We investigate relationships between life history traits and the character
of population dynamics as revealed by time series data. Our classification
of time series is according to 'extinction category,' where we identify thr
ee classes of populations: (i) weakly varying populations with such high gr
owth rates that long-term persistence is likely (unless some extreme catast
rophe occurs); (ii) populations with such low growth rates that average pop
ulation size must be large to buffer them against extinction in a variable
environment; and (iii) highly variable populations that fluctuate so dramat
ically that dispersal or some other refuge mechanism is likely to be key to
their avoidance of extinction. Using 1941 time series representing 758 spe
cies from the Global Population Dynamics Database, we find that, depending
on the form of density dependence one assumes, between 46 and 90% of specie
s exhibit dynamics that are so variable that even large carrying capacities
could not buffer them against extinction on a 100-year time horizon. The f
act that such a large proportion of population dynamics are so locally vari
able vindicates the growing realization that dispersal, habitat connectedne
ss, and large-scale processes are key to local persistence. Furthermore, fo
r mammals, simply by knowing body size, age at first reproduction, and aver
age number of offspring we could correctly predict extinction categories fo
r 83% of species (60 of 72).