The Policy Studies Journal, Vol. 34, No.

3, 2006

Policy Coherence and Policy Domains

Peter J. May, Joshua Sapotichne, and Samuel Workman

Policy scholars generally agree that greater coherence of policies is desirable, but the concept is
under-theorized and has received little empirical examination. This research examines the policy
coherence of 18 policy domains and considers institutional factors that affect variation among them.
There is considerable variation in coherence among substantive, regional, and identity-based policy
domains. Greater degrees of policy coherence exist for policy domains that have dominant congres-
sional committees or have more involvement of lead federal agencies. These findings extend what
policy scholars know about policy subsystems in American policymaking to consideration of the
coherence of policy domains.
KEY WORDS: policy coherence, policy domain, policy subsystem

Introduction

A casual observer of the popular and academic characterizations of public

policy could easily conclude that policymaking institutions in the United States
are incapable of producing coherent policies. Discussions of policy topics almost
inevitably engender complaints about inconsistencies in policies within a given
policy area. These inconsistencies foster what implementation scholars have long
identified as a source of implementation gaps (see Goggin et al., 1990; Mazmanian
& Sabatier, 1983). Yet, more is at stake than implementation success, as policy
coherence also relates to the consistency with which policy intentions are signaled.
As pointed out by Schneider and Ingram (1997), a lack of policy coherence sends
confusing messages to potential policy targets—children, the elderly, women, and
so on—about the importance of their concerns. As such, policy coherence and
policy politics are intertwined.
Fragmentation of policy systems fosters the type of disjointed policymaking
that is common to American politics (see Baumgartner & Jones, 1993, pp. 235–251).
But, pointing out that fragmentation of policy systems undermines policy coher-
ence tells us little about what constitutes coherence or the ways with which
institutional dislocations contribute to it. May et al. (2005) suggest that the array
of interests and issues among different components of a given policy area is
important to consider. They argue coherence is undermined when these pull in

381
0190-292X © 2006 The Policy Studies Journal
Published by Blackwell Publishing. Inc., 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington Road, Oxford, OX4 2DQ.
382 Policy Studies Journal, 34:3

different directions and enhanced when they are more or less in accord. Their
empirical understanding of this, however, is limited to the obscure case of Arctic
policy for which they only have limited measures of key attributes.
We seek to extend this understanding of policy coherence by assessing the
coherence of different policy domains and by considering the role of institutions
in affecting it. Policy scholars have a long-standing concern with differences
among policy domains that Andrew McFarland (2004, p. 6) notes reflects assump-
tions that the structure of power and the nature of the process vary among policy
areas. The interplay of politics and policies within different domains is central to
our theorizing about variation in policy coherence. This leads us to consider the
role of issue and interest concentration and of policy targeting within domains. We
assess these attributes for 18 policy domains and examine institutional factors that
contribute to varying degrees of coherence.

Policy Coherence and Policy Domains

Policy coherence is an elusive concept that is easily understood but difficult to

measure. In common parlance, coherence implies that various policies go together
because they share a set of ideas or objectives. Appalachian policy since 1965 has
consisted of a relatively coherent set of educational, environmental, jobs-related,
and health-related policies that are aimed at enhancing economic development of
the region. This commonality in provisions and goals illustrates a high degree of
coherence. In contrast, many policy areas are marked by inconsistent policies and
little commonality of goals. We have identified these criticisms in commentary
about national policies for children (Meucci, 1997), the disabled (Watson, 1993),
the environment (Nierenberg, 2001), energy (Bleakley & Latoff, 2004), families
(Bogenschneider, 2000), immigration (Wells, 2004), rural areas (Bonnen, 1992),
transportation (Dilger, 2003), urban housing (Koschinsky & Swanstrom, 2001), and
women (Gelb & Palley, 1996). Clearly, policy coherence is a relative term for which
there can be differing degrees of the attribute.
One complication for assessing policy coherence is identifying the policies
that in principle should cohere. The scope of the policies can be as broad as the
consistency of environmental policies as a whole, as specific as the consistency of
particular provisions for addressing environmental impacts of animal feeding
operations, or something in between. We suggest the concept should be applicable
to multiple levels given that the boundaries around policies are often malleable
(see Burstein, 1991). One starting point is to consider coherence as applied to what
are commonly labeled as distinctive policy domains while recognizing that these
differ in scope and sometimes overlap. Policy scholars agree that policy domains
are more or less established areas of policy that give meaning to common problems
and have integrative properties (see Browne, 1995; Burstein; Laumann & Knoke,
1987). Our investigation of policy coherence advances this understanding.
The notion of policy coherence should apply to a diversity of policy areas.
These include substantive areas (e.g., environment, energy, health), policies affect-
ing particular groups (e.g., children, families, the elderly), and policies affecting
May/Sapotichne/Workman: Policy Coherence and Policy Domains 383

particular geographic areas (e.g., rural, urban, specified regions). The policies
suggested by the parenthetical examples are common constructions of different
policy areas even though they differ in the degree to which they crosscut different
issues. One of the research questions we address is whether there are systematic
differences in coherence among substantive, group, and geography-based policies.
A second complication for assessing policy coherence arises from the inability
to directly measure the consistency of policies. In principle, one could gauge the
compatibility of the goals or intent of different policies within a given domain.
This turns out to be exceedingly difficult. As noted by Schneider and Ingram (1997,
pp. 82–84), policy goals can be framed broadly or narrowly, be opaque, or may be
largely symbolic or hortatory (also see Stone, 1997, p. 37). The stated goals or intent
of policies, as found in preambles to statutes, are often so vague as to provide little
basis for assessing consistency. Reading between the lines to infer goals is a highly
subjective approach to assessing policy consistency. Despite a variety of heroic
efforts, we found it impossible to find an effective operationalization of degree of
commonality of goals among policies within different policy areas.1

Coherence Attributes: Issues, Interests, and Policy Integration

Given the inability to directly measure policy consistency, we have opted to

employ a set of indirect measures that—we argue—in combination, contribute to
differing degrees of policy coherence at the domain level. From this perspective,
policy coherence is a latent concept. We draw from policy scholars’ understanding
of the interplay of interests and issues that dominate the depiction of policy
politics in the United States to suggest relevant attributes. As we elaborate here,
this understanding suggests that policies within a given policy area cohere because
of consistent issue framing, supportive policy publics or clients, and integrative
policy properties. Given these basic precepts, we think of issues, interests, and
targeting as variable policy attributes for a given policy area that form the basis for
gauging policy coherence.
Most policy domains are comprised of a variety of issues. For example,
Laumann and Knoke (1987) empirically characterize health policy in the 1970s as
comprising 56 issues and energy policy as made up of 65 issues. The crowding of
the issue space for a given policy area is not as important for policy coherence as
is the degree to which attention is focused on a smaller set of issues. Shifts in
attention among issues undermine policy consistency (see Baumgartner & Jones,
1993). As such, a policy space that has greater issue focus is more likely to have
stronger policy coherence than one that contains a diversity of issues and little
focus. As we demonstrate with this research, the relevant consideration is the
degree of concentration of issues and not the number of issues.
Issues do not arise out of the blue. The underlying interest group involve-
ment is a key aspect of both issue diversity and policy coherence. A policy
space that is comprised of a diversity of interests that seek out issue niches will
appear more as a checkerboard of issues than a mosaic. In contrast, a policy
space that is dominated by a few interests will force attention to a more
384 Policy Studies Journal, 34:3

