This article was downloaded by: [Istanbul Universitesi Kutuphane ve Dok]

On: 23 December 2014, At: 22:43

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK

Public Management Review

Publication details, including instructions for authors
and subscription information:
http://www.tandfonline.com/loi/rpxm20

The logic of governance in

health care delivery
a b
Melissa Forbes , Carolyn J. Hill & Laurence E. Lynn
c
Jr
a
Doctoral Student, Ford School of Public Policy,
Department of Sociology, University of Michigan,
Weill Hall, 4th Floor, 735 State St, Ann Arbor, MI,
48109-3091, USA Phone: +1 734 764 3490 E-mail:
b
Assistant Professor, Georgetown Public Policy
Institute, Georgetown University, 3520 Prospect Street
NW, 4th Floor, Washington, DC, 20007, USA Phone: +1
202 687 7017 E-mail:
c
Sydney Stein, Jr. Professor of Public Management
Emeritus, University of Chicago, Chicago, IL, USA E-
mail:
Published online: 14 Dec 2007.

To cite this article: Melissa Forbes , Carolyn J. Hill & Laurence E. Lynn Jr (2007) The
logic of governance in health care delivery, Public Management Review, 9:4, 453-477,
DOI: 10.1080/14719030701726457

To link to this article: http://dx.doi.org/10.1080/14719030701726457

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the
information (the “Content”) contained in the publications on our platform.
However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or
suitability for any purpose of the Content. Any opinions and views expressed
in this publication are the opinions and views of the authors, and are not the
views of or endorsed by Taylor & Francis. The accuracy of the Content should
not be relied upon and should be independently verified with primary sources
 of information. Taylor and Francis shall not be liable for any losses, actions,
claims, proceedings, demands, costs, expenses, damages, and other liabilities
whatsoever or howsoever caused arising directly or indirectly in connection
with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan, sub-
licensing, systematic supply, or distribution in any form to anyone is expressly
forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014
 Abstract
THE LOGIC OF
A multi-level analytic framework termed a
‘logic of governance’ is used to identify GOVERNANCE IN
systematic patterns of health care governance
HEALTH CARE
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

from the findings of disparate research

studies. Using a subset of 112 studies on
health care service delivery, we use an
‘inside-out’ interpretive strategy to construct
DELIVERY
an empirical overview of health care govern- An analysis of the empirical
ance. This strategy incrementally aggregates
findings from studies of adjacent then of non- literature
adjacent levels of governance until a coherent
overall picture emerges. In general, the
choices of organizational arrangements, ad-
Melissa Forbes, Carolyn J. Hill and
ministrative strategies, treatment quality and Laurence E. Lynn, Jr
other aspects of health care services by Melissa Forbes
policy makers, public managers, physicians, Doctoral Student, Ford School of Public Policy,
and service workers, together with their Department of Sociology, University of Michigan,
values and attitudes toward their work, have Weill Hall, 4th Floor, 735 State St, Ann Arbor
significant effects on how health care public MI 48109-3091, USA, Tel: þ1 734 764 3490
policies are transformed into service-delivery E-mail: [email protected]
outputs and outcomes. Investigations that fail
to account for such mediating effects in
Carolyn J. Hill
research designs or in the interpretation of
Assistant Professor
results may provide inaccurate accounts of
Georgetown Public Policy Institute, Georgetown
how health care governance works.
University, 3520 Prospect Street NW, 4th Floor
Washington, DC 20007, USA
Tel: þ1 202 687 7017
E-mail: [email protected]
Laurence E. Lynn, Jr
Key words Sydney Stein, Jr. Professor of Public Management
Logic of governance, health care delivery, Emeritus, University of Chicago, Chicago,
governance, outcomes IL, USA
E-mail: [email protected]

Vol. 9 Issue 4 2007 453 – 477

Public Management Review ISSN 1471-9037 print/ISSN 1471-9045 online
Ó 2007 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/14719030701726457
 454 Public Management Review

An expanding international network of scholars has begun investigating questions of

public governance using the theories, models, methods, and data of the social and
behavioral sciences (Hill et al. 2005). Their objective is conceptual and empirical
understanding of the complex interrelationships among political institutions, public
organizations and management, and service-delivery processes, outputs and outcomes
in order to better orient public management policy making and practice toward the goal
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

of improved government performance.

Studies of public sector governance employ various definitions of governance (Hill
et al. 2005). Kooiman (1999) identifies common elements in these definitions as ‘the
emphasis on rules and qualities of systems, co-operation to enhance legitimacy and
effectiveness and the attention for new processes and public–private arrangements’.
Osborne (2006) creates a useful analytic framework within which to compare
definitions, arguing that a New Public Governance paradigm is emerging from which to
develop theory and research that will inform practice. The research reported below is
conceived in that spirit.
For purposes of empirical analysis, public sector governance has been defined as
‘regimes of laws, rules, judicial decisions, and administrative practices that constrain,
prescribe, and enable the provision of publicly supported goods and services’ through
formal and informal relationships with agencies in the public and private sectors (Lynn
et al. 2001: 7). More specifically, governance can be conceptualized as a system of
hierarchically ordered institutions which Lynn et al. (2000a, 2000b, 2001) have termed
a ‘logic of governance’. This particular conceptualization addresses how multiple layers
of governing institutions influence public sector outcomes in general and in specific
policy domains.
The result of this line of thinking is an analytic framework that can be used to
assist public management scholarship by suggesting how hierarchical levels of
institutions might be interrelated in explaining public service outputs and outcomes.
Investigators can use such a logic of governance to integrate findings from widely
dispersed but conceptually related literatures from various disciplines, fields, and
subfields. This type of analytic synthesis has the potential to reveal bigger pictures of
what is being learned about governance as well as facilitating inferences about the
effect of public management on performance in individual studies or specialized
literatures.
Previous analyses using a logic of governance (Forbes and Lynn 2005; Hill and Lynn
2005, hereinafter ‘FL’ and ‘HL’) focused primarily on how investigators approached the
study of governance and what kinds of models they used. The question motivating this
article is: Using a logic of governance as an analytic framework, what can we learn
about the governance and, specifically, about the management, of health care delivery
from empirical literatures that address various aspects of this issue? We have chosen
health care as a substantive focus for this inquiry because it is sufficiently broad, and
because health care studies are sufficiently numerous, to capture a variety of putative
causal relationships among different levels in the chain of delegation. Along the way, we
 Forbes et al.: The logic of governance in health care delivery 455

comment on similarities and differences with the delivery of education services to

suggest what cross-sector comparisons might look like.

