ICAE Working Paper Series - No.

46 - March 2016 (Updated: August 2017)

The Power of Scientometrics

and the Development of Economics

Matthias Aistleitner
Jakob Kapeller
Stefan Steinerberger

Institute for Comprehensive Analysis of the Economy

Johannes Kepler University Linz
Altenbergerstraße 69, 4040 Linz
[email protected]
www.jku.at/icae
 The Power of Scientometrics and the Development of Economics«

Matthias Aistleitner*, Jakob Kapeller** and Stefan Steinerberger***

Abstract

Citation metrics and its related indices and rankings become increasingly important in

the evaluation of research. Such indices are part of a more general tendency aiming for

the simplification of complex and interconnected phenomena through quantification.

The purpose of our contribution is to analyze the impact of such quantitative indices on

the further development of science with a special emphasis on economics. In this case

we observe a multitude of interesting effects on both, the level of individual scientists

as well as the global development of the discipline.

Keywords

pluralism, academic reproduction, paradigms, citation metrics, reactivity,

quantification

JEL classification

A14, B50

«
This work was funded by the Institute for New Economic Thinking (INET) under grant numbers
INO1500038 and INO1500039. The manuscript has been accepted for publication in the Journal of Economic
Issues.
* Johannes Kepler University Linz, Institute for Comprehensive Analysis oft he Economy (ICAE),

[email protected]
** Johannes Kepler University Linz, Institute for Comprehensive Analysis oft he Economy (ICAE) &

Department of Economics, [email protected]

*** Yale University, Department of Mathematics, [email protected]

1
 1. Introduction

Understanding the impact of scientometrics on the development of academic

disciplines is a complex problem of great current significance. Its relevance stems from

a recent trend to introduce numerical measurements of scientific performance in order

to evaluate research activities and facilitate comparisons on various levels, e.g. across

different researchers, institutions or publication outlets. These comparisons usually take

the form of rankings, which aim to ‘measure’ the ‘quality’ of universities, scientists,

scientific articles and journals on a single scale.1 This development also points to a

methodological shift inside scientometrics, which has its conceptual origins in an

interpretative analysis of scientific communication aiming to understand the discursive

properties of academic publishing (e.g. De Solla Price 1965, Rip and Courtial 1984).

This approach, which has sometimes been dubbed as ‘cognitive scientometrics’, is

increasingly making way for new forms of ‘evaluative scientometrics’, which try to

define indicators of research quality based on an analysis of citation frequencies inside

the scientific literature or through peer-review- and survey-instruments (Mingers and

Leydesdorff 2015, Adler et al. 2008).

The increasingly prominent role of quantitative evaluation in academia can be

interpreted as a part of a more general social trend oriented towards the numerical

assessment of social issues. One major historical impetus of this process of an increased

‘quantification of social phenomena’ (Espeland and Stevens 2008, p. 401) is to make

social conditions politically predictable and controllable. An archetypical example in

this context is the development of the system of national accounts (SNA) and its

corresponding parameters like the gross domestic product (GDP), which serves the

purpose of assessing a nation’s economic activities and thereby provides a yardstick for

2
measuring economic development. Aside from administrative interests, indicators like

the GDP were also developed to serve scientific purposes, since they facilitate the

quantification and measurement of theoretically postulated aggregate concepts like

GDP, inflation or capital.

From a theoretical perspective one can understand the ‘quantification of social

phenomena’ as a ‘general sociological phenomenon’, which covers ‘the production and

communication of numbers’ and deals with ‘regimes of measurement’ (cf. Espeland

and Stevens 2008, p. 401). From this point of view numbers fulfill two functions: for

one, they serve as symbolic placeholders representing single entities or events (e.g.

‘9/11’). Such placeholders do indeed contribute to the ‘quantification of social

phenomena’ as a general tendency but they by themselves do not yet establish any

specific ‘regimes of measurement’. However, numerical information also directs

attention towards the relative properties of various entities: when assigning numerical

values to different entities of the same class, numerical information allows to create an

ordering that has the double function of unifying different objects across a uniform

scale, which, at the same time, makes it easier to distinguish and differentiate between

these objects (cf. Espeland and Stevens 2008).

Recently, this process of quantification has gained additional momentum within science

due to the introduction of regimes of measurement in the sphere of research evaluation.

The interplay between supply and demand for evaluation of academic performance

have the potential to create ever-new tools for and facets of numerical evaluation

procedures in academia.

3
While the rise of evaluative scientometrics within academia is clearly part of a broader

phenomenon, this paper focuses more particularly on its role and impact within the field

of economics. Its main contribution is to provide a theoretical embedding of the notion

of quantification in a more general account of scientific development and to better

understand its impact on the development of specific research fields and the behavior

of individual researchers. The resulting argument not only provides an integrated

assessment of the ‘power of evaluative scientometrics’ but also supports this

assessment with small case studies focused on economics, which are suitable for

empirically illustrating the underlying argument.

We focus on three specific aspects: firstly, scientometric routines in research

assessment are always embedded in conventional patterns of academic reproduction;

hence, the intensified use of these routines has to be understood against this backdrop,

which is the main focus of section 2. Secondly, scientometric evaluation is based on

mechanical procedures that are easily reproducible and therefore cause incentives for

strategic behavior among researchers. In section 3 we discuss some aspects of reactivity

with a special focus on the impact of reactive evaluation routines on the citation

behavior of economists. Thirdly, we consider the special case of economics as a

‘contested discipline’ (Lee and Elsner 2011). Unlike other social sciences, economics

has only one dominant paradigm: the axiomatic core of so-called mainstream

economics is largely based on neoclassical economic principles. These principles guide

the majority of the economic scientific community. Their conceptual predominance

remains largely unchallenged except for the protest of a small minority of heterodox

economists (Dobusch and Kapeller 2012a) and, more recently, also students of

economics (ISIPE 2014). In this context, section 4 selectively summarizes and extends

4
past approaches analyzing interparadigmatic engagement in economics and discusses

the impact of quantitative evaluation regimes against the backdrop of paradigmatic

divisions in economics. Section 5 concludes and summarizes the main arguments of

this paper.