restricted set of issues, thereby imposing greater policy coherence or at least the
potential for it. This obviously does not mean that the interests in less-crowded
policy area necessarily agree. With fewer interests, there is greater potential for
crafting the type of policy constituency that May et al. (2005) argue is important
for fostering policy coherence. For example, William Browne (1995) suggests
that the traditional issue focus of agricultural policy on farm development
results from an accommodation of diverse actors—including congressional
committees—who promote a common interest in farm policy. As with issue
concentration, we show that the relevant consideration is the degree of interest
concentration within a policy domain.
As suggested by Browne’s (1995) depiction of agricultural policy, the coher-
ence of a policy area depends on how issues and their interests relate. A diverse
set of issues can still cohere if policies contain integrative properties that link issues
and interests. The consistency of policy goals is in itself a potentially powerful
integrative force, as illustrated by Appalachian policy. But, as noted above, goals
are often too vague and too numerous to be useful in this regard. Two other
potential policy glues have been discussed in the literature. One is the degree of
commonality in perceptions of policies and the ideas they contain for a given
policy area. Baumgartner and Jones (1993, pp. 25–26) label this policy image and
document the importance of a consistent image for issue stability and a shift in
image as a factor in policy change. Similarly, Richardson (2000; also see Jordan,
2005) describes the importance of common language for describing policy prob-
lems as a policy glue within policy communities. Baumgartner and Jones measure
policy images for different issue areas over time as having a positive or negative
tone. We suggest that this approach is useful for understanding pressure for
change but is more limited in understanding the role of images as an integrative
force for policy coherence.
A second integrative force is the targeting of policies involving the assignment
of benefits or burdens to specific subsets of the population. Schneider and Ingram
(1993, 1997, pp. 84–89) argue that policy targeting is a key attribute of policy
design that serves important integrative functions. They argue that policy targeting
reifies the policy image (social construction, in their terminology) of the target
group. As such, policies that have more specific targets establish the deservedness
of those groups for policy attention. More diffuse policies fail to provide this
commonality of purpose. Schneider and Ingram (1993) also suggest that targeting
can be employed as a proxy for the political influence of the affected group in that
politicians respond positively to target groups with greater perceived political
power.
The discussion of issues, interests, and targeting provides a basis for thinking
about what constitutes different degrees of policy coherence among policy areas.
On the one hand, as suggested by May et al. (2005) for Arctic policy, a domain
comprised of diverse issues, many competing interests, and little targeting will be
relatively incoherent. On the other hand, a domain comprised of focused issues,
supportive interests, and greater targeting would be relatively coherent. This is
illustrated by William Browne’s (1995) depiction of the focus of agricultural policy
May/Sapotichne/Workman: Policy Coherence and Policy Domains 385

on farm development as reinforced by a supportive policy monopoly around this

policy theme.
The vast array of policy spaces that fall between these endpoints on a
hypothetical continuum is more problematic for gauging the degree of policy
coherence. We argue that policy domains with a diversity of issues or interests are
not necessarily incoherent (also see May et al., 2005). This is because a diverse set
of policy issues and interests can cohere if they are linked by common set of ideas
or greater targeting. Susan Fuhrman (1994) suggests that education policy in the
1990s achieved greater policy coherence by emphasizing education standards that
“not only provide policy direction, but also are intended to unite various programs
and policies so as to overcome the fragmentation that has characterized education
policy”(p. 85). Schneider and Ingram (1997, p. 87) suggest that the greater targeting
of criminal justice policy toward juveniles provided greater consistency in treat-
ment of juveniles.

Hypothesized Differences in Policy Coherence

Policy scholars provide insights about the interplay of issues and interests that
suggest distinctive patterns among policy domains. These notions and our theo-
rizing above provide a foundation for suggesting hypotheses about differences in
policy coherence among different types of domains and about the institutional
forces that shape policy coherence.

Identity and Regional Policies. The common depiction of the interplay of issues and
interests within a given policy area is one of well-developed networks of issues
and interests that have varying degrees of overlap. We argue that these connec-
tions lessen the tug of policy components in different directions and provide a
basis for more coherent policies. But, not all policy areas involve well-developed
networks of interests or issue publics with overlapping interests (see May, 1991).
When the issues and interests diverge within a given policy area, the tug is in
different directions and undermines policy coherence.
Hypothesis 1: The policy coherence of substantive policy domains is greater than
that of identity-based or regional policy domains.
We suggest that identity-based policies—those about children, families, women,
and other groups—and regional polices—urban, rural, and geographically specific
policies—tend to be comprised of a diversity of issues and interests that have little
in common. Because of this, they will not exhibit the degree of coherence that is
obtained for substantively oriented policies. Urban policymaking entails consid-
eration of housing, energy, health, economic development, employment, environ-
mental, transportation, and other issues. Women’s policy entails consideration of
a similar diversity of issues. Because the different subissues often lack common
policy glue, the broader policy area is little more than an artificial agglomeration
that has the label of a geographic region (i.e., rural or urban policy) or demo-
graphic characteristics (i.e., children, family, or women’s policy). Substantive
386 Policy Studies Journal, 34:3

domains like education, the environment, and transportation have greater com-
monalities in issues, interests, and objectives—recognizing that this is a relative
comparison—and as a result will exhibit greater policy coherence than identity and
regional policies.
Situations involving identity-based and regional policies can be imagined
where policy coherence is stronger than might otherwise be the case. This might
be the case if policy targeting is greater in focusing attention on defined groups or
regions (as opposed to diffuse groups like women or regions like rural areas) or
if a focused set of issues is highlighted to provide a stronger policy glue than
simple demographic or geographic identification. For example, policies for veter-
ans are likely to be more coherent because they focus on improved benefits and
opportunities for veterans. Similarly, the coherence of a regional policy could be
increased if there is issue focus and consistent targeting of services to that
geographic local.

Institutional Considerations. Policy domains do not exist in a vacuum. They are

what Burstein (1991, p. 328) labels as components of political systems that have
substantive, organizational, and socially constructed bases. Policy scholars have
long struggled to make sense of the networks of actors or subsystems that operate
within different policy areas (see Berry, 1989; Gais, Peterson, & Walker, 1984). A
key component of this area of research is consideration of the institutional
arrangements that provide structure to policy subsystems. We consider differences
in institutional arrangements as one basis for explaining variation in coherence
among policy areas.
An important caveat to any institutional explanation is the endogenous nature
of these influences. As the term subsystem implies, the makeup of a policy domain
both responds to and reifies institutional arrangements. We treat coherence as a
result of institutional choices and processes. But, it may well be that the institu-
tional structures we attribute as influential result from the same policy and
political demands that foster coherence. Or, it may be a bit of each. Our data are
too limited to sort out these differences, but they are important to keep in mind
when considering our findings.
Hypothesis 2: Congressional institutional influence hypothesis—policy coherence
is greater for policy domains for which there is more dominant congressional
committee involvement.
There are two lines of reasoning that support this proposition. One, based on the
information perspective on Congress (Krehbiel, 1991), suggests that as coordina-
tion costs increase, policies will be less consistent. The costs of coordination go up
as issues within a given policy domain are parceled out to a greater number of
committees. A second argument is that interest group involvement is also shaped
by committee structure. As committees and subcommittees stake claims to issues
in the “congressional jurisdictional battlefield” (Baumgartner & Jones, 1993, pp.
193–215), they engage different interests, issues, and perspectives. The more
committees that are involved, the more clouded is the policy space.
May/Sapotichne/Workman: Policy Coherence and Policy Domains 387

Greater domination of a policy area by particular committees can overcome the

fragmentation fostered by multiple committees. Jeff Worsham (1998) examines this
in studying the interplay of interest group alignments and congressional institu-
tional structure for policies concerning nuclear energy, forestry, and Native Ameri-
cans. He documents how home committees are able to shape to varying degrees
the involvement of groups concerning issues for which they serve as institutional
cornerstones of subsystem arrangements. As additional committees engage similar
issues, Worsham finds that a wider range of interests are brought to the table and
the tone of the discussion changes (also see Talbert, Jones, & Baumgartner, 1995).
This diffusion of committee attention, we suggest, undermines policy coherence.