AN ‘INSIDE-OUT’ ANALYTIC APPROACH

Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

Governance involves means for achieving direction, control, and coordination of

individuals and organizations on behalf of interests they have in common (Vickers 1983;
Wamsley 1990; Lynn et al. 2001). Any democratic governance regime both reflects and
shapes underlying political processes in which actors collectively choose institutional
arrangements intended to serve their interests. Referring to the political processes of
both parliamentary and presidential democracies, Strøm (2000: 266; cf. Lupia and
McCubbins 2000) describes a ‘chain of delegation’ in ‘contemporary democracies, from
voters all the way to civil servants that ultimately implement public policy . . . in which
those authorized to make political decisions conditionally designate others to make such
decisions in their name and place’. The result may be expressed in a set of hierarchical
interrelationships linking the institutional choices of policy makers to service delivery
through intervening levels of management.1
In general, reference to a logic of governance encompassing interactions between
different political/organizational levels is a reminder of the endogeneity of complex
governance processes and institutions. Toonen (1998: 248) argues that ‘reform and
change at one level of analysis presupposes certain conditions at other levels of analysis’.
This analytic approach, in his view, can accommodate ‘more subtle and differentiated
conceptualizations which allow us to go somewhat deeper into the actual operation of
the system instead of simply scratching the surface’ in comparative public
administration research (1998: 237).
In this article, the core interrelationships in the logic of governance are employed as
a framework to analyze health care studies (including occupational and public health
studies) chosen from the HL and FL databases. Specifically, we are concerned with
interrelationships between public policies, public management, service delivery, and
observed outputs/outcomes. We included studies that were specifically concerned with
health sector governance questions involving variables at two or more of these four
levels in the logic of governance. Examples of the types of variables include: structural
variables such as Medicare and Medicaid policies in the United States or policies
governing the UK Public Health Service, management variables such as those concerned
with physicians assuming managerial responsibility, service-delivery variables such as
those representing nurses’ beliefs and values, and outputs and outcomes such as access
to treatment and behavioral change among specific target populations.
Most individual studies do not incorporate multiple levels of explanation, and the
topics of those that do employ multiple levels are diverse. These realities of empirical
research produce findings that are often difficult to integrate or compare. For example,
as depicted in Figure 1, Study A may investigate the effects of policies on outputs or
 Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

456
Public Management Review

Figure 1: Modeling governance relationships

 Forbes et al.: The logic of governance in health care delivery 457

outcomes, Study B may investigate the effects of policies on public management, Study
C may investigate the effects of public management on service delivery, and Study D
may investigate the effects of service delivery on outcomes. These four studies,
moreover, may use different variables to operationalize management and outputs/
outcomes.
Because a straightforward linking of findings across such studies is seldom possible,
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

the logic of governance as an analytic tool can help tease out insights that would
otherwise be obscure. Using this logic, we employ what we call an ‘inside-out’ strategy
to synthesize the findings of diverse health sector studies. This strategy assumes that the
way the logic of governance actually operates is discernible through aggregating the
findings of disparate but sufficiently numerous individual studies. Along the way, we
also note exemplary individual studies with unusually broad implications.
The inside-out strategy works as follows. First, we group the health sector
studies by the types of independent and dependent variables and levels of
governance used in the study. For example, all studies that employ management-
level independent variables and service-delivery-level dependent variables constitute one
group, which we designate the ‘inside’ group because it concerns the primary direct
relationship between management and service delivery. Next we construct groups of
studies that incrementally move ‘outside’ this inner, management-explains-service-
delivery group to add levels of governance directly above management and below
service delivery. These studies use policy-level variables to explain management-
level variables and employ service-delivery-level variables to explain outputs/
outcomes. The next studies further ‘out’ skip a level: policies to service delivery
(skipping management), and management to outputs/outcomes (skipping service
delivery). Studies that skip two levels – examining how policies affect outputs/
outcomes (skipping both management and service delivery) – constitute the
outermost level of studies that we examine.
The inside-out method allows us to identify the different ways that variables are
being modeled, including whether variables at a given level are being used to differing
degrees as dependent or independent variables. Further, based on findings from studies
of adjacent levels of governance, it enables us to consider, first, whether studies that
exclude higher and lower levels in the hierarchy are seriously incomplete and, second,
whether studies that exclude intervening levels of governance might be omitting
important mediating effects. Thus, we can incrementally assemble the various pieces
until a multi-level picture emerges.

HEALTH STUDY DATA

The health-related studies utilized for the analysis in this article are from the HL and FL
logic of governance databases. The full HL database contains 823 studies (Hill and Lynn
2005), and the FL database contains 193 studies (Forbes and Lynn 2005). The studies in
 458 Public Management Review

the databases were identified by the authors from over seventy academic journals
covering the twelve-year period from 1990 to 2001 (inclusive).2 Articles were included
in the databases if they explicitly specified causal or reduced form relationships between
variables from two or more governance levels. The vast majority of studies in the HL
US database use quantitative methods, whereas over 40 percent of the FL international
database comprises qualitative studies. The difference in size and composition of these
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

databases reflects differences in research orientations and methodologies among US and

international management scholars.
The terms ‘c-level’, ‘d-level’, ‘e-level’, and ‘f-level’ are used throughout this section
and in the tables cited below. They refer to hierarchical levels in the logic of governance
as described in Lynn et al. (2000a, 2000b, 2001), where the levels are: (a) citizen
preferences and interests; (b) public choice; (c) policy/structures of formal authority;
(d) discretionary management, organization, and administration; (e) service delivery
and core technologies; (f) outputs and outcomes; (g) stakeholder assessments of
performance.
For the present study, we selected the subset of all health-related studies in the
combined FHL database for analysis. Our selection included all health studies that
examined dependent variables at the management (d-level), service delivery (e-level),
or outputs/outcomes levels (f-level), resulting in a health subsample of 112 studies.
Thirty-three percent (n ¼ 37) of the studies employ non-US data sources and almost
one-fourth (n ¼ 26) rely on qualitative methods. A list of the journals and the number
of health articles drawn from them is listed in the Appendix.
Within these 112 studies, 268 specific findings were reported; in most cases, these
were findings emphasized by the authors. A ‘finding’ is, for example, ‘A Maryland law
setting a minimum maternity hospital stay increased charges’ (Udom and Betley 1998);
‘Competitive contracting for home care services reduced efficiency’ (Abelson et al.
2004); and ‘Retrospective drug utilization review had no relationship with drug
expenditures per recipient’ (Moore et al. 2000). As shown in Table 1, the majority of
the health findings – 146 – have an output/outcome as the dependent variable. Service
delivery is the dependent variable in 75 findings, while management is the dependent
variable in 47 findings.
It is widely believed that publication bias leads to a greater representation of studies
with statistically significant findings than are actually found by researchers: studies that
feature findings of ‘no effect’ are less likely to be published. In the studies we
examined, we found that the rate of null findings was around 20 percent. Thus, most of
the findings have passed some kind of significance test or criterion.3
Additionally, the findings we summarize are not always causal. Most findings are
drawn from quantitative, non-experimental studies. Many include a number of control
variables, which reduce but do not eliminate the possibility that the observed
relationships are spurious. The sheer volume of the evidence, however, is suggestive of
relationships that – even if correlational, not causal – provide insights for policy makers
and managers.
 Forbes et al.: The logic of governance in health care delivery 459

Table 1: Number of findings at each sublevel of management, service delivery, and outputs/outcomes

Levels and sublevels of dependent variables Number

(d) Management
(1) Administrative structures 23
(2) Tools 3
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

(3) Values & strategies 21

Subtotal 47
(e) Service delivery
(1) Program design features 0
(2) Field worker discretion 0
(3) Field worker beliefs/values 3
(4) Admin processes & procedures 9
(5) Work/treatment/intervention 17
(6) Client influence/behavior/preferences 6
(7) Use of resources and/or performance 40
Subtotal 75
(f) Outputs and outcomes
(1) Govt/public sector outputs 8
(2) Market/firm/private sector outputs 24
(3) Individual/society outputs 86
(4) Govt/public sector outcomes 4
(5) Market/firm/private sector outcomes 2
(6) Individual/society outcomes 22
Subtotal 146
Total number of findings at d, e, and f levels 268

Note: This table does not include findings that use a-, b-, f-, or g-level independent variables, and findings that model
governance from the ‘bottom–up’ (e.g. findings that use an e-level independent variable and a d-level dependent variable).