2. Patterns of Academic Reproduction: from Matthew Effects to Path-

Dependency

‘The competition between paradigms is not the sort of battle

that can be resolved by proofs.’ (Kuhn 1962, p. 148)

One classical finding of scientometric research is that academic attention – mostly

measured on the basis of citation frequencies – is highly skewed. The distribution of

attention, influence and prestige among a given quantity of single researchers or

research articles follows a simple underlying structure: most researchers or research

articles receive no or just very little attention, whereas a few researchers or research

articles receive a great deal of attention. From a formal point of view, this specific

distribution of attention, influence and prestige in science has similar characteristics to,

for instance, the distribution of wealth or the attractiveness of websites and follows a

‘power law’ Pareto/Zipf distribution at the top. Such types of distributions are common

within social contexts. To name a few examples, the population of cities, the number

of received phone calls, the number of words used in a text or the sales figures of book

titles tend to follow such a power law (Newman 2006). One main feature of this kind

of distribution is that just a few elements at the top of the respective distribution collect

a disproportionately large share of the variable of interest. One example, derived from

the population structure of Germany, is that the inhabitants of only four different cities

5
comprise about 10% of the German population. In the scientific discourse, a majority

of inner-academic attention focuses only on a small fraction of the respective scientific

literature.2 Figure 1 illustrates the characteristics of such distributions and shows three

potency-distributed measures in descending order – the population of the 82 biggest

cities in Germany, the wealth distribution in Austria, and the distribution of citations to

articles published in the American Economic Review between 1981-19853. To better

illustrate the properties of said samples, we compare these three distribution to that of

the birth weight of newborn babies, which is a normally distributed random variable.

Figure 1 reminds us of an important property of power-law distributions, namely that

the differences between median-, mean- and maximum-values are significantly larger

than they would be for a quantity that follows a normal distribution. Indeed, for some

Pareto distributions with heavy tail the notion of mean is not even defined. The simple

interpretation of this pattern is that in the case of power-law distribution more extreme

values occur and, hence, these extreme values are quantitatively more important than

in the case of normally distributed properties. For distributions following a power law,

we observe a remarkable difference between median and mean as a large amount of the

total quantity is concentrated in the edge of the distribution.

<Figure 1 near here>

In 1965, Derek J. de Solla Price undertook the first systematic study on the distribution

of attention in scientific discourse. Based on an analysis of citation data he postulated

that current generations of scientific results only refer to a small number of past

contributions. However, the mechanism from which this uneven distribution emerges

was not apparent at first and only later clarified by Robert K. Merton’s classical work

on the ‘Matthew Effect in Science’ (Merton 1968): he postulated that the acquisition of

6
prestige and attention given to scientific work is closely correlated with the amount of

attention acquired in the past. Merton based his argument on citation data as well as

interviews conducted with Nobel Prize laureates. The implicit logic of this mechanism,

namely ‘whosoever has, will be given more’ is nowadays aptly called the ‘Matthew

Effect’ in reference to the passage in the gospel of Matthew4.

‘[T]he Matthew effect consists in the accruing of greater increments of

recognition for particular scientific contributions to scientists of considerable

repute and the withholding of such recognition from scientists who have not yet

made their mark.’ (Merton 1968, p. 58)

Matthew effects can be characterized more generally as self-reinforcing effects, which,

in the special case of Matthew effects in science, have the following structure: ‘If an

author/article x is cited, then x will become a more attractive point of reference in the

future.’

Self-reinforcing effects appear in similar form in a series of social contexts, which for

instance has been discussed in detail in economic literature about path dependency

(Sydow et al. 2009; Dobusch and Schüßler 2013). The main thesis of this theoretical

strand is that the establishment of technical, organizational or social standards can lead

to self-reinforcing effects, which lead to a relative or absolute dominance of these

standards. The former kind of dominance (relative dominance) can be observed in most

cases in the form of a distribution of power-law type as discussed above.

Classical applications of path dependency theory can be found in examples such as the

persistence of specific technological standards (e.g. the QWERTY keyboard design;

7
see David 1985), the evolution of monopolies in software markets (Shapiro and Varian

1999), the relative attractiveness of file sharing networks and social networking sites or

the cluster structure of the high-tech industry (Arthur 1994). Moreover, the

development of the Pareto-distribution as a ‘natural’ form of wealth distribution can be

simulated in a theoretical model by integrating self-reinforcing effects (Levy and Levy

2003). Self-reinforcing mechanisms can be empirically identified and range from direct

network effects (the more a standard is used, the more attractive it becomes) over

learning effects (the better a standard has been understood, the more attractive it

becomes) to indirect network effects (if a certain standard is a requirement for the use

of other products, this standard will become more attractive).

Based on the examples presented above, the Mertonian Matthew effect can be

interpreted as a central mechanism of academic reproduction and posits – in terms of

path dependency theory – a so-called a ‘direct network effect’: the attractiveness of a

point of reference correlates with the number of past references. The consequence of

such self-reinforcing effects is unevenly distributed attention inside the scientific

discourse, where a few contributions receive disproportionately large attention, while

many others remain largely unnoticed. It is indeed observed that most scientometric

indicators follow a power-law distribution.