Hypothesis 3: Bureaucratic institutional influence hypothesis—policy coherence is

increased for policy domains for which there is a dominant executive agency.

The logic of this hypothesis is similar to the congressional influence hypothesis

in that executive branch agencies can also act as an institutional basis for policy
consistency. They are important players in policymaking for which the uniformity
of their voice, like that of any coalition of interests, can affect the contours of
policy. This notion is widely accepted in the literature on executive and congres-
sional relationships (see Ripley & Franklin, 1991). John Kingdon (1995, pp. 28–29)
found that agency leaders had substantial influence in the 1970s for agenda setting
about health and transportation issues.
When the views of a few agencies dominate policy discussions, issue defini-
tions are likely to be more focused and thereby contribute to increased policy
coherence. Lisa Miller (2004) documents this dynamic showing how the increased
involvement of criminal justice bureaucrats in congressional hearings since the
mid-1970s has led to “a stabilization of the policy environment” (p. 581) and a
focusing of the criminal justice agenda. When no agency is assigned responsibility
for a given issue, as is the case with women’s and children’s policies in the United
States, there is less ability to distill diverse policy perspectives. This, in turn,
undermines policy coherence.

Data and Measures

Data

Our unit of analysis is a policy domain for which we examine attributes of 18

of them. Table 1 summarizes the domains under study and the key data sources.
These were selected to provide a contrast between substantive, identity-based, and
regional domains. This is an analytic sample of domestic policy domains that allow
for this contrast rather than a statistical sample that attempts to represent all policy
domains.2 The small number of domains clearly limits the power of our statistical
analyses, but the extensive data collection tasks necessitated limiting the number
of domains. The last column of Table 1 indicates our categorization of each policy
area.
388 Policy Studies Journal, 34:3

Table 1. Policy Domains under Study

Policy Domain Data for This Studya Domain Typeb

Hearings Public Laws
Agriculture 30 (12%) 30 (21%) Substantive
Appalachian 17 (100%) 9 (100%) Regional
Aviation 30 (20%) 30 (97%) Substantive
Children 30 (10%) 30 (52%) Identity
Disabled 30 (35%) 24 (100%) Identity
Education 30 (3%) 30 (16%) Substantive
Environment 30 (6%) 30 (64%) Substantive
Family 30 (15%) 30 (57%) Identity
Great Lakes 30 (71%) 6 (100%) Regional
Homeless 30 (34%) 15 (100%) Identity
Immigration 30 (20%) 30 (77%) Substantive
Older Americans 30 (31%) 18 (100%) Identity
Railroad 30 (18%) 30 (79%) Substantive
Rural 30 (9%) 30 (58%) Regional
Transportation 30 (8%) 30 (28%) Substantive
Urban 30 (19%) 18 (100%) Regional
Veterans 30 (6%) 30 (19%) Identity
Women 30 (13%) 30 (52%) Identity
Total 527 (11%) 450 (42%)
a
Cell entries show number of sampled hearings and public laws for
each domain. Percentages in parentheses are the percent of the total
number of hearings or public laws that our sample represents for
each domain. Data are from January 1975 through December 2000.
b
Classification based on topical orientation of each domain.
Substantive = addressing substantive issues. Identity = addressing
issues for a given group. Regional = addressing issues for a given
geographic area or region.

Following the practice of a number of policy scholars (see Baumgartner &

Jones, 1993; Browne, 1995; Laumann & Knoke, 1987), identification of the content
of each domain was based on key word searches. We limit our key word search
to the primary descriptor of each domain, rather than attempting to devise a list
of subtopics of each domain. The latter would have entailed arbitrary choices
about the content of each domain leading to inconsistencies in defining the scope
of different domains. The use of primary descriptors also helps to limit our
attention to the most salient issues that constitute each policy area. We consider
attributes of each domain for the 25-year period from January 1975 through
December 2000. We choose to address policies after 1974 out of convenience in
having access to electronic sources as well as to avoid contamination of data that
would have been introduced by congressional reforms of the early 1970s. This
avoids the confounding influence of House reforms on changing patterns of
testimony at hearings (see Heitshusen, 2000).
Our primary data come from congressional hearings and the public laws that
are relevant to each domain for the period under study. We identify hearings and
public laws for each domain using appropriate key word searches of the Policy
Agendas Project database. This database contains descriptors and coding of issue
May/Sapotichne/Workman: Policy Coherence and Policy Domains 389

attributes for congressional hearings and statutes, among other data sources, for
the post-World War II period (see Baumgartner, Jones, & Wilkerson, 2002). Hear-
ings have long been used as a basis for understanding policy dynamics with
attention to the constellation of different coalitions and the issues that are being
debated (e.g., Baumgartner & Jones, 1993; Jenkins-Smith, St. Clair, & Woods, 1991;
Worsham, 1998; Zafonte & Sabatier, 2004).
We sample relevant hearings and public laws for each domain with random
selection to obtain a sample size of 30 of each item per domain. We include all
items if the universe is less than 30. The sheer number of items among 18 domains
necessitated sampling to make the coding of the data manageable. We select 30
as a threshold number in order to provide sufficient basis for comparison across
domains. The parenthetical entries in Table 1 show the percentages of total
hearings and laws within each domain that are included in our data.
Our data are based on all hearings and public laws for the time period under
consideration with no effort to exclude particular categories of hearings or laws
other than private laws. As such, we have included appropriations and authoriz-
ing activities as well as nonlegislative hearings. Scholars have identified mean-
ingful differences in interest group participation at hearings held by different
legislative branches (Worsham, 1998), among different committee venues
(Baumgartner & Jones, 1993, pp. 193–215), and among different types of hearings
(Talbert et al., 1995). These differences clearly imply that the demands we observe
will vary by the type of forum that is considered. We saw no basis for excluding
one type of forum over another, so we choose to include all.

Measures

The hearings and public laws that we sample for each domain provide the
bases for the measures we develop concerning attributes of policy coherence and
institutional considerations. We aggregated scores for each measure by policy
domain in order to provide a depiction of each of the 18 domains we consider.

Issue Concentration. We conceptualize issue concentration as the degree to which

one or more issues dominate the congressional hearings within a given policy
domain. In order to have a consistent set of potential issues across domains, we
employ the topic codes provided by the Policy Agendas Project for identification
of the issues that are addressed by hearings within each domain under study. This
coding scheme employs 19 different major categories of issues that have proven to
be highly reliable and robust for analyzing issue content (see Baumgartner et al.,
2002).3 The number of relevant issue categories across the domains we consider
range from 4 (Appalachian) to 18 (education, environment, and urban) with a
median of 13. For each policy domain we calculate the proportion of hearings that
fall within each relevant issue category. Based on these proportions, we calculate
issue concentration for each domain using a Herfindahl index.4 Scores approaching
1 indicate high degree of concentration and those approaching zero indicate low
concentration of issues.
390 Policy Studies Journal, 34:3

