Is economics a house divided? Analysis of citation networks.

Onder, Ali Sina ; Tervio, Marko

I. INTRODUCTION

We ask whether the academic discipline of economics is divided into clusters of universities where authors tend to cite authors from the same cluster more than could be expected under idiosyncratic differences in citation patterns. We use citation data between top economics journals from 1990 to 2010 to construct the citation matrix between authors' home institutions. We compare all possible partitions of top universities into two equal-sized clusters. We find a significant division between top universities in this citation network, and it is consistent with what is commonly thought as the divide between "freshwater" and "saltwater" schools.

The likelihood of citing a paper by an author from another university in the same cluster is about 16% higher than the likelihood of citing a paper by an author from the other cluster. We assess the statistical significance of this division using simulations. In each simulated citation network, the likelihood of citation propensities is independent across university pairs, while average citation propensities and the distribution of pairwise deviations from average propensities at each university match their empirical counterparts. The division is statistically extremely significant and is robust to considering different extents of "top universities" and time periods. However, there are significant differences across fields of economics, with macroeconomics and econometrics exhibiting the strongest division, whereas finance and international economics exhibit rather weak division.

II. DATA

A. Data Sources

We use the citation data of articles published in 102 economics journals between 1990 and 2010, where the set of top journals was taken from the classification by Combes and Linnemer (2010). (1) The data were obtained from Thomson Scientific's Web of Science, which is an online database pooling journal articles' data from major databases including Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI), and Arts and Humanities Citation Index (A&HCI). Notes, editorials, proceedings, reviews, and discussions were not included. The resulting data cover 97,526 unique articles with 34,431 unique contact authors and 1,187 unique affiliations associated with these contact authors.

Our data set contains information on articles cited in the reference sections of these articles. Data on cited articles consist of year of publication, name of the journal, and name of the contact author. (2)

B. Construction of the Citation Matrix

We use articles published between 1990 and 2010 and articles cited by them to construct a citation matrix between institutions. Data on contact authors of citing articles also contain their affiliation at the time of publication. However, author affiliations for cited articles are not directly observed. Hence, we construct a career path for each author from 1977 to 2010 by using affiliation information of citing articles. For this task, we also use data on articles published between 1977 and 1989, in order to enlarge the set of cited articles that can be matched with an author affiliation. If an author did not publish in our sample journals in a year, then we use his or her next known affiliation; if no affiliation is observed between the cited year and 2010, then we use the last previously observed affiliation. Using this procedure, we are able to identify 36,189 unique authors of a total of 1,662,212 cited articles in the reference sections of 91,635 unique articles written by 32,572 unique authors. Authors of a total of 753,230 cited articles could not be matched with an affiliation. The observed affiliations form a total of 1,187 citing and 1,192 cited institutions.

We measure citations in units, so that every article conveys one unit of citations, regardless of how many documents it cites. For example, if an article by an author from MIT cites 20 articles and 4 of them by Harvard authors, then this counts as 4/20 = 0.2 units of citations from MIT to Harvard. (3) Cited publications whose author cannot be matched with an affiliation are treated as authored at an institution called "unknown."

Citation data are gathered in the aggregate citation matrix, which gives the sum of unit citations from all articles. The element at row i and column j is the sum of unit citations by authors from institution i to articles by authors from institution j. To analyze subsets of institutions, we just keep the relevant submatrix of the aggregate citation matrix; when analyzing subsets of journals and publication years, we restrict the underlying summation to subsets of articles.

Figure 1 shows the distribution of articles in our data by publication year. Steady increase in the annual number of articles reflects an increase in the number of journals as well as increase in articles per journal-year. Of the 102 journals in the set, 79 were in existence in 1990 and 96 in 2000. The average number of articles published in a journal per year increased from 50 in 1990 to 54 in 2000 and to 73 in 2010. Figure 1 also shows the distribution of unit citations that are used in the construction of our citation matrix by publication year. The number of "cites out" and "cites in" in a given year refers to the amount of unit citation for which an author affiliation could be identified, respectively, for citations made and citations received.

