A Statement on Risk Assessment 
Arthur J. Schwartz 
Associate Professor 
Department of Mathematics 
University of Michigan 
Ann Arbor, Mich. 
In volume 43, number 146 on page 33056 of the Federal Register, 
the director of the NIH states "The work of Robin Holliday in 
assessing statistical probabilities of biological accidents is also 
noteworthy (see appendix P of the October 1977 Environmental Impact 
Statement.)". Earlier in the section on risk assessment (same page) 
he states "many commentators advocate more studies in risk assessment. 
It has been maintained that assumptions about biological containment 
may not be valid and that all components should be tested. Concern 
has been expressed that the biological containment safety systems 
may fail altogether." 
I pointed out during a hearing of the NIH in December 1977 
that the Holliday article was weak on a number of points and far 
from authoritative. To begin with the probabilities that Holliday 
computed were a priori probabilities at best and more likely a col- 
lection of hunches and guess work. The term "statistical proba- 
bilities" would seem to imply a posteriori probabilities based on 
actuarial evidence. The article of Holliday presents no such 
figures . 
The methodology used by Holliday makes blanket, unwarranted 
and imprudent use of the assumption of statistical independence. 
This error generally creates an unduly optimistic picture. Holliday 
considers only a few accident scenarios and ignores the very serious 
problem of design failure and outside disturbances not to mention 
totally unforeseen scenarios. 
I have consulted with other probability theorists, statisticians, 
actuarial scientists, and health physicists. I can find no theoretical 
or empirical support for Holliday's assertions. 
September 16,1978 
[A-169J 