These patterns of academic reproduction do not only hold for the level of individual

researchers and articles, but also apply to scientific institutions (e.g. academic journals)

as well as specific schools of thought and paradigmatic traditions (in the sense of Kuhn

1962). Such a perspective, which applies the basic idea of a Matthew-effects to specific

fields and approaches instead of single authors and contributions, is exemplified by

8
simulation-based study of John D. Sterman and Jason Wittenberg (1999), which

provides a detailed account on the path-dependent properties of scientific evolution.

For example, Sterman and Wittenberg found that the intrinsic quality of the core ideas

of a single paradigm is only of minor importance for its success, which provides a

theoretical rationale for non-linear developments in the field of scientific knowledge

and stands in line with Thomas S. Kuhn’s historical observations. A prime example in

this context is given by the advent of Copernican theory, which in its beginning – in

terms of its empirical explanatory power and precision – lagged behind the geocentric

view. The main reason for the initial superiority of the Ptolemaic theory was that it

could employ a sophisticated theoretical apparatus with numerous correction terms (so-

called ‘epi-cycles’) introduced to improve the empirical performance of the underlying

models.

For the purpose of this paper, it seems promising to take a closer look at the structure

of the Sterman and Wittenberg model, in which three essential positive feedback loops

emerge. These feedback loops stabilize the persistence of paradigms and thereby

provide a theoretical justification of the Kuhnian observation by means of path-

dependence theory. The feedback loops take the form of direct network effects and refer

to the academic labor market, the perceived relative explanatory power of paradigms

and the role of obvious anomalies. In all three cases, a greater number of practitioners

within a paradigm leads to self-reinforcing effects because the number of recruitments,

solved problems and rationalized anomalies is proportional to a paradigm's size. In turn,

these size-dependent factors contribute in a positive way to the attractiveness of the

paradigm and therefore stabilize the dominance of established patterns of thought in the

sense of a path dependency of the paradigmatic development.

9
 ‘The prevalence of positive feedback processes in paradigm development

means that the evolution of the system as a whole is strongly path-dependent.’

(Sterman and Wittenberg 1999, p. 333)

Building on this argument, which is also summarized in Figure 2 we now sketch a

theoretical argument contextualizing the advent of evaluative scientometrics within the

past one-and-a-half decades5 on academic reproduction. Evaluative scientometrics tries

to measure influence inside the sciences by determining standardized citation

frequencies (usually the number of citations received by an author, an article or a

journal), which in turn is interpreted as a measure of quality. The self-referential logic

of this process is striking, since quality is essentially equated with influence and impact.

Measuring impact then serves as a basis for institutional evaluation, which further

redistributes influence inside the scientific community. In this context, the

implementation of ranking systems in order to evaluate research performance leads to

a further concentration of academic attention by attesting a high level of quality to those

authors, articles and research fields whose initial level of paid attention is already high

(Dobusch and Kapeller 2009). Accordingly, a fourth feedback loop can be added to the

model of Sterman and Wittenberg, which can be formulated in the following way: The

bigger a paradigm or research field is, the higher the amount of received citations in

this field will be – ‘big is beautiful’. Finally, this number of received citations is used

as a rarely questioned hallmark of scientific quality and therefore further improves the

attractiveness of the respective paradigm (see Figure 2).

<Figure 2 near here >

10
This additional feedback loop appears in the form of an indirect network effect: The

establishment of scientometric indicators as a standard in terms of scientific quality

evaluation favors ‘bigger’ fields of research and makes them more attractive for those

researchers, who have internalized the ruling quality standards of scientific evaluation.

While these rather general arguments on additional feedback loops in academic

reproduction are rather difficult to empirically illustrate, it is possible to make a clear-

cut argument on the expected effects of this additional feedback-loop of rankings – and

the associated visibility – on academic journals. The inclusion of a journal in a ranking

should boost its visibility and lead to an increase in attractiveness for potential authors,

i.e. result in an increase in received submissions. Regrettably, submission data is hardly

shared and often treated confidentially by editors and/or publishers. Nonetheless, we

managed to acquire submission data for three anonymous economic journals of

comparable size and character, which joined the most important ranking of journals –

Thomson Scientific’s Journal Citation Reports (JCR), in the years 2010/11.6 Their

aggregate average submission numbers in the years before and after inclusion in the

ranking look as follows.

Average number of Average number of Average growth in

annual submissions annual submissions after annual submissions
before inclusion into the inclusion into the JCR after inclusion into the
JCR JCR

87 Submissions/Year 156 Submissions/Year 78.8%

Table 1: Aggregated submission data from three economic journals joining the JCR in 2010/11.

While Table 1 gives a first intuition on the impact of the indirect network effect

introduced by evaluative scientometrics, it focuses on special cases – newcomers, so to

11
say, for whom the inclusion in the JCR obvious implies a boost in attractiveness. For

the established journals, institutions and paradigms the very same effect obviously leads

to a reinforcement of inherent path-dependencies by rewarding the already rewarded

once more and thereby shedding light on those, who are already visible (see also

Demange 2012). Hence, it is possible to state a first finding regarding the power of

(evaluative) scientometrics, as it reinforces existing patterns of academic reproduction

by increasing the number of self-reinforcing feedback-loops operating in the

distribution of academic attention and interests.

3. Evaluation Routines and the Role of Reactive Measurement Procedures

There are a large number of quantitative evaluation methods. Most of them are based

on conventional empirical techniques such as surveys (e.g. the peer-review process for

journal evaluation of the German Academic Association for Business Research7) or

counting event-frequencies (e.g. citation frequencies). In the context of empirical social

research such techniques are required to conform to certain minimal methodological

requirements. Among these minimal requirements are the validity (which means really

measuring what is meant to be measured) and reliability (repeated measurements

should lead to similar results) of measurement techniques. Both aspects require that

empirical measurement procedures are non-reactive, which means that their application

must not influence the observed behavior.