III. ANALYSIS

Our goal is to find out whether institutions can be divided into "clusters" within which authors cite each other more than could be expected under idiosyncratic citation patterns. The existence of discrete clusters is, of course, an abstraction; the point of this exercise is to uncover a dimension of differentiation in the citation patterns of institutions. Self-citations are a serious confounding factor, because citations within an institution are necessarily also within-cluster citations. Over 10% of the cites in our data are institutional self-cites. (4) We ignore all self-citations, effectively replacing the diagonal elements of the citation matrix with zeroes.

To measure clustering, we use a slightly modified version of 2-modularity of Girvan and Newman (2002). (5) For a given partition of institutions to clusters, Q measures the difference between the actual and the expected proportion of cites between clusters, where the expectation is calculated under independently distributed citation patterns. The strongest division in the network is that which maximizes modularity. Our additional normalization takes into account the impact of removing self-citations on expected citation patterns. Without this correction, the expectation benchmark would always predict a significant amount of self-citations. With the correction, expected self-citations are set to zero. Intuitively, the expected citation patterns are calculated under the hypothesis that authors at all institutions distribute their outbound nonself cites at a probability that depends only on target institution, not on sender institution. Analyzing proportions instead of cite counts also serves as a normalization that gives each institution equal weight in defining the strength of deviations from expectation, regardless of its share of all citations.

[FIGURE 1 OMITTED]

Denote the aggregate citation matrix for the set of n institutions by M. The normalized citation matrix T has typical elements

(1) [T.sub.ij] = [M.sub.ij]/[summation over (h[not equal to]i)][M.sub.ih]

and we set [T.sub.ii] = 0. Row i measures citations as proportions of outbound nonself cites from institution i. We define its expectation as the average fraction of nonself citations by departments other than i going to department j: 2 *

(2) [E.sub.ij] = [1/(n-2)] [summation over (h[not equal to]i)][T.sub.hj] for h [not equal to] j

and [E.sub.ii] = 0, for i=1, ..., n. Finally, the citation information that is used in the analysis is contained in the matrix of deviations from expected citation patterns [OMEGA] = T - E.

Table 1 shows the unit citations between the top 20 academic institutions, that is, the matrix M. The background colors represent a heat map of the pairwise deviations from expected citation patterns, that is, the elements of [OMEGA]. If a row department cites a column department more (less) than expected, then the corresponding element is red (blue), while darkness captures the magnitude of the deviation. Consider, for example, the element at first column and second row, 4.6. It is the sum of unit citations made by articles with a contact author at the University of Rochester to articles where the contact author is affiliated with the University of Minnesota. It could mean, for example, that there were 46 articles by Rochester authors that cited Minnesota authors and that 10% of the citations in each of those articles referred to articles by Minnesota authors, giving a total of 4.6 unit citations. Moreover, the relatively dark shade of this cell reveals that 4.6 is clearly above the expected number of unit citations from Rochester to Minnesota, where the expectation is based on the total amount of (nonself) unit citations made and received by these two institutions in our data.

We consider all partitions of the set of n institutions into two equal-sized clusters. (6) Formally, consider any partition of the set of n institutions into subset A and its complement. We measure the strength of the division as

(3) Q(A|[OMEGA]) = [e'.sub.A][[OMEGA]e.sub.A] + (l-[e.sub.A])' [OMEGA](l-[e.sub.A]),

where [e.sub.A] is the membership vector for subset A, equal to unity for members and zero for nonmembers, and l is a vector of ones. This measure gives the sum of total deviations from the expected proportion of normalized citations for within-cluster pairs of institutions. (Deviations add up to zero, so the amount of deviations for between-cluster pairs of institutions is necessarily just the negative of Q and can be omitted.)

We define the strongest division to be the partition of A to two clusters of n/2 institutions that maximizes (3). (7) Thus, for a set of n institutions, with n even, there are [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] distinct ways of dividing them to two equal-sized clusters. We use brute force to select the strongest of all possible partitions.

IV. RESULTS

There are authors from 1,192 institutions in the data. To analyze their possible division, we restrict the analysis to a subset of top institutions. We define the "top" by the ranking of institutions by influence in the network of citations using eigenvalue centrality; for details, see Pinski and Narin (1976). (8) Self-citations are removed before calculating influence. Table 2 lists the influence measure for the top 50 institutions by influence. Our main specification considers the division between the top 20 academic institutions. Selected summary statistics of the citation matrix are also reported in Table 2. Self-cites, which are excluded in the analysis, are reported separately. Cites to articles whose contact author could not be matched with an institution are listed as "cites to unknown." All cites are measured in units-per-citing-article, so the sum of outgoing cites, self-cites, and cites to unknown adds up to the total number of articles published by contact authors from each institution.