Interest Concentration. We conceptualize interest group concentration as the degree

to which one or more categories of witnesses dominate the hearings within a given
policy domain. This is based on a coding of witness lists for each hearing as
obtained from Lexis-Nexis Congressional Universe. Unlike the issue categories for
which we could rely on established categories of issues, we had to identify
relevant categories of interests for each domain. Just as the number of issue
categories varies by domain, we allowed the number of interest categories to vary
by domain. We sought consistency in the scope of the categories and the interests
they represent across domains in order to avoid imposing fragmentation while also
allowing for substantive differences across domains. This was accomplished by
considering four basic types of interests as common to all domains: state and local
government officials, research and expert witnesses, federal agency officials from
the agency responsible for policy implementation of the designated policy area,
and federal officials from other agencies. We also allowed for domain-specific
interests that reflect substantive differences in the domains. For example, Appa-
lachian policy was coded as having business, housing and economic development,
citizen organizations, regional, health, and labor interests in addition to the four
basic categories of interests.
The number of interest categories across the domains range from 6 (environ-
ment) to 13 (family) with a median of 10. For each policy domain, we calculate
the proportion of witnesses that fall within each of the relevant categories. We
obtained 73-percent agreement among coders for the assignment of witnesses to
interest categories for a sample of 30 hearings that involved 296 witnesses. We
measure interest concentration using a Herfindahl index of the concentration of
interests among the categories. This measures the degree to which one or more
categories of interests dominate the hearings for that domain as well as provides
a comparable metric to the one we devise for measuring issue concentration.

Policy Targeting. We consider the extent to which the public laws in a given domain
target particular geographic areas, groups, or entities. We coded the provisions of
each public law according to a 3-point scale of whether a given law targets a
particular region (group or entity), targets a list of areas (groups or entities), or
only addresses a diffuse geographic area (groups or entities). We find that the most
meaningful measure of targeting is the proportion of laws within a given domain
that have explicit targets (i.e., our first category of coding). The stability of this
measure is influenced by some domains (e.g., Great Lakes) that have few public
laws. We obtained 87-percent agreement among coders for the coding of targeting
among a sample of 30 public laws.

Committee Fragmentation. We measure the extent of congressional committee frag-

mentation as the degree to which one or more committees dominate hearings for
each domain. We calculate the proportion of hearings for the domains in our study
that fall within each of the relevant congressional committees as identified by the
Policy Agendas Project database.5 The number of relevant committees across the
domains we consider range from 5 (veterans) to 20 (women) with a median of
May/Sapotichne/Workman: Policy Coherence and Policy Domains 391

13. From these, we calculate a Herfindahl index that shows the concentration of
hearings for each domain within particular committees. Lower scores indicate
greater fragmentation of attention within congressional committees.

Executive Agency Involvement. Our primary interest is the degree to which one or
more federal agencies are dominant influences in policymaking. We devise two
measures. One is a dummy variable that indicates the presence or absence of a lead
federal agency for a given policy area. This designation is based on our review of
the policy history of each policy area. Each domain except for those addressing
policies for children, the Great Lakes, rural areas, and women has a designated
lead agency.6 In order to better depict executive agency involvement, we also
construct a measure that depicts the relative influence of lead agency witnesses for
those domains with lead agencies. This is calculated as the ratio of the number of
hearing witnesses representing the lead agency divided by the number of wit-
nesses representing other federal agencies. Higher scores suggest greater involve-
ment of lead agencies, while lower scores indicate lesser involvement.

Policy Implications. While not a focus of our research, we also sought to develop
measures that could show the implications of differing degrees of policy coher-
ence. We developed two measures based on the coding of public laws within each
policy domain. One measure is the proportion of laws that are largely symbolic.
The values range from 0 (transportation and immigration policies) to 0.67
(women’s policy) with a median of 0.15. A second measure is a measure of policy
stability based on the proportion of laws that entail little or no change in
substantive policy provisions from the (then) existing policy. The values range
from 0 (Great Lakes policy) to 0.87 (transportation policy) with a median of 0.38.
The measures of symbolic content and policy stability are based on a coding of
policy content from review of the title of each public law and the Congressional
Research Service summary of it. We obtain an intercoder agreement of 97 percent
for the symbolic coding and of 80 percent for the policy stability coding for a
sample of 30 public laws.

Findings

Our discussion of the findings follows from the above theorizing about policy
coherence. We first consider the structure of policy coherence among domains as
evidenced by variation in issue concentration, interest concentration, and policy
targeting. We next consider differences in coherence among different types of
policy domains. This is followed by consideration of differences in institutional
influences upon coherence.

Structure of Policy Coherence

We argue that variation in policy coherence results from differing concentra-

tions of interest group involvement, issues, and targeting of resultant policies. We
392 Policy Studies Journal, 34:3

consider each of these to be attributes of the coherence of a policy domain. We

examine how these attributes relate to each other and the extent to which they can
be used to provide a meaningful measure of policy coherence at the domain level.
We find considerable variation in each attribute among the 18 policy areas
under study. The issue concentration score ranges from 0.12 (family policy) to 0.78
(immigration policy) with a median of 0.42. The interest concentration score ranges
from 0.10 (family policy) to 0.23 (Great Lakes and veterans policies) with a median
of 0.18. The proportion of policies within a given domain directly targeting groups
or geographic areas ranges from 0.23 (children’s policy) to 1.0 (Appalachian and
Great Lakes policies) with a median of 0.63.
Figure 1 depicts variation among these three attributes. The degree of issue
concentration is on the horizontal axis. The degree of interest concentration is on
the vertical axis. The extent of direct targeting is shown by the size of the dot next
to the domain name. Less coherent policy domains are in the lower-left quadrant
and more coherent policy domains are in the upper-right quadrant, keeping in
mind differences in targeting. The fairly linear array of policy domains shown in
Figure 1 suggests that coherence can be depicted by one dimension. This obser-
vation is reinforced by the moderately strong correlations among the indicators
(r = 0.65 for ln of issue and interest concentration, p ⬍ 0.01; r = 0.53 for ln of interest
concentration and targeting, p = 0.01; and r = 0.56 for ln of issue concentration and
targeting, p = 0.01).
0.23

Veterans Great Lakes

Disabled Older Americans
Concentration of interests (Herfindahl index)

Transportation

Agriculture
0.19

Aviation
Railroad
Environment Homeless
0.17

Education Immigration
Appalachian

Children
0.13

Rural Women Urban

0.10

Family

0.12 0.18 0.28 0.45 0.78

Concentration of issues (Herfindahl index)

Figure 1. Structure of Policy Coherence for 18 Domains.

Note: Axes are scaled on a ln scale with relevant Herfindahl scores indicating the degree of issue
and interest concentration among domains. The extent of targeting is indicated by the size of the
dot for each domain.
May/Sapotichne/Workman: Policy Coherence and Policy Domains 393

