Seeing the Forest for the Trees: An Atlas of the Politics–Administration Dichotomy
| Published date | 01 March 2014 |
| DOI | http://doi.org/10.1111/puar.12163 |
| Date | 01 March 2014 |
| Author | Ion Georgiou |
Ion Georgiou’s interdisciplinary research
spans the problem structuring methods of
operations research, network development
and analysis, and the history of decision-
making methodologies. He has published on
topics as diverse as railroad transportation
policy, case-based experiential learning,
decision-making methodology, and the
history of management. In public admin-
istration, he recently offered a detailed
analysis of the arguments used by Herbert
Simon against Luther Gulick in the 1940s
(published in Public Administration).
E-mail: iongeorgiou@gmail.com
156 Public Administration Review • March | April 2014
Public Administration Review,
Vol. 74, Iss. 2, pp. 156–175. © 2014 by
The American Society for Public Administration.
DOI: 10.1111/puar.12163.
Ion Georgiou
Fundação Getulio Vargas, Brazil
Recent years have seen attempts to make sense of the
politics–administration dichotomy. Triangulating among
historical research, empirical observations, new models of
interaction between politicians and administrators, and
the division of the literature into “schools,” novel ways of
understanding and examining the dichotomy have devel-
oped. ese have been largely thematic and have revealed
the extent of a literature spanning more than 120 years.
Because of its size, a complementary structural analysis
of the literature now not only is conceivably useful but
also can off er means for approaching it. is article off ers
an atlas—that is, a series of visual maps, accompanied
by associated statistics and interpretations—that can
assist researchers in their travels through the territory of
the dichotomy. Ten ways of tackling the literature are
presented, culminating in an initial reading list that
covers the breadth of dichotomy research, thus providing
an epistemological foundation for those who wish to enter
the territory.
Recent years have seen attempts to make sense
of the politics–administration dichotomy.
Triangulating between historical research
(Overeem 2012), empirical observations (Demir
2009a), new models of interaction between politi-
cians and administrators (Svara 1985, 2001, 2006a,
2006b), and the division of the literature into
“schools” (Demir 2009b), novel ways of viewing, cri-
tiquing, understanding, and examining the dichotomy
have developed. ese have been largely thematic and
have revealed the voluminous
extent of a literature spanning
more than 120 years. Because
of its size, a complementary
structural analysis of the litera-
ture now not only is conceivably
useful but also can off er means
for approaching it. To this end,
what is off ered here is an atlas—
that is, a series of visual maps,
accompanied by associated sta-
tistics and interpretations—that
can assist researchers in their
travels through the territory of the politics–adminis-
tration dichotomy.
e atlas is designed as a networked bibliography,
extracted specifi cally from the American literature,
of journal articles pertinent to the dichotomy for the
period 1887–2010. e network is subjected to fi ve
analyses chosen for their broad relevance to biblio-
graphic networks. Some highlights of the results from
these analyses include, for example, (1) the extent
of the complementarity turn in dichotomy studies,
triggered in the mid-1980s; (2) the identifi cation of a
subgroup of classical papers that continue to com-
mand attention in contemporary dichotomy research;
and (3) the qualifi ed overall dominance of Public
Administration Review set against a growing diversi-
fi cation of journals publishing dichotomy studies. In
addition, the application of network theory enables
the identifi cation of an initial reading list that covers
the breadth of the literature’s 123 years, thus provid-
ing an adequate epistemological foundation for those
who wish to enter the territory. Ultimately, 10 means
for approaching, or entering, the dichotomy literature
are identifi ed.
In order to appreciate the fi nal product for what it is,
the fi rst task is to qualify the term “networked bibli-
ography” and the criterion employed for the inclusion
of journal articles within it. For convenience, in what
follows, the bibliographic network to be presented will
be referred to as the “dichotomy
network.”
Terms of Reference
As with models in general
(Morgan and Morrison 1999),
networks are representational
tools for investigating some
aspect of the world. eir
accuracy refl ects, and results
from, design criteria, not least
concerning issues of data inclu-
sion and exclusion, as well as of
Seeing the Forest for the Trees: An Atlas of the Politics–
Administration Dichotomy
What is off ered here is an
atlas—that is, a series of
visual maps, accompanied
by associated statistics and
interpretations—that can
assist researchers in their trav-
els through the territory of
the politics–administration
dichotomy.
Seeing the Forest for the Trees: An Atlas of the Politics– Administration Dichotomy 157
approach does not focus on the evolution of a literature and, con-
versely to coupling, proposed co-citation, whereby citees are paired
based on the frequency with which they appear jointly in citers.
With co-citation, as the literature, and especially the distribution
of citations, evolves, so do clusters of citees, allowing for longitudi-
nal explorations of changes in a fi eld of scholarship. Variations in
coupling and co-citation computations are discussed by Batagelj and
Cerinšek (2013). Here, coupling and co-citation will be pursued
based on the classic approaches by Kessler and Small, and interpre-
tations of the results will be off ered.
Main path analysis uncovers a chronologically ordered stream of
papers, spanning the entire network from beginning to end, which
appear on a greater number of citation paths than others. eir
prominence is hypothesized as rendering explicit the essential
stream of research within a given literature. ree variations were
fi rst proposed by Hummon and Doreian (1989) in the late 1980s,
and a number have been developed since (Hummon, Doreian, and
Freeman 1990; Liu and Lu 2012; Lucio-Arias and Leyesdorff 2008).