A. Clustering Results

The strongest division is depicted in the last columns of Table 2 for n = 12, 16, 20, 24. We call the cluster that includes Harvard "the saltwater cluster" and the other "the freshwater cluster." Most departments always show up in the same cluster. Chicago, Northwestern, Penn, and Rochester are always in the freshwater cluster; MIT, Stanford, Princeton, Berkeley, and Columbia are always in the saltwater cluster. The only institutions whose cluster membership varies by specification are Yale and Michigan. The division is the same as was found in hiring/placement data in Tervio (2011). (9)

The magnitude of the division can be illustrated by considering the relative propensities to cite within and between clusters. Among the top 20 academic institutions, the average number of unit citations between a pair of institutions in different clusters is 11.76, while the average for institution pairs in the same cluster is 13.67, that is, 16.2% higher. Among the top 16 academic institutions, the average number of unit citations between a pair of institutions in different clusters is 14.91, while the average for institution pairs in the same cluster is 17.32, that is, 16.1% higher.

We also applied two alternative clustering algorithms, the Louvain method (using the Pajek software package) and MapEquation (see Rosvall and Bergstrom 2011 for details). For n = 24, both methods yield the same division as our analysis, when restricted to yield two clusters of equal size. Without this restriction, the Louvain method moves Michigan and UCSD to the freshwater cluster, while MapEquation finds that the division to equal-sized clusters is in fact optimal. For n = 20, Pajek finds the same clusters as we do, whereas MapEquation finds no division at all (i.e., just one cluster). Both algorithms find one cluster optimal for n = 16 and n = 12.

B. Strength of Attachment

The relative strength of attachment to the saltwater and freshwater clusters can be measured for any institution that hosts authors that publish in our sample of journal articles. More precisely, redefine [OMEGA] to include all departments and not just the top n. We define the "salt content" of department i as

(4) [S.sub.i] = {([e'.sub.i][OMEGA][e.sub.S])/[(l - [e.sub.i])'[e.sub.s]]} -{([e'.sub.i][OMEGA][e.sub.F])/[(l - [e.sub.i])'[e.sub.F]]},

where [e.sub.i] is the ith unit vector and [e.sub.s] and [e.sub.F] are the membership vectors of saltwater and freshwater clusters. The divisors account for the removal of self-cites; top institutions are themselves members of a cluster and have one less potential citation partner in their own cluster. Finally, "relative salt" is obtained by subtracting the mean salt content of all departments (.385).

Table 2 lists the "relative salt" measure for the 50 most influential institutions. It measures the average deviation from the expected share of outgoing citations to saltwater members in excess of the share going to members of the freshwater cluster. True to name, the saltiest of saltwater schools appear to be Berkeley and MIT, while Minnesota and Rochester are the freshest of the fresh. Chicago appears surprisingly "neutral" along with Stanford, Yale, and Columbia. Note that self-citations were removed from the analysis and Chicago is by far the most heavily cited freshwater department, so a disproportionate share of its citations to the freshwater cluster is ignored in the analysis. Outside academia, the Federal Reserve Bank appears quite "fresh," while World Bank and IMF are somewhat "salty."

[FIGURE 2 OMITTED]

The joint pattern of attachment to clusters and influence in the citation network is depicted in Figure 2. The rough pyramid shape of the scatter plot shows that more influential institutions appear to be less "partisan" in terms of the salt/fresh division.

V. IS THE DIVISION STATISTICALLY SIGNIFICANT?

Given the large number of possible partitions, it would often be possible to find partitions where the division appears strong even for a random pattern of deviations. It could also be that the anecdotal evidence of a division in economics is based on people attributing meaning to essentially random variation. To test the statistical significance of the division, we have to take into account that the partition has been selected from the set of possible partitions precisely in order to maximize the strength of the apparent division. Our concern is not that we would find spurious clustering due to random variation at the level of citations or publications but rather we might confound a random collection of strong links between departments with clustering.