We employ Principal Component Analysis to derive the dimensionality of

policy coherence from the three coherence attributes. A single dimension accounts
for 72 percent of the variation in the data. This result provides an Eigenvalue of
2.16 and strong factor loadings for each of the component items: issue concentra-
tion (ln scale 0.87), interest concentration (ln scale 0.86), targeting (0.81). A
two-dimensional solution is less satisfying as the Eigenvalue for the second
dimension is only 0.49 and the resultant components are not clearly distinguished
when employing a variety of rotation schemes. We conclude that a one-
dimensional representation of coherence is appropriate and we employ the
principal component scores as the derived measure of coherence.
One concern is that our measure of policy coherence is an artifact of the way
that we constructed our measures of interests and issues. As more interests are
involved in a given domain, they are more likely to bring more issues to the table,
thereby undermining coherence. As fewer interests are involved, they potentially
provide a better basis for targeting those interests for policy benefits (Schneider &
Ingram, 1993). We argue that the concentration of each is important for charac-
terizing coherence and not the absolute number. Otherwise, we would simply be
saying that single-issue and interest domains are more coherent than multiple-
issue and interest domains. Consider the difference between the agriculture and
the rural policy domains. By our coding, each has a wide diversity of categories
of interest groups (9 and 12, respectively) and issues (15 and 16, respectively). But,
as shown in Figure 1, agriculture policy has much stronger issue and interest
concentration than rural policy. This comports with the concentration of agricul-
ture policy on farm development and serving farm interests (Browne, 1995), and
of rural policy having a diffuse constituency and issues (Browne, 2001).
The difference between considering absolute number and concentration for
assessing coherence is evident from the observed relationships. The Pearson
correlation between our measure of coherence and the number of categories of
interest groups among policy domains is -0.48 (one-tailed p = 0.02) and between
coherence and the number of categories of issues among policy domains is -0.46
(one-tailed p = 0.03). The fact that these correlations are of much lower magnitude
than those based on concentration measures (the above factor loadings) suggests
that more is at play in fostering coherence than simply involving fewer interests
or issues. It also suggests greater internal reliability for the measure of coherence
when considering the concentration of issues and interests.
Another concern is that our measures are not picking up what we label as
coherence but simply reflect issue or policy salience. We noted above that our key
word search is intentionally biased toward the more salient aspects of each policy
area. But, we think our measures have little to do with salience. The pattern in
Figure 1 suggests a weak relationship. What we find to be more coherent domains
like Appalachian and Great Lakes policies have hardly been more salient areas
than less coherent domains like environmental and urban policies. Moreover, it is
reasonable to expect that increased salience fosters heated-up policy spaces that
engage new interests and issues, as is often the case after crises (see Birkland,
1997). This would tend to undermine coherence by diffusing issue and interest
394 Policy Studies Journal, 34:3

concentration. This is evidenced by a Pearson correlation of -0.32 (p = 0.19 two-

tailed) between the number of hearings for each domain (ln scale) and our policy
coherence score.
The validity of the policy coherence measure rests on the logic of our
construction of the measure and on face validity of the results. Figure 2 shows the
array of coherence scores among the 18 policy domains. The less coherent domains
at the bottom of Figure 2 make intuitive sense. Each of these domains—family,
rural, children, and women’s policies—lack a dominant policy, engender a diver-
sity of issues and interests, and have diffuse targeting. As noted earlier, scholars
commenting on each of these policy areas bemoan the lack of policy consistency.
The relatively high coherence scores for some of the domains located at the top
of Figure 2 are less intuitive. The high coherence of Great Lakes policy reflects a
strong focus on environmental issues, focused involvement of state officials, and
direct targeting for only six statutes. Rabe and Zimmerman (1995; also see Rabe,
1996) note the fledgling federal efforts to provide an integrative Great Lakes’
environmental policy. They suggest that policy integration has been achieved in
recent years through shared state-level actions concerning pollution prevention as
a “unifying rallying cry for reform” (p. 71). The coherence of the older American’s
policy area, comprised of 18 laws, reflects the importance of the Older American’s

Great Lakes

Older Americans
1.0
Appalachian
Transportation
Agriculture
Aviation
Veterans
Immigration
Homeless
Environment Education
Railroad
0.0
Disabled
Coherence score

Urban

Women
–1.0

Children
Rural

–2.0

Family

Figure 2. Policy Coherence Scores for 18 Domains.

Note: Coherence scores are Principal Component Analysis scores for a single dimension based on
interest concentration, issue concentration, and policy targeting. Values are relative scores arrayed
from highest to lowest.
May/Sapotichne/Workman: Policy Coherence and Policy Domains 395

Act of 1965 (Public Law 89–73) in focusing issue attention, the dominance of the
elderly lobby in hearings, and the more specific targeting of services under the
various amendments to the Older American’s Act. The Appalachian policy domain
reflects what we consider to be a strong example of coherent regional policy. The
nine statutes in our database are amendments to the Appalachian Regional
Development Act of 1965 (Public Law 89–4) that target economic development
programs for that region. As discussed by Bradshaw (1992), the coherence of this
policy domain is enhanced by the presence of an active, focused constituency for
these programs.

Categories of Policy Domains

We hypothesize that the policy coherence of substantively oriented policy

domains is greater than that of identity-based or regional policy domains (hypoth-
esis 1). In particular, we expect substantive domains to have greater issue focus
and less diversity in interest group involvement. Table 2 presents a comparison of
mean scores for each type of policy domain.
The distinctions among domain types are not as clear-cut as we hypothesized.
The differences in mean coherence scores do not reach conventional statistical levels
of significance, although the pattern in mean scores is consistent with the hypothesis.
Insight about the domain distinctions can be found by considering differences in the
individual attributes of coherence. This is particularly the case when taking into
account the degree of policy targeting. The last column of the table shows the
statistical significance levels for differences in means when introducing policy
targeting as a covariate. These findings, recognizing the low power of the statistical
tests, show that substantive policies have greater concentration of issues than the

Table 2. Coherence among Different Types of Domains

Type of Policy Domain ANOVA F-tests

Substantive Regional Identity p-Valuea Alt p-Valueb
Policy coherence attributes
Issue concentrationc 0.52 (0.07) 0.34 (0.07) 0.31 (0.08) 0.11 0.06
Interest concentrationd 0.18 (0.01) 0.17 (0.02) 0.17 (0.02) 0.60 0.14
Targeting of policiese 0.66 (0.04) 0.81 (0.16) 0.52 (0.08) 0.11 —f
Overall coherence scoreg 0.40 (0.11) 0.10 (0.63) -0.45 (0.46) 0.29 —
Number of cases 7 4 7
Notes: Cell entries are mean scores for policies within a given type of domain, standard errors are in
parentheses. Higher values indicate greater amounts of the attribute.
a
One-way ANOVA for means among policy domains.
b
ANCOVA for means among policy domains controlling for degree of policy targeting.
c
Herfindahl score for issues addressed, ln values used in calculations of F-tests.
d
Herfindahl score for interests represented, ln values used in calculations of F-tests.
e
Proportion of Public Laws that directly target categories of individuals, groups, or areas.
f
Covariate not relevant for this comparison.
g
Principal component score based on the three attributes of policy coherence.
Analysis of Variance (ANOVA)
Analysis of Covariance (ANCOVA)
396 Policy Studies Journal, 34:3

other types (p = 0.06), but do not differ with respect to diversity of interest group
involvement (p = 0.14). Geographic policies evidence 1.6 times greater targeting than
that of identity-based policies (one-tailed T-test p = 0.05).
Closer inspection of Figures 1 and 2 above suggests two contrasting worlds
of identity-based and regional policies. One is comprised of diverse issues and
interests with limited targeting that make for less policy coherence (families,
women, children, rural, and to a lesser extent urban and disabled policies). The
other is comprised of greater issue focus, less diversity in interest involvement,
and stronger targeting of policies (Appalachian, Great Lakes, homeless, older
Americans, and veterans). The mean coherence scores for the former are on
average one-half the mean scores for the latter. Issue concentration for more
coherent identity-based policy domains is three times that of less coherent identity-
based domains. The comparable factor for regional domains is two. Interest
concentrations differ for each by a factor of 1.5 and 1.4 respectively. The compari-
son of coherence scores for substantive domains with these subcategories provides,
not surprisingly given the post hoc nature of the comparison, significant differences
in mean coherence scores (F-test p-value ⬍0.01).