This idealized methodological viewpoint is widely contradicted by the fact that the

reactivity phenomenon occupies a central role in inner-scientific evaluation routines.

Reactivity means that the application of an investigation instrument can lead to a

12
change in behavior of the observed subjects. In the social sciences reactivity is primarily

seen as a methodological problem, which may cause biases if data is collected

repeatedly. In the case of methodologically guided evaluations this bias translates into

a problem of individuals anticipating the specific evaluation criteria, which leads to a

change in behavior in accordance the criteria imposed for purposes of evaluation.

A significant contribution for the understanding of the consequences of reactive

measurement procedures in scientific evaluation comes from Wendy N. Espeland and

Michael Sauder (2007). In a comprehensive study about the effects of a law school

ranking introduced by the magazine U.S. News, they find that evaluative measures,

especially rankings, can have a strong influence on the social environment as well as

on single actors in the scientific community.

Specifically, the authors show that behavior and perception of relevant agents in the

US-law school context (deans and faculty, students, public institutions, donors, etc.) is

influenced by two central mechanisms of reactivity: self-fulfilling prophecies and

commensuration. The first mechanism, self-fulfilling prophecies, means that the law

school-ranking leads to a behavioral change which further reinforces and polarizes the

ranking position: the ranking assumes a self-affirming character. The ordering of the

law schools according to a ranking score signalizes decisive differences which induces

a behavioral change of the social environment (students, public, donors, etc.) and

furthermore leads to a reinforcement of the respective ranking position or trend

(downward or upward) in the ranking process. For example, former ranking scores are

not only used to determine the allocation of financial resources, but also impacts peer-

review processes itself when former ranking positions are interpreted as indicators of

13
quality by reviewers. Towards the lower end of a ranking list, the power of such a self-

fulfilling prophecy can lead to a downward spiral: the lower the ranking position, the

lower the equipment with financial resources through external financiers (which play a

central role in the US-higher education system), the harder it becomes to move up or

even maintain one's position in the ranking.

For law schools the internalization of such factors leads to a stronger focus on aspects

that influence the position in the ranking list. At the same time, aspects that are

irrelevant for the ranking are neglected (Espeland and Sauder 2007, p. 11-14).

‘Rankings create self-fulfilling prophecies by encouraging schools to become

more like what rankings measure, which reinforces the validity of the

measure.’ (Espeland and Sauder 2007, p. 15)

The second mechanism, commensuration, is characterized by a transformation of

human cognition, caused by a numerical-competitive framing of a complex social

object. Here, the conceptual clarity of rankings plays a central role: it suggests that the

comparative analysis of the relative performance of educational institutions is achieved

by means of a down-to-earth and trustworthy yardstick.

‘[Commensuration] changes the locus and form of attention, both creating and

obscuring relations among entities. Commensuration is characterized by the

transformation of qualities into quantities that share a metric, a process that is

fundamental to measurement.’ (Espeland and Sauder 2007, p. 16)

14
In the context of the law school ranking, the commensuration mechanism is triggered

by the hierarchical relationship that emerges from an explicit comparison of scores. In

the context of economics journal rankings serve as main reference points for research

evaluation (Lee 2007; Bloch 2010; Corsi et al. 2010) and, hence, as the main medium

for commensuration making the journals the prime level for comparisons of all sorts.

‘By simultaneously unifying and distinguishing objects [...] rankings classify,

reward and punish, and organize interventions.’ (Espeland and Stevens 2008,

p. 416)

In Espeland and Stevens’ case the imposition of a ranking leads to a series of specific

patterns of action on the side of the law schools, like increasing expenses for marketing

to raise the chance of successful future peer review processes8, creating specific

administrative departments to obtain information about the employment status of

graduates or lowering ‘acceptance rates’ to signal selectivity to external assessors. As

students' test results are relevant for the ranking, many law schools increased the

number of merit-based scholarships in order to attract better students and decreased the

importance of other evaluation criteria when selecting students, and so on. In short, the

U.S. News-Ranking proved to be highly reactive.

Following this logic, we would expect ranking-based evaluation procedures in

economics to induce similar systematic incentives. Since journal rankings play a core

role in economics and provide an arena for the mechanisms of commensuration and

self-fulfilling prophecies, such effects should eventually materialize on the level of

economic journals: we would expect reactive effects induced by the increasing

15
visibility and importance of journal rankings to affect journals more strongly than single

authors or articles as journals reside in the focal point of said rankings. This argument

implies a transmission of prestige towards scientific journals: if correct, all articles of

a highly ranked journal should become attractive simultaneously leading to a more

balanced distribution of citations within a highly-ranked journal. Figure 3 shows that

this expectation indeed holds for the case of economics by means of a period-based

comparison of citations to articles in five high-quality economic journals. We compare

the number of articles receiving at most one citation within a fixed time period (papers

published in 1981-1985 and 2004-2008, respectively, with citations counted up to 1990

and 2013, respectively.)

<Figure 3 near here>

The insight of de Solla Price (1965) that the majority of the scientific literature – in our

case the majority of articles in a journal – receives little to no attention from future

articles was already mentioned in Chapter 2 and has been confirmed empirically by

several works (Garfield 2006, p. 91; Nature 2005, Seglen 1996). In light of this, the

results shown in Figure 3 are quite surprising as they indicate a substantial reduction

in the number of articles that are neglected in terms of citations. However, at this point

one could argue that the change of this citation pattern may be due to an overall increase

of the citation frequency. Put differently, the decrease in neglect articles could point to

an intensification and diversification of economic research. Hence, we provide a more

nuanced analysis in Figure 4 by inspecting the changes in shares of received citations

across the whole distribution of papers published in each period. The resulting pattern

again supports our theoretical expectations – a decrease in the share of citations (by

roughly 9%) dedicated to the upper limit of the distribution is complemented by a

16
corresponding increase in lowest eight deciles.