We measure the statistical significance of the division by comparing the strength of the strongest division found in the actual sample to its bootstrapped distribution. The bootstrap distribution is obtained by generating random permutations of the deviation matrix [OMEGA] and measuring the strength of the strongest division found for each permutation. In these permutations, we randomly reorder the off-diagonal elements of [OMEGA], separately for each column, treating all possible permutations as equally likely. These simulated deviation matrices describe a world where the average share of incoming citations is held fixed for each university, but deviations from average nonself citation patterns are idiosyncratic. In the simulation, the distribution of pairwise deviations [[OMEGA].sub.ij] is the same as in actual data, but a tendency to cite a particular institution more does not imply a tendency to cite another particular institution more.

[FIGURE 3 OMITTED]

The strongest partition under the random benchmark always appears "statistically significant" to a naive test that treats the strongest partition as given. We conducted 10,000 simulations for each n = 12,16,20 and 2,000 for n = 24. In all of these simulations, there is only one instance where it is possible to find a division as strong as we find in the actual data, for n = 16. Therefore, we conclude that the division is statistically very significant. This simulation also helps illustrate the 16% magnitude of the "excess" within-cluster citations by showing how far it is in the tail. In these simulations, the strongest partition results in a magnitude this large in 0.3% of the cases, the 95th percentile of the excess is 8.5%, and the 99th percentile is 13.1%.

VI. SUBSAMPLES

A. Time Periods

We repeat the cluster analysis for a subset of citation years, using a rolling 10-year window starting from 1990-1999 and ending in 2001-2010, with the set of departments fixed at the top 16 academic departments as calculated for the whole time period. The clusters in the strongest division are exactly the same throughout the period, but there appears to be a secular trend toward a weaker division. The excess percentage of cites for within-cluster pairs (over between-cluster pairs) declines from 18.9% to 13.7% between the first and the last window. After running 10,000 simulations for each window, we find that the division is always statistically very significant, but with p value increasing from 0 to .0009 over time.

The time series results are summarized in Figure 3, which plots the strength of attachment to saltwater cluster (as defined in that period) against the last year of the 10-year time window. A noticeable development is the increasing "saltiness" of Chicago. Toward the end of the period, Chicago has a higher relative propensity to cite authors at saltwater schools than the average of all institutions. Despite this, Chicago shows up in the freshwater cluster in every time period, because it is so heavily cited by other freshwater departments. Even though Chicago appears "more salty" than some of the saltwater departments, an alternative partition where it switches places with a weakly attached saltwater department would result in more cross-cluster citing and make the division weaker.

B. Fields

We analyze the citations between the subset of four most influential field journals for nine fields, with journal fields defined by Combes and Linnemer (2010). Unfortunately, we do not have the JEL codes by article, so we do not include articles in general interest journals. The definition of "most influential" journals is based on the same influence measure as for institutions in the previous section, calculated from the matrix of unit citations between all 102 journals in our data. See Table A1 in the Appendix for summary statistics by journal. We also list our influence measures for these journals so as to provide an alternative ranking based on the citation patterns between them.

Table 3 shows the strongest division in each field. The analysis is in each case conducted for the 16 most influential academic departments in the citation network of that field. We define the p value as the fraction of simulations where the strongest division to two clusters is as strong as or stronger than the one found in actual data. With all journals included, this p value is .0001. Among the fields, macroeconomics and econometrics have the strongest division, at p = .000. Micro theory (.025), public economics (.027), and growth/development (.058) also exhibit a clear division, while the remaining fields show only weak evidence for a division. (10) In terms of the excess likelihood of citing same-cluster authors, the highest "biases" are found in econometrics (40.8%) and macroeconomics (32.6%), while for a moderately clustered field like micro theory, this "bias" is only 13.8%. To illustrate the size of these bias measures, we also list the 95th percentile of the same measure in the simulations under the random benchmark; they vary between 8% and 16% by field.