PUBLIC GOVERNANCE IN THE HEALTH CARE LITERATURE

The health care studies in our subset do not attempt to explain all aspects of governance
at the management, service-delivery, and output/outcome levels. The uneven coverage
of sublevels, shown in Table 1, suggests that we know more about the determinants or
correlates of some aspects of governance than about others. For example, the most
commonly used management (d) dependent variables were:

. administrative structures: for example, diversification of services, converting or

closing of hospitals; and
 460 Public Management Review

. values and strategies: for example, administrative or operational expenses;

entrepreneurialism, participatory decision making, normative legitimation or
endorsement.

At the service-delivery (e) level, the most commonly used dependent variables were:

.
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

resource use and ‘performance’, economic and otherwise: for example, hospital
efficiency or profitability, clinical effectiveness, total or per-patient costs,
expenditures; and
. characteristics of work, treatment, intervention, or services: for example, length
of stay, Cesarean deliveries, and placement of patients on waiting lists.

Finally, the most commonly used results-oriented (f) dependent variables were:

. individual or societal outputs: for example, number of events or procedures

(such as births, knee replacements, prenatal care visits, low-birthweight births,
emergency room visits), adequacy of care, access to care, total expenditures;
. business or private sector outputs: for example, profits, bed supply,
expenditures; and
. individual or societal outcomes: for example, mortality and morbidity rates.

In summary, empirical health care governance research is primarily concerned with

explaining only certain aspects of governance within the management, service-delivery,
and output/outcome levels.
The types of independent variables being used to explain the dependent variables are
similarly uneven. Table 2 shows the general types of independent variables used to
explain these dependent variables. We can see, for example, that policy (c-level)
explanations are by far the most common for any of the three types of dependent
variables. Over 80 percent of the management findings used formal structures as an
explanatory factor, along with almost two-thirds of the service-delivery findings and
almost 60 percent of the output/outcome findings. The table also indicates that within-
level explanations were rarely employed.
The modeling patterns among the health studies are not necessarily generalizable
to governance studies in other subfields, however. Among the studies of education
governance in the FHL database, for example, about the same percentage of studies
used policy/formal structures as used within-level management variables as
explanatory factors to model management-level dependent variables. Additionally,
out of 67 service-delivery findings in the education subset, policy explanations were
employed in only slightly over one-third of the cases, compared to two-thirds in the
health subset. Out of 76 output/outcome findings in the education subset, formal
structures are used in just over 10 percent of the cases, compared to 60 percent in
the health subset.
 Forbes et al.: The logic of governance in health care delivery 461

Table 2: Aggregate-level logic-of-governance relationships in the health subset (cells show number of
findings)

Level of dependent variable

Level of independent variable (d) Management (e) Service delivery (f) Outputs/outcomes
(c) Public policy 38 49 84
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

(d) Management 9 20 25
(e) Service delivery – 6 37
Total # findings at this level 47 75 146

Note: This table does not include findings that use a-, b-, f-, or g-level independent variables, and findings that model
governance from the ‘bottom–up’ (e.g. findings that use an e-level independent variable and a d-level dependent variable).

Comparing results from the health and education subsets highlights differences
among fields both in governance structures and processes, and in the emphases that
researchers place on different levels of governance in their modeling. In education
research, greater emphasis is placed on management’s effect on service delivery and on
outputs/outcomes, whereas policy/formal structure tends to be more heavily
emphasized in health care research. (This aspect of the education literature may be
changing, however, with the introduction of the No Child Left Behind Act in the
United States, which explicitly promotes greater large-scale data collection on
education outcomes.)

DIRECT GOVERNANCE

More specific insights concerning health care governance and management can be
obtained by employing the inside-out analytic strategy described earlier.
The three innermost groups of health care studies, which model interactions at
adjacent levels, are concerned with relationships between management and service
delivery (d to e), service delivery and outputs/outcomes (e to f), and public policies and
management (c to d). We first consider the twenty findings concerning management
explanations for service delivery. The primary sublevel of service delivery being
modeled, accounting for just over half of the findings in this group, is concerned with
performance and the use of resources (top panel, Table 3). The remaining studies in this
subset focus primarily on service-delivery-level policies and processes as well as field-
level employee attitudes and beliefs. Unlike the distribution for all studies shown in
Table 1, studies in this subset tend not to examine dependent variables of work/
treatment/intervention.
Two kinds of management-to-service-delivery explanations are prominent, as shown
in Table 3. Half of the twenty findings use a management structure explanation. Boyne
 Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

462
Table 3: Explaining service delivery using different governance levels (cells show number of findings)

Service-delivery (e) sublevels (dependent variable)

(6) (7)
(1) (2) (3) (4) (5) Client Use of
Program Field Field Admin Work/ influence/ resources
design worker worker processes & treatment/ behavior/ and/or Row
features discretion beliefs/values procedures intervention preferences performance total
Public Management Review

Management independent variable (d-level)

(1) Administrative structures 2 1 1 6 10
(2) Tools 2 2 4
(3) Values & strategies 1 2 3 6
Total d-to-e findings 0 0 3 4 1 1 11 20
Public policy independent variable (c-level)
(1) Hierarchy/structure
(i) Type of ownership 10 6 16
(ii) Level/type of govt 1 1 2
(iii) Internal govt entities 1 3 4
(iv) Political atmosphere 0
(2) Mandated behavior 0
(3) Policy design and elements 3 6 3 12 24
(4) Fiscal situation 1 1 2
(5) Other 1 1
Total c-to-e findings 0 0 0 4 16 5 24 49
 Forbes et al.: The logic of governance in health care delivery 463

et al. (2003), for example, devote an entire chapter of their public management reform
book to the relationship between New Public Management (NPM) health care reforms
in the UK and the efficiency, responsiveness, and equity of service delivery. Another 30
percent use ‘management values and strategies’ explanations, such as trust or expertise
(for example, Cole 2003). ‘Management tools’ explanations are used least often (as
they are in the equivalent education subset). Although the sample of health care findings
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