<Figure 4 near here>

Overall, these findings weakly support our hypothesis that academic attention is

undergoing a shift away from attention focused on single authors and contributions to

attention being focused on highly ranked journals; further work to better disentangle

the sources of the observed patterns would be highly appreciated. In a first step,

additional evidence on this issue can be obtained from more large-scale empirical

analysis, which indicates that the distribution of attention and, hence, the discursive

properties of economics are increasingly concentrated upon a limited number of

authors, departments and journals (Glötzl and Aigner 2015). This latter trend thereby

stands in contrast to the general development of scientific discourse, which shows

patterns of an increasing diversification of attention for most disciplines (Larivière et

al. 2009).

Regarding the power of scientometrics, we find in this section that the currently

practiced form of scientometrics has the power to influence the behavior of researchers

and scientific institutions. In the case of economics, this identified power crystallizes

especially – as shown at the end of this chapter – in the institutional field of economic

journals.

4. Economic Discourses and the Influence of Scientometrics on the Future of

Economics

The scientific discourse in economics is different from other social sciences because it

17
is dominated by a single paradigm – neoclassical economic theory – which significantly

shapes academic teaching and economic research. We refer to Backhouse (2005) or

Dobusch and Kapeller (2009) for a discussion of the historical roots of this

paradigmatic dominance. In the domain of a more narrowly defined research discourse

especially the 1970s and 1980s stand out: in this period, not only Keynesian

macroeconomic approaches were driven out by neoclassical and monetaristic theories,

but also the journal culture experienced a significant theoretical narrowing. This led to

a virtually complete exclusion of critical and alternative scientific contributions from

the economic discourse; such articles were – with some exceptions – rejected in review

procedures. The resulting confrontation caused by this exclusion from an ‘official’

economic discourse led to the foundation of – nowadays eminent – heterodox-economic

journals, like the Cambridge Journal of Economics, the Journal of Post Keynesian

Economics or the Journal of Economic Issues (King 2003, p. 134-136). In this sense,

economics is still a ‘contested discipline’ (Lee and Elsner 2011). For the same reason

it is not surprising that alternative or heterodox schools of thought only constitute a

small fraction inside the economic discipline, since they are confronted which such

‘exclusion routines’ on several levels.

Of course, the description of this specific constellation per se does not suffice to draw

a conclusion regarding the representation of alternative economic approaches within

the mainstream economic discourse. As a consequence, the question of how heterodox

ideas are regarded within the mainstream discourse has to be answered primarily from

an empirical perspective. Past works analyzed the interaction between heterodoxy and

mainstream by comparing citation patterns associated with mainstream and heterodox

economic journals. A representative example of this literature is reproduced in Figure

18
5, which compares the relative citation flows between 26 economics journals (13

highly ranked mainstream and 13 heterodox) over a period of twenty years (1989-

2008).9 It shows that heterodox journals exhibit a quite balanced citation pattern

(heterodox and mainstream journals are cited equally), while the citation behavior of

mainstream journals is drastically in favor of other mainstream journals. In this view,

heterodoxy is more open or pluralist, whereas orthodoxy is relatively closed or monistic

– an assertion, that is reinforced by an inspection of absolute citation flows, i.e. net

transfer of ideas, as conducted below.

<Figure 5 near here>

A more detailed analysis of the data from Figure 5 furthermore shows that the

percentage of citations from the top thirteen heterodox journals exported into

mainstream journal literature considered here (2,85% of total references) is driven

heavily by statistical outliers. Measured in absolute figures, 2,85% represent 753

citations. Of these, the majority (613 citations) is caused by only three journals that

hold a special position within the economic discourse.10 Within the remaining 23

journals only a minimal transfer of ideas in the form of 140 citations within a period of

20 years can be found.

Even if Figure 5 serves as sufficient evidence for the underrepresentation of alternative

economic approaches in mainstream economic literature, it is still not clear whether the

observed pattern can be explained by a paradigmatic divide or by a strong focus within

economic discourse on articles published in highly ranked journals. After all, the

orthodox sample represents the top 13 from the Journal Citation Report 2007, whereas

the heterodox sample is scattered between ranking position 17 and 130. Considering

19
the high level of self-reference as well as a strong elitist orientation inherent to the

scientific discourse in economics, this is a strong argument. These aspects were recently

documented by Fourcade et al. (2015), who showed in detail that the economic

literature is – compared to other scientific disciplines – less inclined to refer to other

disciplines (self-reference) and has a stronger focus on a small group of top journals

whose authors primarily stem from a small and homogeneous quantity of universities

(concentration; see also Hodgson and Rothman 1999). This does not only refer to a

more focused attentiveness within the economic discipline but also a relatively tightly

structured and hierarchical internal organization (elitist orientation). Therefore, it is not

surprising that an analysis of the development of economic ‘top-journals’, recently

published in the Journal of Economic Literature, was limited to the observation of only

five journals (Card and DellaVigna 2013).

We will now replicate the procedure shown in Figure 5 with a corresponding control

group in order to understand to which extent the neglect of alternative theoretical

approaches is due to an elitist approach towards journal rankings. Figure 6 shows a

replication of Figure 5, where the heterodox sample is replaced by a control group

consisting of those thirteen mainstream journals in the JCR 2007 ranking which are –

in each case – one position below the heterodox journal sample. The result in Figure 6

shows that the observed discursive pattern cannot exclusively be explained by the

relative ranking position of the heterodox journals; on the contrary, the citation

frequency of the control group is more than three times higher than in the heterodox

sample. Indeed, paradigmatic factors seem to play a central role and suggest a

systematic discursive exclusion of alternative theoretical approaches.

