We use the Harvard catalogue of seismic moment tensor solutions to inv
estigate the statistical properties of incremental static stress cause
d by earthquakes. Using Okada's (1992) program for a point dislocation
in a half-space, we compute normal and shear stress on nodal planes f
or each earthquake in the catalogue, as well as stress invariants in t
he focal zone of each event. These accumulated stress values are calcu
lated at the location of any future reference earthquake (pre-stress)
and then compared with the stress level measured at the same point aft
er the event (post-stress). Comparison of the statistical distribution
s for pre- and post-stress indicates that the normal stress level has
little influence on an earthquake occurrence. On the other hand, the s
hear stress in the focal zone of an event is significantly higher befo
re an earthquake than after that earthquake (the stress caused by the
reference earthquake is not taken into account in this comparison). Ea
rthquakes are more likely to be induced by incremental shear stress if
the stress and the moment tensor of an ensuing event are consistent,
i.e. the incremental stress has the same sign as the seismic moment te
nsor of the reference earthquake. Similar results are obtained if we c
ompare the first and second pre- and post-stress invariants: earthquak
e triggering is not influenced by the average normal stress (first inv
ariant); contrary to that, the average shear stress (second deviatoric
invariant) is significantly larger before an earthquake than after it
. The distribution of hypocentres from the PDE catalogue confirms the
above conclusions: the incremental stress caused by events in the Harv
ard list and measured at hypocentre locations of PDE events, is higher
for the pre-stress shear component, but shows no significant differen
ce for compressional/dilatational stresses. These findings, if interpr
eted in a typical framework of the Coulomb failure criterion, would su
ggest that the effective friction coefficient mu is close to zero. Sin
ce this effect is observed in various seismotectonic regions for shall
ow and intermediate earthquakes (with a depth of as much as 300 km), t
he conventional explanation for the low value of mu-high pore pressure
-is less plausible. We conjecture that these results can be explained
by the fractal pattern of earthquake fault geometry which is due to th
e fault self-organization in conditions of high lithostatic and tecton
ic shear stress.