Institutional Forces

We hypothesized that the degree of policy coherence is shaped by a set of

congressional and executive branch institutional forces. In particular, we sug-
gested that greater dominance of congressional committees within policy
domains (hypothesis 2) and greater executive agency dominance (hypothesis 3)
would help foster greater policy coherence. Table 3 addresses the relationship
between congressional committee involvement and policy coherence. The table

Table 3. Congressional Committee Influence on Coherence

Committee Involvementa Statistical Results

Low Committee High Committee T-test p-Valueb Correlation (p-Value)c
Concentration Concentration
Coherence scored -0.64 (0.34) 0.64 (0.14) ⬍0.01 0.69 (⬍0.01)
Issue concentratione 0.25 (0.04) 0.55 (0.06) ⬍0.01 0.72 (⬍0.01)
Interest concentrationf 0.16 (0.01) 0.19 (0.01) 0.02 0.65 (⬍0.01)
Targeting of policiesg 0.55 (0.08) 0.72 (0.06) 0.05 0.38 (0.06)
Number of cases 9 9 18
Note: Cell entries are mean scores for a given institutional setting with standard errors in parentheses.
Higher values indicate greater amounts of the attribute.
a
Concentration of committee involvement in hearings (Herfindahl scores) as split into low and high
categories.
b
T-test for means between low and high involvement, one-tailed p-values.
c
Pearson’s correlation between coherence attribute and ln of committee Herfindahl score; one-tailed
p-values.
d
Principal component score based on the three attributes of policy coherence.
e
Herfindahl score for issues, ln values used in calculations of T-test and correlation coefficient.
f
Herfindahl score for interests, ln values used in calculations of T-test and correlation coefficient.
g
Proportion of public laws that directly target categories of individuals, groups, or areas.
May/Sapotichne/Workman: Policy Coherence and Policy Domains 397

shows differences in the mean coherence scores for policy domains that have
fragmented committee involvement, in comparison to the mean scores of policy
domains that have dominant committees.
The differences in mean coherence scores are in the expected direction. Those
domains that have greater committee dominance exhibit higher coherence scores,
greater issue concentration, somewhat greater interest group concentration, and
greater targeting of policy provisions. The last column reports correlations between
committee concentration scores and scores for attributes of policy coherence. Each
of these relationships is moderate to very strong and statistically significant at
p-values of 0.06 or less. These findings clearly support the hypothesis that
congressional committee dominance of a policy area is associated with greater policy
coherence. As noted earlier, the potential endogenous nature of this relationship
limits drawing causal inferences about institutional influences. Nonetheless, these
results further support the findings of policy scholars (Talbert et al., 1995; Worsham,
1998) that congressional venues matter in shaping activity within policy domains.
Table 4 addresses the relationship between executive branch agency involve-
ment and policy coherence. The table shows differences in mean coherence scores
for policy domains that do not have a lead federal agency in comparison to mean
scores of policy domains that have a responsible bureau.
The differences are in the expected direction but only the differences for the
overall coherence score and for issue concentration are detectable at tolerable
statistical levels of significance. The last column reports correlations between the
degree of agency involvement in committee hearings for those domains with lead
federal agencies and scores for attributes of policy coherence.7 These show mod-
erate relationships between agency involvement and overall coherence (at a less
tolerable level of statistical significance) and between agency involvement and

Table 4. Executive Branch Influence on Coherence

Executive Branch Involvementa Statistical Results

No Responsible Responsible T-test Correlation
Agency Agency p-Valueb (p-Value)c
Coherence scored -0.70 (0.69) 0.20 (0.22) 0.06 0.39 (0.08)
Issue concentratione 0.24 (0.08) 0.44 (0.05) 0.03 0.08 (0.39)
Interest concentrationf 0.16 (0.02) 0.18 (0.01) 0.13 0.31 (0.14)
Targeting of policiesg 0.54 (0.17) 0.67 (0.05) 0.16 0.58 (0.02)
Number of cases 4 14 14
Note: Cell entries are mean scores for a given institutional setting with standard errors in parentheses.
Higher values indicate greater amounts of the attribute.
a
Whether or not there is a lead federal agency for addressing a given policy domain.
b
T-test for means between levels of bureaucratic involvement, one-tailed p-values.
c
Pearson correlation between coherence attribute and ln of the degree of responsible agency involve-
ment in congressional hearings; one-tailed p-values.
d
Principal component score based on the three attributes of policy coherence.
e
Herfindahl score for issues, ln values used in calculations of T-test and correlation coefficient.
f
Herfindahl score for interests, ln values used in calculations of T-test and correlation coefficient.
g
Proportion of public laws that directly target categories of individuals, groups, or areas.
398 Policy Studies Journal, 34:3

targeting of policies. Taken together, these findings about executive branch influ-
ences on policy coherence are more mixed than are those for congressional
committee involvement.

Implications of Policy Coherence

While we have not sought in this research to assess the implications of

different degrees of policy coherence among domains, it is instructive to comment
about these implications to the extent that our data permit. Policy coherence can
be expected to ferment, or at least to be consistent with, policy stability. Stated
differently, the same forces—the existence of policy monopolies (Baumgartner &
Jones, 1993, pp. 6–9)—that promote coherence also promote policy stability. Our
data evidence this. The correlation among domains between policy coherence and
the proportion of policies within each domain that evidence little or no change is
0.47 (p = 0.03).8
Policy coherence can also be expected to foster more substantive policy
provisions in response to the consideration of more focused issues, concentrated
interest involvement, and greater targeting. This is evidenced in a more limited
way with the finding that the correlation among policy domains between policy
coherence and the proportion of policies within each domain that are largely
symbolic or nonsubstantive is -0.36 (p = 0.08).9 Not surprisingly, identity-based
policy domains contain greater proportions of symbolic measures by a factor of 2.9
in comparison to regional policy domains and a factor of 4.5 in comparison to
substantive domains (F-test p-value ⬍0.01).

Conclusions

More than 35 years ago, in one of the first set of commentaries about political
scientists’ efforts to characterize variation in public policy, Robert Salisbury (1968)
suggested that “[c]omparisons of demand patterns among states, issues, and groups
would be illuminating by itself, and ought to have theory-building conse-
quences . . .” (p. 172). This research follows that reasoning in addressing issue and
interest demands for different policy domains. Our analysis shows that the degree
of coherence of a policy domain can be characterized along a single dimension that
reflects a combination of the extent of issue focus, concentration of interests, and
targeting of resultant policies. Increased coherence of a policy area is associated with
greater policy stability and more frequent substantive policy provisions.
We find considerable variation among domains in the degree of policy
coherence. We fail to find strong support for expected systematic differences in
coherence levels among substantive, regional, and identity-based domains. Sub-
stantive policy areas (e.g., agriculture, environment, transportation) typically evi-
dence greater coherence, but a few cases of strong coherence in each of the other
types of domains contribute to the lack of anticipated, systematic differences.
Instead, we find two contrasting worlds of identity-based and regional policies.
One is comprised of diverse issues and interests with limited targeting that make
May/Sapotichne/Workman: Policy Coherence and Policy Domains 399