<Figure 6 near here>

20
Of course, this aspect is not the only weakness in our examination about

interparadigmatic discourse in economics. The time period considered here – namely

before the crisis – could be another possible point of criticism. The financial and

economical crisis could be understood as a central anomaly in sharp contrast to basic

postulates of neoclassical theory (such as the efficient market hypothesis and the

associated arbitrage-based thinking). It is certainly conceivable that the crisis has

changed the perception of basic facts of economy and has lead to a more inclusive

pluralist approach.

A theoretical answer is provided by Kuhn (1962), who argues that a dominant paradigm

facing a significant anomaly will try to resolve this dissonance by an adaption of already

established theses, models and methods and, at the same time, aim to avoid fundamental

debates. While it is beyond question that economic research has changed in some way

in response to the recent crisis (e.g. Young 2014), a more nuanced analysis seems

necessary to assess whether these changes also led to an increased reception of

alternative economic approaches in mainstream outlets or whether this reaction follows

a Kuhnian pattern of an ‘internal’ adaption of existing models. Figure 7 provides a

further replication of Figure 5 which focuses – instead of the pre-crisis period – on

current journal literature from the period 2009 to 2013.

<Figure 7 near here>

Here again – supporting the prediction of Kuhn – no substantial change in the citation

behavior of the dominant paradigm can be observed; indeed, the behavior of the

21
economic mainstream remains widely constant. So far, the anomaly of the financial and

economic crisis has not intensified the reception of alternative theoretical approaches

within the mainstream economic discourse.

The following thesis can now be established: inner-scientific criteria, with citation

metrics and associated institutionalized evaluation routines among them, exert a

stronger influence on the distribution of attention in economics than the actual

economic development (the most drastic example being the financial and economic

crisis). One major reason for this lies in the ‘size-bias’ of citation metrics: by definition,

citation metrics certify quantitatively meaningful research disciplines like mainstream

economics as high-quality disciplines. Especially in the case of economics as a

‘contested discipline’ characterized by an extremely unequal distribution of initial

resources, it is obvious that the consideration of scientometric criteria will further

stabilize the dominant economic paradigm.

First indications for such a development, for example, can be found in France, where

citation metrics have started to play a significant role in the centralist appointment

policy of professors in economics: citation metrics were included in a formal scoring

system in 2005. Assuming the same level of productivity, heterodox economists

achieve – due to this ‘size-bias’ of citation metrics – a significantly lower score, making

the appointment of heterodox economists appear less attractive. This circumstance is

indeed reflected in the French appointment policy: parallel to the introduction of the

scoring system, the relative amount of heterodox economists newly appointed fell from

almost 18% in the period of 2000-2004 to 5% in the period of 2005-2011 (FAPE 2014).

Similar trends can also be identified in other countries where evaluation routines based

22
on citation metrics are used. For example, in Great Britain the so-called Research

Assessment Exercise, which evaluates thematically related institutes, also influences

the appointment policy of these institutes. Hereby, ranking criteria are anticipated in

favor of mainstream economics, leading to a positive discrimination of mainstream-

economic applicants (see also Lee 2007, who analyzes the situation at the beginning of

the 2000s).

Summarizing, we see that another effect of evaluative scientometrics lies in the

stabilization of the dominant role of a prevailing scientific paradigm and therefore

promotes an increasing homogenization of scientific disciplines.

5. Conclusion

The purpose of this examination was to provide a theoretical overview on the

consequences of scientometric evaluation routines with a special emphasis on

economics. Three main findings could be identified: firstly, evaluative scientometrics

contribute to a reinforcement and stabilization of existing patterns of academic

reproduction and further increase any existing bias in the distribution of attention,

prestige and resources (Chapter 2). Secondly, there exist various instances of reactive

effects, which are a natural consequence of employing evaluative scientometrics and

actively influence the behavior of scientific agents (Chapter 3). Thirdly evaluative

scientometrics significantly contribute to the stabilization of a dominant economic

paradigm and limit the influence of alternative or critical approaches within the

scientific discourse (Chapter 4). Illustrative evidence was supported to indicate how the

discussed mechanisms can be assessed empirically and to provide inspiration for further

23
research on the role of evaluative citation metrics in academic reproduction and

scientific development.