Batagelj’s (2003) variation—known as search path count (SPC)—
has been generally adopted (Jo et al. 2009; Liu et al. 2013; Lu et al.
2012) and serves as the foundation for more advanced variations (Liu
and Lu, 2012). Batagelj’s SPC is also included in the network analy-
sis software Pajek (De Nooy, Mrvar, and Batagelj 2011, 282–88),
which was used for developing and exploring the dichotomy net-
work. SPC has become a central feature of citation network explo-
rations, and accordingly, it will be applied to the dichotomy network.
Design Criterion of the Dichotomy Network
Citation networks are usually constructed based on wholesale
extractions from digitally searchable bibliographic databases. ese
databases are essentially archival deposits, and their nature, as well
as the practice of extracting citations en masse (e.g., Lucio-Arias and
Leydesdorff 2007, 2008), raises some specifi c modeling problems
summarized by Marsden (2005, 24–25).
For example, citation counts may be inaccurate because of changes
in journal names or nonstandard journal abbreviations. Editorial
policies aff ect which indexing services track which journals, render-
ing no database unqualifi edly comprehensive and the availability
of records therein eclectic. Citations themselves do not simply
contribute to the fl ow of knowledge: they have been found to gener-
ate information cascades resulting in unfounded authority claims
(Greenberg 2009). Furthermore, citations may be used for other
purposes, such as window dressing, self-citation, paying homage,
politically motivated fl attery, off ering corrections, opening disputes,
describing methodologies, or simply off ering a literature review
(Hummon and Doreian, 1989, 40). Citation practices, and the
possible meanings of citations, have also been found to diff er across
research areas (Hargens 2000).
Given such variety in citation characteristics, Marsden notes that
“[r]elatively few explicitly methodological studies of archival data
appear in the network literature” (2005, 24), resulting in a network
developer having to use those analytical tools that appear to be most
relevant to the context at hand and to the objectives concerning
the fi nal product. In this respect, Marsden concludes, “Assessments
of data quality, regardless of source, will be facilitated if research-
ers clearly articulate their concepts of the ‘true scores’ they seek to
technique particular to network modeling itself. ose most relevant
to the dichotomy network are outlined here.
Networks are underpinned by the abstract mathematical fi eld
known as “graph theory” (Gross and Yellen 2006; Newman 2010)
and thus are amenable to quantitative explorations. eir basic rep-
resentational objects are vertices connected by lines. Symmetrical/
reciprocal relationships between vertices are represented by undi-
rected lines, or “edges,” while asymmetrical relationships are repre-
sented by directed lines, or “arcs.” Both types of lines will be used
here, with edges appearing as a result of certain particular analyses.
e fundamental advantage of networks lies in relational explora-
tions. In this respect, the dichotomy network is a bibliographical
network in which the arcs represent citation links between journal
papers, which, in turn, are represented by vertices. In this “citation
network,” a paper at the tail of an arc—the “citee”—is cited by
a paper at the head of the arc—the “citer.” e terms “citee” and
“citer” will refer to papers as opposed to any individual author. In
graph-theoretical terminology, the sum of arrows leaving a citee
constitute its “outdegree.” Conversely, the sum of arrows reaching a
citer constitute its “indegree.” is useful shorthand will be adopted.
Interest in citation networks stems from Garfi eld’s (1955, 108, 111)
initial considerations on general bibliometrics, which can be traced
back to Lotka (1926). Garfi eld proposed “a bibliographic system
for science literature that can eliminate the uncritical citation of
fraudulent, incomplete, or obsolete data by making it possible for
the conscientious scholar to be aware of criticisms of earlier papers.”
It is worth noting that Garfi eld did not propose a system that, of
itself, would perform the desired elimination and thus substitute for
scholars’ own judgments. Garfi eld intended a map of the “associa-
tion of ideas” that could assist “conscientious” scholars in their docu-
mentary search. He also pointed out that the “system” could be used
to calculate “impact factors,” but he concluded, almost as a warning,
that the system itself as well as its results are “just a starting point in
literature research.” e dichotomy network is off ered as just such a
starting point.
is initial warning on the limitations of citation networks has
spawned extensive refl ections in its literature. ey are summarized
by Garfi eld (1979b) and MacRoberts and MacRoberts (1989). e
most basic point is that the network itself does not say anything
about the content of the citations. Jo et al. note that “the quality of
a citation is unknown without further contextual examination of
the citation context,” adding that “a citation network never provides
qualitative information on the relationship among articles except
for the existence of a citation” (2009, 519). In other words, the
network can only off er results that speak of its structure, that is, the
positioning of citations in relation to each other. Researchers can
then use such results as pointers for inquiring into, and interpreting
the content of, the items cited. Two particular structural approaches
developed for bibliographic networks will be pursued here: (1) the
conjoining of papers through “coupling” and “co-citations” and (2)
the detection of the “main path” through the network.
Couplings were fi rst proposed by Kessler (1963) in the early 1960s
as bibliographic pairings of citers based on the citees they share. Ten
years later, Small (1973) pointed out that this backward-looking
To continue reading
Request your trial