Low-Reliability Systems and Th eir Improvement 83
that are error-tolerant and that have goals that are diffi cult to measure (March
and Olsen 1995). Such theories acknowledge that, although eff ectiveness is a goal,
agencies anticipate and accept some degree of failure (Frederickson and LaPorte
2002).
Th e prospect, however, of using the literature on high-reliability systems
to understand how to improve the performance of other agencies is tempting.
H. George Frederickson and Todd LaPorte (2002) argue that according to the
typical view, reliability is increased either by maintaining bureaucratic order as
traditionally understood in public administration or by rejecting that notion
through purposeful redundancy. To shift away from these traditional under-
standings, Frederickson and LaPorte assert that one needs then to examine the
operations of high-reliability organizations so as to be able to model their inter-
nal and external attributes.
An important caveat to all of this is the fact that errors are still possible, even
as we learn more about high-reliability organizations. “False positives” occur
when an organization expends resources to counter a nonexistent threat; “false
negatives” occur when an organization disregards unlikely risks until a catastro-
phe happens. Balancing these risks is diffi cult (Kettl 2007), but since a false nega-
tive produces disaster, agencies and politicians will be biased toward preventing
those. Th is increases the costs of administration, because minimizing this form
of error will increase the number of false positives; therefore, eff orts to minimize
risk, even in high-reliability organizations, come at some cost.
An additional issue is what Frederickson and LaPorte (2002) term “the prob-
lem of rationality.” As false negatives are avoided over time, there will be increased
pressure to reduce administrative costs (incurred as the result of committing false
positives). Kettl (2007) terms this “punctuated backsliding,” in which a catastro-
phe places attention on avoiding further false negatives, but then vigilance—and
funding—diminishes over time. Given that one of the key attributes of high-
reliability organizations is adequate funding, this is a serious problem.
Additionally, improving agency performance oft en hinges upon more than
the attributes (identifi ed previously) of high-reliability organizations. For ex-
ample, some programs have essentially an indefi nite time horizon. Agencies in
charge of monitoring nuclear waste, closed mines, or water resources must be
able to do so for generations (LaPorte and Keller 1996). Not only do these is-
sues demand high-reliability systems, but they also demand institutional con-
stancy. As a result, true high-reliability organizations will probably remain
rare. Although low-reliability organizations can be improved through the study
of the high-reliability systems, there is some danger in treating them as high-
reliable organizations. Performance simply will not match expectations. Karl
Weick, Kathleen Sutcliff e, and David Obstfeld (2008) note that where eff orts at
improvements through Total Quality Management have failed, it is likely due
to insuffi cient decisionmaking infrastructure (the sixth through ninth items in
the previous section). One might also consider here the manner in which the No