in this subset is not large, these findings provide a positive indication that decisions and
actions at the managerial level affect health service delivery in various ways. Only five
of the twenty findings were ‘null’ results – that is, the independent variable showed no
statistical relationship with the dependent variable. Given this, we infer that managerial-
level variables, especially the structural arrangements chosen by managers and
managers’ values and strategies, do affect service delivery.
We next consider the interrelationships between service delivery and outputs and
outcomes (e to f) based on the subset of thirty-seven findings (top panel, Table 4). The
most common dependent variables in this subset are individual/society outputs (twenty-
three of the thirty-seven findings). The most frequent type of service-delivery variables
used to explain these outputs are aspects of the treatment or intervention, such as
length of stay or placement of patients on waiting lists. Because of their large number,
the findings for this particular governance relationship drive the overall results that we
observe for the more general service-delivery-to-outcomes/outputs subset of studies.
The remaining health studies are primarily concerned with the way in which program
design features affect individual outputs/outcomes such as the number of activities or
procedures performed. In education studies, by contrast, researchers are about equally
as likely to use program design features, aspects of work/treatment/intervention, or
use of resources (such as teacher/student classroom ratio) to explain outputs/
outcomes. Based on a larger number of findings in this group of studies, we can
conclude that certain aspects of service delivery, especially treatment characteristics,
have an influence on health outputs and outcomes.
Up to this point, we have examined two adjacent subsets of findings: management-
to-service (top panel, Table 3), and service-delivery-to-outputs/outcomes (top panel,
Table 4). By combining information from both of these inquiries, we can check whether
the same types of service-delivery variables are being explained (by management-level
variables) as are used to explain outputs/outcomes. The third and fifth columns of
Table 5 show the distribution of service-delivery-level findings. In addition to the
number of findings shown earlier, this table also reports the proportion of findings at
each sublevel. For example, eleven studies in the management-to-service-delivery
subset examined a dependent variable depicting use of resources, over half of the
twenty studies in that subset.
In fact, the studies that attempt to explain service delivery use different service-
delivery-level variables than studies that use service delivery to explain outputs/
outcomes.4 Findings produced from studies in the former set tend to model
performance/uses of resources (55%), administrative processes and procedures (20%),
 Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

Table 4: Explaining outcomes using different governance levels

464
Output/outcome (f) sublevels (dependent variable)

(1) (2) (3) (4) (5) (6)

Public Private Indiv./ Public Private Indiv./
sector sector society sector sector society Row
outputs outputs outputs outcomes outcomes outcomes total

Service-delivery independent variable (e-level)

(1) Program design features 7 2 9
(2) Field worker discretion 0
Public Management Review

(3) Field worker beliefs/values 2 2

(4) Admin processes & procedures 2 1 3 6
(5) Work/treatment/intervention 14 3 17
(6) Client influence/behavior/preferences 2 1 3
(7) Use of resources and/or performance 0
Total e-to-f findings 0 2 23 3 0 9 37
Management independent variable (d-level)
(1) Administrative structures 1 2 2 1 6
(2) Tools 4 4 4 12
(3) Values & strategies 5 1 1 7
Total d-to-f findings 1 6 11 1 0 6 25

Public policy independent variable (c-level)

(1a) Type of ownership 7 2 9
(1b) Level/type of govt 0
(1c) Internal govt entities 0
(1d) political atmosphere 0
(2) Mandated behavior 3 1 2 2 8
(3) Policy design and elements 7 13 42 3 65
(4) Fiscal situation 1 1
(5) Other 1 1
Total c-to-f findings 7 16 52 0 2 7 84
 Forbes et al.: The logic of governance in health care delivery 465

Table 5: Overlap of service-delivery sublevels when they are used as dependent or as independent variables

C-to-e level findings D-to-e level findings E-to-f level findings

(e is dependent (e is dependent (e is independent
variable) variable) variable)
Service-delivery (e-sublevel)
variables Number Proportion Number Proportion Number Proportion
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

(1) Program design features 0 – 0 – 9 0.24

(2) Field worker discretion 0 – 0 – 0 –
(3) Field worker beliefs/values 0 – 3 0.15 2 0.05
(4) Admin processes & procedures 4 0.08 4 0.20 6 0.16
(5) Work/treatment/intervention 16 0.33 1 0.05 17 0.46
(6) Client influence/behavior/ 5 0.10 1 0.05 3 0.08
preferences
(7) Use of resources and/or 24 0.49 11 0.55 0 –
performance
Totals 49 1.00 20 1.00 37 0.99

Note: Proportions may not sum to 1.00 due to rounding. Chi-square tests of independence indicate that the distribution of
e-level variables shown in each of these columns is statistically different from the distribution in each of the others:
p 5 .01 for c-to-e and d-to-e distributions;
p 5 .001 for c-to-e and e-to-f distributions;
p 5 .001 for d-to-e and e-to-f distributions.

and employee beliefs and values (15%). We see limited overlap with the service-
delivery variables used to explain outputs/outcomes: here, the explanations are, in
order of frequency, characteristics of the treatment or intervention (46%), program
design features (24%), and administrative processes and procedures (16%). While these
results should be viewed with caution because of the small number of findings, they do
suggest a disjunction between our understanding of what aspects of service delivery are
influenced or determined by higher-level institutions and what aspects of service
delivery in turn affect eventual outcomes. An implication of significant differences in
variable distributions is that researchers who are attempting to explain phenomena
at different levels of governance are, in effect, employing different theoretical
orientations.
The third type of direct governance relationship we consider addresses how public
policy influences management (c to d-level). Our health study subset contains thirty-
eight such findings (Table 6). Hierarchical/structural factors, especially those
concerning ownership status, are as likely to be used as an explanatory factor as
policy design or policy elements such as Medicaid or Medicare policies or NPM
reforms. Together, these two types of public policies constitute 84 percent of the
explanatory factors for this subset. D’Aunno et al. (2000), for example, found that
hospital ownership made some difference in the likelihood that a hospital would make
 466 Public Management Review

Table 6: Indirect governance linking public policy to management (cells show number of findings)

Management dependent variable (d-level)

Public policy (1) (3)

independent Administrative (2) Values & Row
variable (c-level) structures Tools strategies total
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

(1) Hierarchy/structure
(i) Type of ownership 12 4 16
(ii) Level/type of govt 1 1
(iii) Internal govt entities 3 2 5
(iv) Political atmosphere 0
(2) Mandated behavior 0
(3) Policy design and elements 1 3 12 16
(4) Fiscal situation 0
(5) Other 0
Total of c-to-d findings 16 3 19 38

divergent changes in the services they offer. Publicly owned hospitals were significantly
less likely to convert their services than private non-profit hospitals, but private non-
profits did not differ from for-profit hospitals in rates of conversion. This finding
supports the commonly held view that public organizations are fraught with more
bureaucracy than private organizations, which may cause them to resist structural
change. However, Boyne’s (2003) review of thirty-four public management studies did
not reveal definitive empirical evidence supporting the public management beliefs that
public organizations display greater bureaucracy, more red tape, and less managerial
autonomy than private organizations.
Although ownership and policy design are the primary explanatory factors in this
subset, they are differentially used in modeling management sublevels. The most
common predictors of administrative structures chosen by managers (such as the
conversion or closing of hospitals or diversification of services) is the type of ownership,
but type of ownership is used to explain only four out of nineteen findings related to
managerial values and strategies. Instead, the determinants of managerial values and
strategies are primarily policy design variables such as National Health Service reforms
(in Great Britain) and Medicare’s capital prospective payment system (in the United
States). These public policy variables accounted for twelve out of nineteen findings
related to managerial values and strategies. Given the attention to privatization and
other NPM reforms in recent years, it is not surprising that ownership (predominantly
of hospitals) is of interest to health policy researchers.
As with the comparison of adjacent subsets of findings above, we can now consider
the overlap in use of management-level variables, that is, whether the same types of
 Forbes et al.: The logic of governance in health care delivery 467

management variables are being explained (c to d) as are used to explain (d to e and f-

levels) (see Table 7).
Both studies that use management as a dependent variable and studies that use
management as an independent variable tend to focus on administrative structures and
values and strategies. Proportionally, management-level dependent variables tend to
fall in the values and strategies subcategory somewhat more (in 50 percent of the
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

findings) than do studies that use these variables in explanatory models (28 and 30
percent of the findings). The lack of overlap for management-level variables shown in
this table is not as marked, however, as service-delivery variables (d to e and e to f)
shown in Table 5.5