24
References

Adler, Robert, Ewing, John and Peter Taylor. “Citation Statistics.” Report from the
International Mathemaical Union, 2008. Accessed March 10, 2015.
http://www.mathunion.org/fileadmin/IMU/Report/CitationStatistics.pdf
Arthur, Brian W. Increasing returns and path-dependence in the economy. Michigan:
University of Michigan Press, 1994.
Backhouse, Roger E. “The Rise of Free Market Economics: Economists and the Role
of the State since 1970.” History of Political Economy 37, 5 (2005): 355-392.
Bloch, Harry. “Research Evaluation Down Under: An Outsider's View from the Inside
of the Australian Approach.” The American Journal of Economics and Sociology 69, 5
(2010): 1530-1552.
Card, David and Stefano DellaVigna. “Nine facts about top journals in economics.”
Journal of Economic Literature 51, 1 (2013): 144-161.
Corsi, Marcella, D’Ippoliti, Carlo and Federico Lucidi. “Pluralism at Risk? Heterodox
Economic Approaches and the Evaluation of Economic Research in Italy.” The
American Journal of Economics and Sociology 69, 5 (2010): 1495-1529.
David, Paul A. “Clio and the economics of QWERTY.” The American Economic
Review 75, 2 (1985): 332–337.
Demange, Gabrielle. “On the influence of a ranking system.” Social Choice and
Welfare 39, 2 (2012): 431-455.
Diamond, Arthur M. “The Core Journals of Economics.” Current Contents, 1 (1989):
4-11.
Dobusch, Leonhard and Jakob Kapeller. “Why is Economics not an Evolutionary
Science? New Answers to Veblen's old Question.” Journal of Economic Issues 43, 4
(2009): 867-898.
Dobusch, Leonhard and Jakob Kapeller. “Heterodox United vs. Mainstream City?
Sketching a framework for interested pluralism in economics.” Journal of Economic
Issues 46, 4 (2012a): 1035-1057.
Dobusch, Leonhard and Jakob Kapeller. “A guide to paradigmatic Self-marginalization
- Lessons for Post-Keynesian Economists.” Review of Political Economy 24, 3 (2012b):
469-487.
Dobusch, Leonhard and Elke Schüßler. “Theorizing Path Dependence: A Review of
Positive Feedback Mechanisms in Technology Markets, Regional Clusters and
Organizations.” Industrial and Corporate Change 22, 3 (2013): 617-647.
Espeland, Wendy N. and Michael Sauder. “Rankings and Reactivity: How Public
Measures Recreate Social Worlds.” American Journal of Sociology 113, 1 (2007): 1-
40.
Espeland, Wendy N. and Mitchell L. Stevens. “A Sociology of Quantification.”
European Journal of Sociology 49, 3 (2008): 401-436.

25
FAPE. “Evolution of Economics Professors’ Recruitment since 2000 in France: The
End of Pluralism.” Working Paper, 2014. Accessed February 30, 2014.
http://www.heterodoxnews.com/HEN/attach/hen167/FAPE_pluralism_france.pdf
Fourcade, Marion, Ollion, Etienne and Yann Algan. “The superiority of economists.”
Journal of Economic Perspectives 29, 1 (2015): 89-114.
Garfield, Eugene. “The History and Meaning of the Journal Impact Factor.” Journal of
the American Medical Association 295, 1 (2006): 90-93.
Glötzl, Florentin and Ernest Aigner. “Pluralism in the Market of Science? A citation
network analysis of economic research at universities in Vienna.” Ecological Economic
Papers 5, 2015. WU Vienna University of Economics and Business, Vienna. Accessed
August 09, 2017. http://epub.wu.ac.at/4730/1/EcolEcon_WorkingPaper_2015_5.pdf
Hodgson, Geoffrey M. and Harry Rothman. “The editors and authors of economics
journals: a case of institutional oligopoly?” The Economic Journal 109, 453 (1999):
F165-F186.
ISIPE. “An international student call for pluralism in economics.” 2014. Accessed July
26 , 2016. http://www.isipe.net/open-letter/
King, John E. A History of Post-Keynesian Economics since 1936. Cheltenham:
Edward Elgar, 2003.
Larivière, Vincent, Gingras, Yves and Éric Archambault. “The decline in the
concentration of citations, 1900–2007.” Journal of the Association for Information
Science and Technology 60, 4 (2009): 858-862.
Lee, Frederic S. “The Research Assessment Exercise, the state and the dominance of
mainstream economics in British universities.” Cambridge Journal of Economics 31, 2
(2007): 309-325.
Lee, Frederic S. and Wolfram Elsner. Evaluating Economic Research in a Contested
Discipline: Ranking, Pluralism, and the Future of Heterodox Economics. John Wiley
& Sons, 2011.
Levy, Moshe and Haim Levy. “Investment talent and the pareto wealth distribution:
Theoretical and experimental analysis.” The Review of Economics and Statistics 85, 3
(2003): 709-725.
Kuhn, Thomas. The Structure of Scientific Revolutions. Chicago: University of Chicago
Press, 1962.
Merton, Robert K. “The Matthew Effect in Science.” Science 159, 3810 (1968): 56-63.
Mingers, John and Loet Leydesdorff. “A Review of Theory and Practice in
Scientometrics.” Working Paper, 2015. Accessed 23 February 2015.
http://arxiv.org/abs/1501.05462
Nature. “Not-so-deep impact.” Nature 435, 7045 (2005): 1003-1004.
Newman, Mark E. J. “Power laws, Pareto distributions and Zipf's law.” 2006. Accessed
December 10, 2014. http://arxiv.org/abs/cond-mat/0412004
Rip, Arie and Jean-Pierre Courtial. “Co-Words Maps in Biotechnology: An Example
of Cognitive Scientometrics.” Scientometrics 6, 6 (1984): 381-400.

26
Seglen, Per O. “Why the impact factor of journals should not be used for evaluating
research.” British Medical Journal 314, 7079 (1997): 498-502.
Shapiro, Carl and Hal R. Varian. Information Rules: A Strategic Guide to the Network
Economy. Boston: Harvard Business School Press, 1999.
De Solla Price, Derek J. “Networks of Scientific Papers.” Science 149, 3683 (1965):
510-515.
Sterman, John D. and Jason Wittenberg. “Path Dependence, Comeptition, and Success
in the Dynamics of Scientific Revolution.” Organization Science 10, 3 (1999): 322-
341.
Sydow, Jörg, Schreyögg, Georg and Jochen Koch. “Organizational Path Dependence:
Opening the Black Box.” Academy of Management Review 34, 4 (2009): 689–709.
Young, Brigitte. “Financial Crisis: Causes, policy responses, future challenges.”
Directorate-General for Research and Innovation, European Commission, Brussels,
2014. Accessed February 24, 2016. http://ec.europa.eu/research/social-
sciences/pdf/policy_reviews/KI-NA-26554_EN-C.pdf

27
Figures

Figure 1: Comparing four distributions (author’s own graph)11

Figure 2: The role of evaluative scientometrics as additional feedback loop (dashed line) in stabilizing
a scientific paradigm (author’s own graph based on Sterman and Wittenberg 1999, 333).

