Temporal distribution of earthquakes is key to seismic hazard assessment. However, for most fault systems shortness of large earthquake catalogues makes this assessment difficult. Its unique long earthquake record makes the Dead Sea fault (DSF) exceptional to test earthquake behaviour models. A paleoseismological trench along the southern section of the DSF, revealed twelve surface-rupturing earthquakes during the last 8000 years, of which many correlate with past earthquakes reported in historical chronicles. These data allowed us building a rupture scenario for this area, which includes timing and rupture length for all significant earthquakes during the last two millenaries. Extending this rupture scenario to the entire DSF south of Lebanon, we were able to confirm the temporal-clustering hypothesis. Using rupture length and scaling laws, we have estimated average co-seismic slip for each past earthquake. The cumulated slip was then balanced with long-term tectonic loading to estimate the slip deficit for this part of DSF over the last 1600 years. The seismic-slip budget shows that the slip deficit is similarly high along the fault with a minimum of 2 meters, which suggests that an earthquake cluster might happen over the entire region in the near future. Successions of intense periods of seismic activity rupturing significant length of a fault followed by longer periods of seismic quiescence have been documented along several strike-slip faults 1-6 , suggesting that temporal clustering of earthquakes might be a common behaviour for major strike-slip faults. Testing this assumption, however, has long been hampered by the lack of consistent earthquake time series for long-enough fault sections. Moreover, temporal clustering remains a critical issue for seismic hazard models that only started to be addressed in the most recent modelling attempts.