Health care governance: A first approximation

A story is beginning to emerge from studies of direct governance in health care. Both
the administrative structures chosen by public managers and managerial values and
strategies affect service delivery. In particular, these uses of managerial discretion affect
patterns of resource use and the efficiency with which services are delivered. These
aspects of public management are themselves influenced by the structures of formal
authority in which public managers operate. It is reasonable to expect, in other words,
that the public management and service-delivery levels of governance play mediating
roles between the authoritative decisions of policy makers and the outcomes and
outputs of the services they authorize, although the importance of these roles
will depend on the extent and significance of the discretion available to managers and
service workers.

Table 7: Overlap of management sublevels when they are used as dependent or as independent variables

C-to-d level findings D-to-e level findings D-to-f level findings

(d is dependent (d is independent (d is independent
variable) variable) variable)
Management variable
(d-level) Number Proportion Number Proportion Number Proportion

(1) Administrative structures 16 0.42 10 0.50 6 0.24

(2) Tools 3 0.08 4 0.20 12 0.48
(3) Values & strategies 19 0.50 6 0.30 7 0.28
Totals 38 1.00 20 1.00 25 1.00

Note: Chi-square tests of independence indicate that the distribution of e-level variables shown in each of these columns is not
statistically different for the first two types, but is statistically different (marginally so in one case) for the others:
p ¼ .22 for c-to-d and d-to-e distributions;
p 5 .01 for c-to-d and d-to-f distributions;
p ¼ .101 for d-to-e and d-to-f distributions.
 468 Public Management Review

The story is more subtle, however. Some of those aspects of service delivery affected
by management may affect other aspects – for example, frontline worker attitudes may
affect their treatment of clients, which, in turn, directly affect outputs and outcomes.
Michael Lipsky’s (1980) research on street-level bureaucracy suggests that peer
assessment can have a controlling influence on frontline-worker behavior. Thus,
frontline workers’ attitudes toward a management decision might well affect how these
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

workers implement that decision. For example, Kendrick et al. (1995) studied whether
the design of a Smoking Cessation in Pregnancy (SCIP) Project affected smoking quit
rates among low SES women and found no statistically significant difference in quit rates
between participants in intervention clinics and non-participants. Although the
comprehensiveness of the intervention was not explicitly considered as a variable in
the study, the authors’ discussion of their findings concedes that the program design –
which required nurses both to deliver the intervention and to collect data from
participants – affected their ability to implement the program as it was designed.
To further illustrate how these interactions might work, we utilize specific examples
from the health studies in our database, beginning with Doolin’s (2001) study of
doctors serving as clinical directors in New Zealand. This NPM study found that
organizational restructuring of New Zealand health care enterprises and the devolution
of managerial control to senior clinicians, because it was resisted by practicing
clinicians, caused almost no change in administrative processes and policies at the
service-delivery level. The primary mode of operation in clinical units continued to be
based on professional and collegial relations.
Now consider a study that employs a service-delivery independent variable to explain
client outcomes. The most common subject matter for this type of study in our health
subset is the impact of prenatal programs on client behavior and birth outcomes. Joyce
(1999), for example, found no statistically significant effect on birthweight or smoking
reduction among pregnant women who participated in a Medicaid Prenatal Care
Assistance Program. Additionally, there was only a small positive effect of program
participation on the mother’s number of prenatal care visits.
While these two studies differ in their subject matter and even in the country in
which they were conducted, the levels modeled provide insight into broader
governance relationships. Doolin (2001) is essentially interested in how administrative
changes introduced by management affect frontline worker attitudes and activity. What
if, instead of NPM-related changes, a study had examined how administrative changes in
Medicaid affected the attitudes and practices of doctors and nurses? Through this
thought exercise, we can infer part of the reason why Joyce (1999) found no effects of
Medicaid’s Prenatal Care Assistance Program. The failure might be a result of negative
attitudes on the part of doctors and nurses or of administrative policies and practices
that were caused by management actions that were not included in the study’s model.
In turn, these attitudes could make the program less accessible or appealing to pregnant
women on Medicaid, who might not have changed their behavior, resulting in less
positive birth outcomes for their babies.
 Forbes et al.: The logic of governance in health care delivery 469

INDIRECT GOVERNANCE

Having established that the various levels of governance may well play a mediating role
within the logic of governance, we are now in a position to introduce higher-level
complexities by considering studies that skip one or more potentially mediating levels
of governance: findings about how formal authority influences service delivery (c to e),
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

findings about how management affects outputs or outcomes (d to f), and findings about
how public policies affect results (c to f). Because these findings skip a level of the logic,
we refer to them as ‘indirect’ governance.

Is management needed to implement policies?

The lower panel of Table 3 shows detailed information on the forty-nine policy-to-
service-delivery (c to e) findings. Consistent with the overall pattern in Table 1, the
main type of service-delivery dependent variable in this subset examines performance
or resource use such as hospital efficiency or profitability, clinical effectiveness, and
expenditure levels. Next most commonly explained are aspects of work, treatment, or
intervention such as length of stay and placement of patients on waiting lists.
The primary variables used to explain service delivery are type of ownership and
policy designs and policy elements. The predominance of policy design independent
variables may be in part a result of the disproportionately large number of health studies
focusing on Medicaid and Medicare policy in the United States. Together, these two
sublevels account for over 80 percent of the types of formal structure variables used to
explain service delivery. Yet Table 3 also shows that the patterns of their use in
explaining service delivery are somewhat different: type of ownership is used relatively
more often to explain aspects of the treatment or intervention (ten of the sixteen
findings), while policy design is emphasized relatively more often in explanations of
resource use or performance at the service-delivery level (twelve of twenty-four
findings).
The health studies that use ownership status to model changes at the service-delivery
level provide another opportunity to apply an inside-out approach to understanding
how ownership operates within the logic of governance. All of the health studies that
use ownership status to explain changes at the service-delivery level (but exclude
management variables) either employ dependent variables of treatment/intervention or
use of resources/performance. One study, Ettner and Hermann (2001), employs both
types of dependent variables. They find that the length of hospital stay for elderly
Medicare beneficiaries is longer in non-profit hospitals than in for-profit hospitals, but
there is no statistically significant relationship between hospital ownership status and
average total costs of care.
To infer omitted variables that might mediate the relationship between ownership
and service delivery, we can turn to d-level management factors. As discussed
 470 Public Management Review

previously, many of the health studies that employ administrative structure as an

independent variable focus on the efficiency of service delivery. Thus, it is reasonable to
suppose that administrative structures in particular may mediate the effect between
ownership and performance. Several of the health studies in the dataset employ both
policy and management-level independent variables to explain service-delivery
performance changes (for example, Holahan et al. 1991; Shi 1996; Ruef and Scott
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

