Matrix population models have entered the mainstream of conservation biolog
y, with analysis of proportional sensitivities (elasticity analysis) of dem
ographic rates becoming important components of conservation decision makin
g We identify areas where management applications using elasticity analysis
potentially conflict with the mathematical basis of the technique, and we
use a hypothetical example and three real data sets (Prairie Chicken [Tympa
nuchus cupido], desert tortoise [Gopherus agassizii], and killer whale [Orc
inus orca]) to evaluate the extent to which conservation recommendations ba
sed on elasticities might be misleading. First, changes in one demographic
rate can change the qualitative ranking of the elasticity values calculated
from a population matrix, a result that dampens enthusiasm for ranking con
servation actions based solely on which rates have the highest elasticity v
alues. Second, although elasticities often provide accurate predictions of
future changes in population growth rate under management perturbations tha
t are large or that affect more than one rate concurrently, concordance fre
quently fails when different rates vary by different amounts. In particular
, when vital rates change to their high or low values observed in nature, p
redictions of future growth rate based on elasticities of a mean matrix can
be misleading, even predicting population increase when the population gro
wth rate actually declines following a perturbation. Elasticity measures wi
ll continue to be useful tools for applied ecologists, but they should be i
nterpreted with considerable care. We suggest that studies using analytical
elasticity analysis explicitly consider the range of variation possible fo
r different rates and that simulation methods are a useful tool to this end
.