28
 Neglected papers in top economic journals

Quarterly Journal Review of American Journal of

Econometrica
30%
of Economics Economic Studies Economic Review Political Economy

25%
10.49 %

20%
one citation
10.33 % 12.51 % zero citations
15%

8.64 % 7.01 %
10%
18.35 %

5% 10.7 % 9.89 %
6.91 % 7.32 %
3.44 % 0.99 %
2.21 %
1.28 % 2.48 %
0.49 % 0.85 % 1.1 % 1.33 %
0% 0.49 %
1981-1985 2004-2008 1981-1985 2004-2008 1981-1985 2004-2008 1981-1985 2004-2008 1981-1985 2004-2008
(citations (citations (citations (citations (citations (citations (citations (citations (citations (citations
up to 1990) up to 2013) up to 1990) up to 2013) up to 1990) up to 2013) up to 1990) up to 2013) up to 1990) up to 2013)

Figure 3: Comparison of citations to published articles in five prestigious economic journals (authors’
own graph based on data from Thomson Scientific). Note: Published document types such as editorials,
book reviews or biographic items were excluded in this analysis.

Figure 4: Relative change of citation patterns in top journals (journal sample from Figure 3). (authors’
own graph based on data from Thomson Scientific).

2.85%

97.15% Orthodox Heterodox 52.42%

47.58%

Figure 5: Interaction pattern between mainstream economics and heterodoxy (1989-2008; taken from
Pre-crisis Flow of citations between Top 13 Heterodox and Top 13 Orthodox
Dobusch and Journals
Kapeller 2012b, p. 474).

29
8.91%
 97.15% Orthodox Heterodox 52.42%

2.85% 47.58%

Pre-crisis
97.15% Flow of citations between Top 13
Orthodox Heterodox52.42%
Heterodox and Top 13 Orthodox
Journals

47.58%
8.91%
Pre-crisis Flow of citations between Top 13 Heterodox and Top 13 Orthodox
Journals

91.09% Orthodox Control Group 33.85%

8.91%

91.09% Orthodox
66.15%
Control Group 33.85%

Figure 6: Interaction pattern between top mainstream journals and a control group (1989-2008; author's
own calculation based on data
Pre-crisis from
Flow of citations 66.15%
Thomson Scientific).
between
12
Top 13 Heterodox and Control Group

Pre-crisis Flow of citations between Top 13 Heterodox and Control Group

2.39%
2.39% (2.85%)
97.61% Orthodox Orthodox 74.42%
97.61% (97.15%) Orthodox Heterodox 74.42% (52.42%)

24.58%
24.58% (47.58%)

Figure 7: Interaction pattern

Post-crisis
Post-crisis of between
FlowFlow
citations mainstream
of citations
between between
Top Topeconomics
13and
13 Heterodox Top 13and
Heterodox andheterodoxy (2009-2013; author's own
Top 13 Orthodox
Orthodox
calculation based on data from Thomson Scientific; values from Figure 5 in brackets).

1
1

30
Notes

1
Examples for internationally known rankings in academia are the Times Higher Education World Reputation
Ranking or the QS World University Ranking. In the field of journal evaluation rankings like the Journal Citation
Reports from Thomson Scientific stand out.
2
For economists, such distributions have been known for a long time, especially in the context of the analysis of
income and wealth distributions, where they are discussed on the basis of the Pareto-formula.
3
The citations assigned to these articles stem from 1981-1990.
4
‘For to every one who has will more be given, and he will have abundance; but from him who has not, even what
he has will be taken away.’ (Mt 25,29)
5
Although citation data has already been widely used in the 20th century to assess the relative impact of academic
literature (in economics, already the traditional ‘Diamond list’ of distinguished economics journals, published in
Diamond 1989, was based on citation data), its direct institutional impact has strongly increased in the last fifteen
years due to the introduction of various means of quantitative research evaluation (e.g. Lee 2007) and the
associated emergence of popular journal rankings, most notably ISI’s/Thomson Scientific’s „Journal Citation
Reports“ (starting in 1999).
6
We obtained this data through personal correspondence with a series of editorial offices.
7
See also: http://vhbonline.org/service/jourqual/vhb-jourqual-3/
8
Therefore, the brochures are sent primarily to other law schools (cf. Espeland und Sauder 2007, 26).
9
The basis for our sample was the Web of Science database from Thomson Reuters. For the selection of the
relevant journals the Journal Citation Report 2007 was used, whereby the thirteen best ranked journals were
interpreted as ‘top 13 orthodox’ journals. The ‘top 13 heterodox’ where identified by means of the Heterodox
Economics Directory (www.heterodoxnews.com/hed).
10
On the heterodox side of the sample, the Journal of Economic Behavior and Organization, which is attractive
both for heterodox and mainstream articles account for 340 exported citations. On the mainstream side of the
sample, two journals from the field of economic geography (Journal of Economic Geography und Economic
Geography) are included, which import further 273 citations. In this connection, the relative openness of economic
geography for heterodox approaches is disproportionately higher than in strict mainstream economics.
11
Data sources: The population of the German cities is taken from the Mathematica10-database. The estimate of
property assets is based on data from the Household Finance and Consumption Survey (HFCS) of the Austrian
national bank. The number of citations to the American Economic Review is retrieved from the Web of Science
database and the birth weight data of the newborns represents the collected number of births in the first two weeks
of February 2015, which took place in an Austrian hospital. All data are available on request.
12
A comparison of the relative change in citation behavior of the orthodox journals with data from Figure 5 shows
a statistically significant difference on the 1%-level.

31