1998), lending further support to the claim that management variables cannot be
ignored when studying the effect of ownership on service delivery, much less on
outputs and outcomes.
Many studies of indirect governance either do not explicitly consider the influence
of intervening levels or do not control for them. Health studies that employ policy
design independent variables to explain service delivery, but omit management
variables, examine changes in a wider variety of frontline phenomena, and some even
examine multiple outcomes at the front line. Udom and Betley (1998), for example,
found that Maryland’s minimum maternity hospital stay policy increased the average
length of hospital stay for mothers, but average charges for vaginal and cesarean
deliveries increased only slightly. Similarly, Evans et al. (2001) examine how
management accounting control techniques – in the form of physician profiling that
measures their patients’ average length of stay – influence the number of procedures
performed per patient day and patient length of stay. The question that naturally
arises with such studies is the extent to which management factors might have had a
mediating effect on service-delivery outcomes had they been included in the research
designs.
Because most health studies employing policy design elements focus on changes in
managerial values and strategies, we consider specific ways in which managerial
strategies and values are used as independent variables to explain service-delivery
changes. Succi and Alexander (1999), for example, found that management efforts to
include physicians in hospital management and governance led to a decline in hospital
efficiency. However, Surender et al. (2002) found that managers’ positive beliefs about
the strength of medical evidence increased clinical effectiveness of evidence-based
medicine projects in Wales. While these studies address different service-delivery
outcomes, they highlight a parallel insight: managerial values and strategies influence the
way services are delivered.
Applying the logic of governance framework to these studies makes it is possible to
propose instances when management variables are likely to have mediating effects on
causal relationships between non-adjacent levels of governance, making their omission
problematic. It is also helpful in identifying research models where management
variables are less likely to play a mediating role, and where their omission may not
invalidate findings. The Udom and Betley (1998) and Evans et al. (2001) studies are
illustrative of this distinction. While both studies use policy design independent
variables, the directness of the causal links between the independent variables and
service delivery are quite different.
 Forbes et al.: The logic of governance in health care delivery 471

In the Udom and Betley study, the link between a policy design requiring mandatory
hospital stays for new mothers and actual length of stay seems relatively clear. Setting a
mandatory minimum length of stay for mothers increases the overall average length of
stay. Physicians and hospital managers do not have authority to discharge new mothers
before the mandatory minimum length of stay has been reached. In contrast, in the
Evans et al. (2001) study, the relationship between physician profiling, the number of
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

operations performed, and changes in length of stay is arguably less direct. It is possible
that management values play a mediating role in decisions regarding the number of
procedures to conduct each day. Regardless of physician profiling, hospital managers
may feel that there are limits to the number of daily procedures doctors can safely
perform, thus eliminating some of the effect of profiling on number of procedures.
Further, this mediating effect might be influenced by type of ownership.
While intermediate levels in the logic of governance do not always play a mediating
role, these two studies suggest that the mediating role cannot simply be assumed away.
Researchers studying the effect of public policies on service-delivery outcomes ought to
at least be mindful that management can have mediating effects on causal relationships
between policy designs and service delivery when there is room for managerial
discretion in implementing the design. Public management theories about relationships
between policy design and service delivery should accommodate the possibility of
managerial influences that are not being adequately incorporated into research designs.
As Boyne (2003) has argued elsewhere, public management scholars need to make a
more concerted effort to model indirect effects when appropriate.

Management: Results oriented?

A second type of indirect governance is the influence of management on the outputs or

outcomes of public agencies (d to f-level), with twenty-five such findings in our dataset.
This type of finding completes the loop between the management, service delivery, and
outcomes levels of governance. Table 4 shows that, consistent with the overall pattern,
the most commonly modeled type of dependent variable in this subset comprises
individual/societal outputs. Both individual outputs and outcomes tend to be explained
most often by managerial tools, with the remaining studies split about equally between
explanations using administrative structures and managerial values and strategies. For
comparison purposes, we note that education studies overwhelmingly use values and
strategies as the primary explanation for outputs and outcomes.
Another question is whether the same types of management-level variables are being
used to explain both service-delivery variables and outputs/outcomes in the health
studies. The last columns of Table 7 address this question. We see some evidence that
management-level variables used to describe service-delivery outcomes (d to e) tend to
emphasize administrative structures (50% of the studies), while management-level
variables used to describe outputs/outcomes tend to emphasize tools (48%).6
 472 Public Management Review

The fact that the management-to-outputs/outcomes subset consists of only twenty-

five findings (just a few more than the management-to-service-delivery category)
implies a relative lack of emphasis on the direct influence of management on client
outcomes. In contrast, education studies frequently model student performance as
influenced by management variables, such as principals’ decisions or strategies. One
explanation for this difference between health and education studies might reflect
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

differences in the governance relationships of interest to the two research communities.

While many of the health studies examine large-scale individual trends using national
datasets, the education literature more often focuses on the school as the unit of analysis
and studies smaller geographic units given the highly localized nature of the US
education system. Studies in our data set of management effects on outcomes in
education are primarily concerned with how the number of teachers of color and
teacher empowerment affects student outcomes (e.g. Meier 1993; Weiher 2000;
Nielson and Wolf 2001).
The health studies that examine how management might affect program outcomes
usually include service-delivery variables as well; that is, their indirect governance
specification masks the fact that they often account for intervening factors. These
service-delivery variables are usually not modeled hierarchically with the management
variables, though, so possible interactions between the two levels are not considered.
The relatively small number of health studies that model relationships between
management and outcomes while omitting service-delivery variables makes it difficult
to argue conclusively that researchers examining the effect of management on outcomes
should also include service-delivery variables in their analyses. It is not clear whether
health studies use different-level variables in their models because they are modeling
different types of governance relationships or if researchers are omitting important
mediating or higher-level variables from their analysis for other reasons, such as data
limitations.

Are policies self-enforcing?