for less policy coherence. The other is comprised of greater issue focus, greater
concentration of interest involvement, and stronger targeting of policies.
We suggest that policymaking institutions are important intermediaries in
shaping the policy coherence of policy domains because of their role in shaping the
interplay of issues and interests. Consistent with prior findings about the influence
of congressional committees on the scope of policy deliberations (Talbert et al.,
1995; Worsham, 1998), we find greater coherence among policy domains for which
there is greater concentration of committee involvement in holding hearings. This
makes sense as such dominance provides focus to policy agendas and content
while also channeling interest group influence. We also find support for an
institutional influence of dominant executive branch agencies in shaping policy
coherence although the findings are not as clear-cut as they are for the congres-
sional committee influence.
These findings extend what policy scholars know about policy subsystems and
policy monopolies in American policymaking (see Baumgartner & Jones, 1993;
Berry, 1989; Gais et al., 1984) to consideration of policy coherence among policy
domains. Although subsystems have often been associated with negative images
of policymaking, our findings show that the forces commonly associated with
healthy subsystems also contribute to increased policy coherence. Supportive
institutional structures comprised of lead committees and agencies along with
focused issue attention and interest group involvement foster targeted policies and
greater policy coherence. Our data do not permit sorting out connections among
institutional structures, issues, and interests in a refined enough fashion to assess
causal links or the limits to these institutional influences. It may be that the same
factors that contribute to policy coherence also shape institutional structures,
rather than institutions acting as intermediaries. In either case, we demonstrate
that the interplay of issues and interests and the demand patterns that this creates
are central considerations for policy coherence.
Many concepts in the policy literature have undergone limited empirical
assessment. We view one contribution of this research as advancing the empirical
understanding of the policy coherence of policy domains. Although the findings
are proscribed by the limited number of policy areas that we consider and the
inability to examine each in detail, the results have good face validity and are
generally consistent with our theorizing. This gives us confidence that our choices
about conceptualization, indirect indicators, measurement, and assessment provide
at least a reasonable approximation of the concept when applied at the level of
policy domains. We specifically show that the diversity of issues and interests is
not as important to consider in characterizing coherence of domains as is the
degree of concentration of each. This helps explain why equally diverse domains
like agriculture and rural policies have very different degrees of policy coherence
as comports with substantive depictions of these domains.
Consideration of the coherence of policy domains is of more than academic
interest. Our findings draw attention to different political environments and the
challenges they pose for achieving policy consistency. The fragmentation of
institutions and the lack of coherence for some domains is not an artifact of
400 Policy Studies Journal, 34:3

inattention or simply a by-product of other policy priorities. As May et al. (2005)

argue, the lack of an effective constituency for policy areas such as rural policy,
children’s policy, and women’s policy help explain why institutions have not been
created to support those domains. Policy coherence is limited because of the lack
of strong underlying institutions, and strong underlying institutions do not evolve
because of the lack of pressures for coherent policy.
Greater coherence does not necessarily imply better policy. This is because
policy coherence tells us little about the appropriateness of a given set of policies
to the problem or issue area at hand. Stated differently, both bad and good policies
can have relatively high, or low, degrees of coherence. (Dictators are often effective
in creating relatively coherent policies that many would judge as less than
desirable.) This normative observation, of course, is no different than noting that
strong implementation does not turn a bad policy into a good one (see Majone &
Wildavsky, 1979). Nonetheless, as we demonstrate with this research, increased
policy coherence has positive attributes that most analysts would consider impor-
tant bases for better policy acceptance and implementation.

Peter J. May is a professor of political science at the University of Washington

and associated with the Center for American Politics and Public Policy.
Joshua Sapotichne is a graduate fellow at the Center for American Politics and
Public Policy at the University of Washington.
Samuel Workman is a graduate fellow at the Center for American Politics and
Public Policy at the University of Washington.

Notes

Some of the data used here were originally collected by Frank R. Baumgartner and Bryan D. Jones with
the support of National Science Foundation grant numbers SBR 9320922 and SBR 0111443. Neither the
NSF nor the original collectors of the data bear responsibility for the analyses reported here. Earlier
versions were presented at the 2005 annual meetings of the American Political Science Association and
the Association for Public Policy Analysis and Management. Comments by William Gormley, Kathryn
Harrison, Harald Saetren, Christopher Weible, David Weimer, John Wilkerson, and anonymous review-
ers are appreciated.
1. For each of the 450 public laws in our sample, we coded the primary objectives as one of five
categories—organizational or procedural, benefit related, social regulatory, economic regulatory, or
symbolic—as well as other attributes of objectives within each primary category (e.g., reduction of
existing benefits or rights, expansion of existing benefits, or creation of new benefits or rights).
Because most statutes had only implicit objectives, we found it difficult to reliably classify statutes
and often had to infer the primary objective. This exercise reified broad distinctions in policy types.
Goals within the domains we considered tended to be primarily benefit-related (e.g., Appalachian,
children, homeless) or social regulatory in nature (e.g., environment, immigration) with some
domains having goals split between these two categories (e.g., transportation). Use of the minor
codes was limited by the limited number of public laws within many domains. The distinctions in
primary objectives had little empirical import when attempting to correlate stability or other
measures of variation in policy goals with the various other attributes of policy coherence that we
consider. We concluded that our measures of policy goals were too general to be meaningful.
2. It is hard to imagine what a probability-based sample of policy domains would look like given the
lack of an existing list and the difficulty of constructing a master list of policy domains. Although
our selection of domains is somewhat ad hoc, the choices reflect a variety of domestic policy areas.
We present statistical findings in order to convey the variation in results while recognizing that this
is not a probability-based sample.
May/Sapotichne/Workman: Policy Coherence and Policy Domains 401

3. The intercoder reliability for the major topic coding for hearings under the Policy Agendas Project
is reported by the authors to be greater than 95 percent (Baumgartner et al., 2002, p. 35). The 19 major
topic codes are: agriculture; banking and commerce; civil rights or liberties; defense; education;
energy; environment; foreign trade; government operations; health; housing and community devel-
opment; international affairs and aid; labor, employment, and immigration; law, crime, and family;
macroeconomics; public lands; science and technology; social welfare; and transportation. For
further documentation, see the Policy Agendas Project web site at [Link]
accessed on August 3, 2005.
4. This score is computed by summing across all categories of interests the squared proportion of
witnesses that fall within each category of interests.
5. The Policy Agendas Project categorizes committees, taking into account changes in committee name
and jurisdiction over time, into a set of major and minor committees. Of relevance for this research
are the designated 26 major House committees and the 17 major Senate committees that are
identified as in existence in the time period for our data. For further documentation, see the Policy
Agendas Project web site at [Link] accessed on August 3, 2005.
6. Had we designated the lead agencies based on agency testimony at hearings, we would have biased
our results toward greater institutional influence on policy coherence. As such, we sought an
independent designation. The designated lead agencies are: agriculture (Department of Agriculture),
Appalachian (Appalachian Regional Commission), aviation (Federal Aviation Administration), chil-
dren (none), the disabled (Social Security Administration), education (Department of Education),
environment (Environmental Protection Agency), family (Department of Health and Human Ser-
vices), the Great Lakes (none), the homeless (Department of Housing and Urban Development),
immigration (Immigration and Naturalization Service), older Americans (Department of Health and
Human Services), railroad (Federal Railroad Administration), rural (none), transportation (Depart-
ment of Transportation), urban (Department of Housing and Urban Development), veterans (Vet-
erans Administration), and women (none).
7. These are only for those domains that have designated responsible bureaus. The natural log of
agency involvement is used to meet linearity assumptions.
8. This is the Pearson correlation between overall policy coherence scores and the proportion of public
laws in each domain that evidence no or little change. We exclude the Great Lakes policy domain
in this calculation because it only contained six public laws (the lowest of any domain) and was an
outlier in the scatterplot depicting this relationship.
9. This is the Pearson correlation between overall policy coherence scores and the proportion of public
laws that contain symbolic or nonsubstantive provisions in each domain. We also exclude the Great
Lakes policy domain from this comparison. The square-root transformation is employed for the
proportion of laws that are symbolic in order to meet linearity assumptions for the correlation and
homogeneity of variance assumptions for comparisons among groups.

