Competing Explanations for Bureaucratic Preferences
| Author | Andrew B. Whitford |
| Published date | 01 July 2007 |
| Date | 01 July 2007 |
| DOI | http://doi.org/10.1177/0951629807077568 |
| Subject Matter | Articles |
COMPETING EXPLANATIONS FOR
BUREAUCRATIC PREFERENCES
Andrew B. Whitford
ABSTRACT
This study defines and compares three broad theories that seek to explain
bureaucratic preferences. I first argue that each of these explanations is com-
plex – that no single measurable attribute encapsulates the entire theory. Second, I
argue that these explanations are non-nested – that at least one attribute represent-
ing a given theory cannot be expressed as resulting from attributes that represent
one of the other theories. Based on the theory of comparisons of non-nested mod-
els, I assess these three competing explanations with the Likelihood Dominance
Criterion, an approach for assessing the total explanatory power of a given theory
relative to that offered by other theories.
The comparisons take place in the context of the bureaucratic implemen-
tation of policies governing the remediation of hazardous waste at the state
level in the United States. The comparisons show that bureaucratic prefer-
ences are best explained by the organizational capacity and constraints expla-
nation rather than more proximate political and task environment theories.
In total, the agency’s rules, capacity, and characteristics form a better total
explanation of the observed variance in bureaucratic preferences than either
proximate state politics or the agency’s task environment.
KEY WORDS .bureaucracy .model selection .organization .political
control .preferences
Introduction
Economists have long given thought to approaches for uncovering the revealed
preferences of a government bureaucracy. Most notably, McFadden (1975,
1976) offered a theory and methodology for uncovering the underlying distribu-
tion of decision rules that a state transportation agency followed when all that
could be observed were the intermediate or final choices the agency made. Since
McFadden’s pioneering work, numerous studies have examined the choices
made by agencies – almost always with the ultimate goal of estimating the
‘reasoning’ behind the choices (the choices’ determinants). Key to this enter-
prise are the assumptions that the bureaucracy acts ‘as if’ it is a single decision
maker, with a stable set of preferences, which chooses the action at each deci-
sion point leading to the greatest expected utility. The actual decisions reveal
Journal of Theoretical Politics 19(3): 219–247 Copyright Ó2007 Sage Publications
DOI: 10.1177/0951629807077568 Los Angeles, London, New Delhi and Singapore
http://jtp.sagepub.com
the agency’s preferences when a statistical model is employed. Perhaps not
surprisingly, many studies on ‘revealed bureaucratic preferences’ address the
decisions of federal and state agencies in environmental protection policy,
including pesticides regulation (Cropper et al., 1992), Superfund (Gupta et al.,
1996), toxic substances (Van Houtven and Cropper, 1996), the endangered
species program (Simon et al., 1995; Metrick and Weitzman, 1996), and leaking
underground storage tanks (Berrens et al., 1999). Other agencies studied in this
way include the Consumer Product Safety Commission (Thomas, 1988) and the
Federal Trade Commission (Weingast and Moran, 1983).
These studies (along with others not claiming to ‘reveal’ bureaucratic prefer-
ences but still investigating agencies’ decisions) make varying claims about the
dependence of bureaucratic decisions and preferences on the agencies’ organi-
zational, political, and bureaucratic task environments. When specified, each
of these theories can help us better understand the sources of bureaucratic pre-
ferences. However, our full understanding of the explanatory power of a given
explanation (organizational, political, or task environment) is fundamentally
constrained by two epistemological problems. The first problem is discovering
the standard of justification for a knowledge claim about a specific theory – why
we can claim to have evidence for a mechanism’s effect in fact. This is trans-
lated into a methodological problem: how to test the explanatory power of an
entire explanation – and not just its components – when the explanation is
encompassed by a complex set or combination of causal variables, especially
if the total effect of that combination is more informative than its individual
pieces.
In this study, I define and compare three broad theories of the determinants
of bureaucratic preferences that are expressed as organizational, overhead poli-
tical control, and task environment explanations. Organizational capacity and
constraints theory suggests that an agency’s organizational attributes – its rules,
regulations, and reward mechanisms – determine its decisions (e.g., Downs,
1967; Thompson, 1967). The overhead political control model suggests that the
proximate preferences and influences of external political forces determine the
decisions of an agency (e.g., Joskow, 1974; Noll, 1985; Scholz and Wei, 1986).
Task environment theory argues that the problem and situational imperatives
facing an agency determine the choices it makes (e.g., Warwick, 1975; Lipsky,
1980; Wilson, 1989). Each of these theories has no single measurable attribute
in the real world: each theory has multiple measurable attributes, so testing a given
theory requires simultaneously assessing its measurable complexity.
The second knowledge problem is encountered when assessing the explana-
tory power of one theory relative to that of another theory. In the language of
statistics, model selection procedures or non-nested hypothesis testing allow one
to assess model performance without having to assume that one theory is prima
facie true, or that two theories can be artificially combined into a single explana-
tion (MacKinnon, 1992; Clarke, 2001; Pesaran and Weeks, 2001). Essentially,
220 JOURNAL OF THEORETICAL POLITICS 19(3)
To continue reading
Request your trial