free of unrealistic baseline drift to yield reliable spectral ordinates up to at least 10 s. These findings are further corroborated by considering a set of synthetic accelerograms contaminated by random long-period noise. The results of this study suggest that high-pass filtering the digital acceleration record from a cut-off period selected to suppress baseline drifts on the displacement waveform appears to be in most cases too conservative and unduly depletes reliable information on long-period spectral ordinates. Based on these results, new predictive equations for long-period ground motions are obtained and implemented in probabilistic seismic hazard analyses. To illustrate some representative results, ground motion hazard maps in terms of long-period spectral displacements for the Southern Calabrian Arc (Italy) are shown.