References

Baumgartner, Frank R., and Bryan D. Jones. 1993. Agendas and Instability in American Politics. Chicago:
University of Chicago Press.
Baumgartner, Frank R., Bryan D. Jones, and John D. Wilkerson. 2002. “Studying Policy Dynamics.” In
Policy Dynamics, ed. Frank R. Baumgartner, and Bryan D. Jones. Chicago: University of Chicago
Press, 29–46.
Berry, Jeffrey M. 1989. “Subgovernments, Issue Networks, and Political Conflict.” In Remaking American
Politics, ed. Richard Harris and Sidney Milkis. Boulder, CO: Westview Press, 239–60.
Birkland, Thomas A. 1997. After Disaster: Agenda Setting, Public Policy, and Focusing Events. Washington,
DC: Georgetown University Press.
Bleakley, Timothy, and Robert G. Latoff. 2004. “Rethinking U.S. Energy Policy.” The McKinsey Quarterly
13 (3): 50–60.
Bogenschneider, Karen. 2000. “Has Family Policy Come of Age? A Decade Review of the State of U.S.
Family Policy in the 1990s.” Journal of Marriage and Family 62 (4): 1136–59.
402 Policy Studies Journal, 34:3

Bonnen, James T. 1992. “Why Is There No Coherent U.S. Rural Policy?” Policy Studies Journal 20 (2):
190–201.
Bradshaw, Michael. 1992. The Appalachian Regional Commission: Twenty-Five Years of Governmental Policy.
Lexington, KY: University of Kentucky Press.
Browne, William P. 1995. Cultivating Congress: Constituents, Issues, and Interests in Agricultural Policy-
making. Lawrence, KS: University Press of Kansas.
———. 2001. The Failure of National Rural Policy, Institutions and Interests. Washington DC: Georgetown
University Press.
Burstein, Paul. 1991. “Policy Domains: Organization, Culture, and Policy Outcomes.” Annual Review of
Sociology 17: 327–50.
Dilger, Robert Jay. 2003. American Transportation Policy. Westport, CT: Prager Lawrence.
Fuhrman, Susan H. 1994. “Clinton’s Education Policy and Intergovernmental Relations in the 1990s.”
Publius, The Journal of Federalism 24 (3): 83–97.
Gais, Thomas L., Mark A. Peterson, and Jack L. Walker. 1984. “Interest Groups, Iron Triangles and
Representative Institutions in American National Government.” British Journal of Political Science 14
(2): 161–85.
Gelb, Joyce, and Marian Lief Palley. 1996. Women and Public Policies: Reassessing Gender Politics.
Charlottesville, VA: University Press of Virginia.
Goggin, Malcolm L., Ann O’M. Bowman, James P. Lester, and Laurence J. O’Toole, Jr. 1990. Implemen-
tation Theory: Toward a Third Generation. Glenview, IL: Scott, Foresman and Company.
Heitshusen, Valerie. 2000. “Interest Group Lobbying and U.S. House Decentralization: Linking Infor-
mation Focus to Committee Hearing Appearances.” Political Research Quarterly 53 (1): 151–76.
Jenkins-Smith, Hank C., Gilbert K. St. Clair, and Brian Woods. 1991. “Explaining Change in Policy
Subsystems: Analysis of Coalition Stability and Defection Over Time.” American Journal of Political
Science 35 (4): 851–80.
Jordan, Grant. 2005. “Bringing Policy Communities Back In? A Comment on Grant.” British Journal of
Politics and International Relations 7 (3): 317–21.
Kingdon, John W. 1995. Agendas, Alternatives and Public Policies, 2nd ed. New York: Harper Collins
Publishers.
Koschinsky, Julia, and Todd Swanstrom. 2001. “Confronting Policy Fragmentation: A Political Approach
to the Role of Housing Nonprofits.” Policy Studies Review 18 (4): 111–27.
Krehbiel, Keith. 1991. Information and Legislative Organization. Ann Arbor: University of Michigan Press.
Laumann, Edward O., and David Knoke. 1987. The Organizational State: Social Change in National Policy
Domains. Madison: University of Wisconsin Press.
Majone, Giandomenico, and Aaron Wildavsky. 1979. “Implementation as Evolution.” In Implementation,
2nd ed., ed. Jeffrey L. Pressman and Aaron Wildavsky. Berkeley: University of California Press,
177–194.
May, Peter J. 1991. “Reconsidering Policy Design: Policies and Publics.” Journal of Public Policy 11 (2):
187–206.
May, Peter J., Bryan D. Jones, Betsi E. Beem, Emily A. Neff-Sharum, and Melissa K. Poague. 2005.
“Policy Coherence and Component-Driven Policymaking: Arctic Policy in Canada and the United
States.” Policy Studies Journal 33 (1): 37–63.
Mazmanian, Daniel A., and Paul A. Sabatier. 1983. Implementation and Public Policy. Glenview, IL: Scott,
Foresman and Company.
McFarland, Andrew. 2004. Neopluralism: The Evolution of Political Process Theory. Lawrence: University
Press of Kansas.
Meucci, Sandra. 1997. “What Is Children’s Policy, Anyway?” Social Justice 24 (3): 105–24.
Miller, Lisa L. 2004. “Rethinking Bureaucrats in the Policy Process: Criminal Justice Agents and the
National Crime Agenda.” Policy Studies Journal 32 (4): 569–88.
Nierenberg, Danielle. 2001. “U.S. Environmental Policy: Where Is It Headed?” World Watch 14 (4): 12–22.
May/Sapotichne/Workman: Policy Coherence and Policy Domains 403

Rabe, Barry G. 1996. “An Empirical Examination of Innovations in Integrated Environmental Manage-
ment: The Case of the Great Lakes Basin.” Public Administration Review 56 (4): 372–81.
Rabe, Barry G., and Janet B. Zimmerman. 1995. “Beyond Environmental Regulatory Fragmentation:
Signs of Integration in the Case of the Great Lakes Basin.” Governance: An International Journal of
Policy and Administration 8 (3): 58–77.
Richardson, Jeremy. 2000. “Government, Interest Groups and Policy Change.” Political Studies 48 (5):
1006–25.
Ripley, Randall B., and Grace A. Franklin. 1991. Congress, the Bureaucracy, and Public Policy, 5th ed.
Pacific Grove, CA: Brooks Cole Publishing.
Salisbury, Robert H. 1968. “The Analysis of Public Policy: A Search for Theories and Roles.” In Political
Science and Public Policy, ed. Austin Ranney. Chicago: Markham, 151–175.
Schneider, Anne L., and Helen Ingram. 1997. Policy Design for Democracy. Lawrence: University of
Kansas Press.
———. 1993. “Social Construction of Target Populations: Implications for Politics and Policy.” American
Political Science Review 87 (2): 334–47.
Stone, Deborah A. 1997. Policy Paradox: The Art of Political Decision Making. New York: W.W. Norton and
Company.
Talbert, Jeffery C., Bryan D. Jones, and Frank R. Baumgartner. 1995. “Nonlegislative Hearings and
Policy Change in Congress.” American Journal of Political Science 39 (2): 383–405.
Watson, Sara D. 1993. “Holistic Policymaking: ‘Neo-Liberalism’ as Illustrated by the Women’s and
Disability Rights Movements.” Policy Studies Journal 21 (4): 752–65.
Wells, Miriam J. 2004. “The Grassroots Reconfiguration of U.S. Immigration Policy.” International
Migration Review 38 (4): 1308–48.
Worsham, Jeff. 1998. “Wavering Equilibriums: Subsystem Dynamics and Agenda Control.” American
Politics Quarterly 26 (4): 485–513.
Zafonte, Matthew, and Paul Sabatier. 2004. “Short-Term Versus Long-Term Coalitions in the Policy
Process: Automobile Pollution Control, 1963–1989.” Policy Studies Journal 32 (1): 75–107.