We now consider the outermost group of studies, those that examine the relationships
between public policies and service-delivery results (c to f). The bottom panel of
Table 4 summarizes the relationships in the eighty-four findings of this type. Because
these types of findings skip two levels in the logic (management and service delivery),
the potential for biased findings due to omitted mediating effects increases. Once again,
the most frequent dependent variables being modeled are individual/society outputs,
such as the number of events (for example, births), adequacy of care, access to care,
and total expenditures. We also observe a fair number of private sector outputs in the
models, such as profits, bed supply, and expenditures. Most of the findings (sixty-five
out of eighty-four) were explained by formal structures such as eligibility or program
features of state or federal programs (e.g. Medicaid, Medicare and National Health
 Forbes et al.: The logic of governance in health care delivery 473

Insurance programs), managed care or fee-for-service options in insurance plans, or

health sector reforms at the national or state levels.
This subset of public policy-to-outputs/outcomes (c to f) findings constitutes the
greatest degree of indirect governance we consider in this analysis. As such, these
findings potentially are the most vulnerable to model misspecification due to omitted
interactions with intervening levels of management and service delivery. They often
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

leave unexplored the mechanisms by which the public policy operates to produce the
observed effect. What can be made, for example, of Bradley’s (2003) conclusion from a
number of studies that non-profit hospitals provide better-quality services than their
private counterparts? Surely it is not simply the tax status that explains these
differences. What aspects of management or service delivery associated with tax status
account for the results? Bradley also found that health insurance did not seem to be
affected as much by ownership as by some interaction variable between provider and
hospitals.
It is not the case that all policy-to-outputs/outcomes findings suffer from omitted
variable bias. Studies that test the relationship between policy designs and societal
outcomes (such as the way in which Medicare and Medicaid policy changes affect health
care costs and usage) often omit management and service-delivery variables because the
particular policy under investigation does not allow for substantial managerial or
frontline discretion; it is virtually self-enforcing. Such policies are rather rare, however.

HEALTH CARE GOVERNANCE: WHAT IS LOGICAL?

We began this inquiry by asking: what can we learn about the governance and,
specifically, the management of health care delivery from empirical literatures that
address various aspects of this issue? With few exceptions, policies, management, and
service delivery are endogenous in complex health care governance processes. Specific
efforts to account for such endogeneity or explore it directly are quite rare, however.
We draw this conclusion based on two primary patterns we observed.
First, the types of variables at a particular level in the logic of governance that are
being explained (that is, used as dependent variables) tend not to overlap the types of
variables at those same levels that are used to explain (that is, used as independent
variables). This implies that researchers who are attempting to explain phenomena at
different levels of governance are employing different theories or logic models in their
explanations. It may be too much to ask of any particular researcher or study to
untangle complex endogeneity among governance processes. But given the relatively
disjointed pattern of results that we found, it is difficult even to craft a larger, reliable
picture from the available individual studies.
Second, findings related to what we characterize as ‘indirect governance’ may often
omit important mediating effects of management and/or of service delivery. A large
number of the findings we examined concerned indirect governance (i.e. c to e, d to f,
 474 Public Management Review

and c to f), where levels in the logic are omitted. Some of these findings were produced
by research that did control for intervening levels, but many did not. Although we note
that the specific patterns of health care are not necessarily generalizable to other fields
such as education or social services, the potentially important influence of omitted
mediating effects in studies of ‘indirect governance’ are likely to arise across almost all
fields of public policy.
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

Our analytic synthesis did not allow us to infer hierarchical interrelationships among
specific types of variables. Nonetheless, the general importance of mediating variables
when policies are not self-enforcing seems clear. For example, Surender et al. (2002) is
an exemplary qualitative study showing the mediating effects of managerial values and
strategies on service-delivery outcomes. The authors relied on interviews and survey
data to arrive at conclusions regarding the effect that managers’ opinions had on the
success of government-initiated Evidence-based Medicine Projects in Wales.
Additionally, LaMontagne and Kelsey’s (2001) study of the effect of OSHA’s 1984
ethylene oxide standard on the chance of unsafe worker exposure to ethylene oxide
demonstrates how multi-level models can be created to explain program outcomes in a
more holistic manner. Using survey data and quantitative methods, they test whether
the number of part-time workers (e-level), management’s decision to install combined
sterilizer-aerators (d-level), or hospital ownership structure (c-level) make a difference
in worker exposure incidents.
In general, health care outputs (services delivered and utilized) and outcomes
(indicators of health status) are the resultant, among other things, of complex
interactions among hierarchical levels of health care governance: public policies that
define and allocate resources to programs and organizations and establish terms for their
operation; the use of discretion by public managers to organize the work of their
agencies and implement particular values and strategies; and the values, attitudes,
treatment methods, and resource utilization skills of service workers. Data limitations
are likely to preclude comprehensive investigations of the full range of these
interactions. It is all the more important, therefore, that investigators be mindful of the
consequences of conceptually incomplete research designs when interpreting their
findings and drawing conclusions from them. Such mindfulness will sharpen the
contributions of individual studies to the overall understanding of health care
governance.

NOTES
1 For a fuller development of the ideas in this section, see Lynn et al. (2000a, 2000b, 2001).
2 A list of the articles included in the database is available from the authors upon request.
3 Although the overall null finding rate was around 20 percent, there were exceptions. With regard to
resources or performance, measured either at the level of primary work, or with outputs related to
individuals/society, or to the private sector, the rate of null findings was markedly higher: from 25 to 36
percent. It is possible that these types of dependent variables have higher public salience, and thus findings
 Forbes et al.: The logic of governance in health care delivery 475

related to them are of more interest regardless of the presence or direction of effect. It may also be the case
that conventional wisdom or existing policies presume (or have been shown to exhibit) relationships between
certain explanatory factors and their ‘outcomes’. A finding of ‘no effect’ is thus of policy and management
interest because it is counterintuitive. In contrast, variables at the management or primary work levels may be
either less salient or determined to a greater degree through discretion. For determinants of these types of
variables, findings of ‘no relationship’ may be of less interest to practitioners and researchers: the relationships
examined may be more specialized and contingent and conventional wisdom may not exist or may be less
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

salient.
4 A Chi-square test shows that the distributions of e-sublevel variables used in d to e findings and e to f findings
are statistically different (p 5 .001).
5 A Chi-square test of independence shows no statistical difference in the distributions of d-sublevel variables
used in c to d findings and in d to e findings (p ¼ .224).
6 A Chi-square test of independence indicates that the distributions of e-sublevels examined in d to e and in d to f
findings are marginally statistically different (p ¼ .101).

