This paper uses regression techniques to take a second look at a class
ic risk-perception data set originally collected by Paul Slovic, Sarah
Lichtenstein, and Baruch Fischhoff. As discussed in earlier studies,
the attributes expected mortality, effects on future generations, imme
diacy, and catastrophic potential all significantly affect risk rating
s. However, we find that perceived risk and dread show different regre
ssion patterns; most importantly, only perceived risk ratings correlat
e with expected mortality. In addition, average risk ratings are found
to be significantly affected by perceived individual benefits, which
suggests that perceptions of risk are net rather than gross indicators
of harm.