REFERENCES
Abelson, J., Tedford Gold, S., Woodward, C., O’Connor, D. and Hutchison, B. (2004) Managing Under
Managed Community Care: The Experiences of Clients, Providers and Managers in Ontario’s
Competitive Home Sector, Health Policy. 68: 3 pp359 – 72.
Boyne, G. A. (2003) Sources of Public Service Improvement: A Critical Review and Research Agenda, Journal of
Public Administration Research and Theory. 13: 3 pp367 – 94.
Boyne, G. A., Farrell, C., Law, J., Powell, M. and Walker, R. M. (2003) Evaluating Public Management Reforms:
Principles and Practice, Buckingham, UK: Open University Press.
Bradley, M. (2003) What Do We Know about Ownership? Working paper, Texas A&M University.
Cole, M. (2003) The Health Action Zone Initiative: Lessons from Plymouth, Local Government Studies. 29: 3
pp99 – 118.
D’Aunno, T., Succi, M. and Alexander, J. A. (2000) The Role of Institutional and Market Forces in Divergent
Organizational Change, Administrative Science Quarterly. 45: 3 pp679 – 703.
Doolin, B. (2001) Doctors as Managers: New Public Management in a New Zealand Hospital, Public Management
Review. 3: 2 pp231 – 54.
Ettner, S. L. and Hermann, R. C. (2001) The Role of Profit Status under Imperfect Information: Evidence from
the Treatment Patterns of Elderly Medicare Beneficiaries Hospitalized for Psychiatric Diagnosis, Journal of
Health Economics. 20: 1 pp23 – 49.
Evans, J. H., III, Hwang, Y. and Nagarajan, N. J. (2001) Management Control and Hospital Cost Reduction:
Additional Evidence, Journal of Accounting and Public Policy. 20: 1 pp73 – 88.
Forbes, M. and Lynn, L. E., Jr (2005) How Does Public Management Affect Government Performance? Findings
from International Research, Journal of Public Administration Research and Theory. 15: 4 pp559 – 84.
Hill, C. J. and Lynn, L. E., Jr, (2005) Is Hierarchical Governance in Decline? Evidence from Empirical Research,
Journal of Public Administration Research and Theory. 15: 2 pp173 – 95.
Hill, C. J., Lynn, L. E., Jr, Proeller, I. and Schedler, K. (2005) Introduction to a Symposium on Public
Governance, Policy Studies Journal. 33: 2 pp203 – 11.
Holahan, J., Dor, A. and Zuckerman, S. (1991) Has PPS Increased Medicare Expenditures on Physicians?, Journal
of Health Politics, Policy and Law. 16: 2 pp335 – 62.
Joyce, T. (1999) Impact of Augmented Prenatal Care on Birth Outcomes of Medicaid Recipients in New York
City, Journal of Health Economics. 18: 1 pp31 – 67.
 476 Public Management Review

Kendrick, J. S., Zahniser, C., Miller, N. and Salas, N. (1995) Integrating Smoking Cessation into Routine Public
Prenatal Care: The Smoking Cessation in Pregnancy Project, American Journal of Public Health. 85: 2
pp217 – 22.
Kiser, L. L. and Ostrom, E. eds (1982) ‘The Three Worlds of Action: A Metatheoretical Synthesis of
Institutional Approaches’ in E. Ostrom Strategies of Political Inquiry, Beverly Hills, CA: Sage Publications.
Kooiman, J. (1999) Social-Political Governance, Public Management: An International Journal of Research and Theory.
1: 1 pp67 – 92.
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

LaMontagne, A. D. and Kelsey, K. T. (2001) Evaluating OSHA’s Ethylene Oxide Standard: Exposure
Determinants in Massachusetts Hospitals, American Journal of Public Health. 91: 3 pp412 – 17.
Lipsky, M. (1980) Street-Level Bureaucracy, New York: Russell Sage Foundation.
Lupia, A. and McCubbins, M. (2000) Representation or Abdication? How Citizens Use Institutions to Help
Delegation Succeed, European Journal of Political Research. 37: 3 pp91 – 307.
Lynn, L. E., Jr, Heinrich, C. J. and Hill, C. J. (2000a) ‘Studying Governance and Public Management: Why?
How?’ in C. Heinrich and L. Lynn (eds) Governance and Performance: New Perspectives, Washington, DC:
Georgetown University Press.
—— (2000b) Studying Governance and Public Management: Challenges and Prospects, Journal of Public
Administration Research and Theory. 10: 2 pp233 – 61.
—— (2001) Improving Governance: A New Logic for Empirical Research, Washington, DC: Georgetown University
Press.
Meier, K. (1993) Latinos and Representative Bureaucracy Testing the Thompson and Henderson Hypothesis,
Journal of Public Administration Research and Theory. 3: 3 pp393 – 414.
Moore, W. J., Gutermuth, K. and Pracht, E. E. (2000) Systemwide Effects of Medicaid Retrospective Drug
Utilization Review Programs, Journal of Health Politics, Policy and Law. 25: 4 pp653 – 88.
Nielsen, L. B. and Wolf, P. J. (2001) Representative Bureaucracy and Harder Questions: A Response to Meier,
Wrinkle, and Polinard, Journal of Politics. 63: 2 pp598 – 615.
Osborne, S. P. (2006) The New Public Governance? Public Management Review. 8: 3 pp377 – 87.
Ruef, M. and Scott, W. R. (1998) A Multidimensional Model of Organizational Legitimacy: Hospital Survival in
Changing Institutional Environments, Administrative Science Quarterly. 43: 4 pp877 – 904.
Shi, L. (1996) Patient and Hospital Characteristics Associated with Average Length of Stay, Health Care
Management Review. 21:2 pp46 – 61.
Strøm, K. (2000) Delegation and Accountability in Parliamentary Democracies, European Journal of Political
Research. 37: 3 pp261 – 89.
Succi, M. J. and Alexander, J. A. (1999) Physician Involvement in Management and Governance: The
Moderating Effects of Staff Structure and Composition, Health Care Management Review. 24: 1 pp33 – 44.
Surender, R., Locock, L., Chambers, D., Dopson, S. and Gabbay, J. (2002) Closing the Gap between Research
and Practice in Health: Lessons from a Clinical Effectiveness Initiative, Public Management Review. 4: 1
pp45 – 61.
Toonen, T. A. J. (1998) Networks, Management, and Institutions: Public Administration as ‘Normal Science’,
Public Administration. 76: 2 pp229 – 52.
Udom, N. and Betley, C. L. (1998) Effects of Maternity-Stay Legislation on ‘Drive-Through Deliveries’, Health
Affairs. 17: 5 pp208 – 15.
Vickers, G. (1983) The Art of Judgment: A Study of Policy Making, London: Harper & Row.
Wamsley, G. L. (1990) ‘Introduction’ in G. L. Wamsley and J. F. Wolf (eds) Refounding Public Administration,
Newbury Park, CA: Sage Publications.
Weiher, G. R. (2000) Minority Student Achievement: Passive Representation and Social Context in Schools,
Journal of Politics. 62: 3 pp886 – 95.
 Forbes et al.: The logic of governance in health care delivery 477

APPENDIX

List of journals

Journal Number of articles

Academic Medicine 1
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 22:43 23 December 2014

Administrative Science Quarterly 3

American Journal of Public Health 17
British Journal of Management 1
Children & Youth Services 1
European Journal of Political Research 1
Health Affairs 3
Health Care Management Review 5
Health Policy 4
Health Policy and Planning 6
Hospital & Community Psychiatry 1
International Journal of Health Care Finance and Economics 2
International Journal of Public Administration 1
Journal of Accounting and Public Policy 3
Journal of Human Resources 4
Journal of Politics 1
Journal of Health & Social Behavior 1
Journal of Health Economics 21
Journal of Health Politics, Policy and Law 7
Journal of Management Studies 1
Journal of Public Administration Research and Theory 2
Journal of Public Health Medicine 3
Journal of Public Policy 1
Journal of Public Policy & Marketing 1
Journal of the American Medical Association (JAMA) 1
Journal of Policy Analysis and Management (JPAM) 2
Local Government Studies 1
Medical Care 1
Medical Care Research & Review 2
Mental Health Services Research 1
New England Journal of Medicine 1
Public Administration 1
Public Administration Quarterly 1
Public Choice 1
Public Management Review 3
Public Money and Management 3
Public Policy and Administration 1
Publius 1
Social Forces 1

Total 112