Table 4 shows the variation in the cluster membership of top departments across fields and highlights the differences from the saltwater-freshwater division found in the overall sample (as seen in Table 2). Clearly, there is significant variation in the memberships across fields, even if we only considered those where the division is statistically significant. Some groups of departments like Berkeley-Harvard-MIT and Chicago-Northwestern-Rochester are quite consistently found together, whereas Stanford, Yale, Columbia, and Michigan appear very inconsistent in their affiliations. The seemingly random affiliation of the latter departments is consistent with the fact that they are only weakly attached to their cluster in the main analysis (i.e., they have "relative salt" close to zero, see Figure 2). As our clustering method forces all departments to belong to one cluster or another, it is not surprising that weakly attached departments swing about rather randomly between clusters. The real outlier among the fields is econometrics, where the division is significant and yet looks very different from that found in the full sample (e.g., it is the only field where MIT is not on the same side with Harvard). The division in macroeconomics is almost identical to the overall division. This raises the question whether the overall division is driven by the division in macro. For this reason, we construct a sample that combines the field data but leaves out macro. The second to last column of Table 4 shows that the resulting division is almost identical to the overall division. Moreover, we analyze top five general interest journals and find a division very similar to the overall division. Divisions that we find in both cases (top five general interest and all fields excluding macro) have high statistical significance (p value is .003 for both).

VII. CONCLUSION

Stanford economist Robert E. Hall first came up with the freshwater/saltwater term in the 1970s, based on the then workplaces of a group of leading macroeconomists with a distinctive style of research: Robert E. Lucas at Chicago, Thomas Sargent at Minnesota, and Robert Barro at Rochester. (11) More recently, Gregory Mankiw (2006) has argued that the freshwater/saltwater division had become an issue of the past already by the 1990s, because "... science progresses retirement by retirement. As the older generation of protagonists has retired or neared retirement, it has been replaced by a younger generation of macroeconomists who have adopted a culture of greater civility" (p. 38). We do not have a measure of civility, but, in terms of the citation flows between economics departments, the saltwater/freshwater division is clearly not yet a matter of the past.

The network of citations in economics in articles published during the 1990s and 2000s exhibits a division where authors are significantly less likely to cite articles by authors at universities across the divide. The division adheres to the common notions of "freshwater" and "saltwater" schools. We find a 16% excess likelihood of citing same-cluster authors, which is statistically very significant, but, in terms of magnitude, very far from having two isolated schools of thought. When restricting the citations to top field journals, the strongest divisions are found in macroeconomics and econometrics. Citation data cannot reveal whether the divisions are based on methodological or ideological differences, but it seems clear that a purely geographical explanation would not work. Some of the divisions may be explained by a tendency to cite former colleagues and mentors, as the same division has earlier been found (Tervid 2011) in the network of PhD placements. Hiring networks and specialization could conceivably explain divisions in a field like econometrics where it would be harder to argue for ideological reasons behind the clustering.

ABBREVIATIONS

A&HCI: Arts and Humanities Citation Index

SCI-EXPANDED: Science Citation Index Expanded

SSCI: Social Sciences Citation Index

doi: 10.1111/ecin.12164

Online Early publication November 9, 2014

APPENDIX: SUPPLEMENTARY TABLES

APPENDIX: SUPPLEMENTARY TABLES

TABLE A1
Summary Statistics and Influence by Journal, 1990-2010

Rank   Journal Title                  Articles   Cites In   Cites Out

1      Econometrica                       995    3,071.69      297.64
2      American Economic Review         3,222    2,751.80      959.97
3      Journal of Political Economy       857    1,895.59      305.10
4      Quarterly Journal of               783    1,476.68      258.37
       Economics
5      Review of Economic Studies         778    1,036.27      348.94
6      Journal of Finance               1,373    1,245.49      538.25
7      Journal of Economic Theory       1,764      981.51      671.10
8      Journal of Financial             1,018      867.63      426.41
       Economics
9      Journal of Econometrics          1,721      781.87      586.13
10     Journal of Monetary              1,130      748.34      462.90
       Economics
11     Rand Journal of Economics          766      536.83      323.02
12     Review of Economics and          1,195      646.46      458.65
       Statistics
13     Journal of Public Economics      1,500      598.69      536.08
14     Journal of Economic                753      418.99      161.28
       Perspectives
15     Review of Financial Studies        762      340.58      400.31
16     International Economic             964      431.28      448.29
       Review
17     Economic Journal                 1,449      608.13      526.92
18     Journal of Economic                 75      440.67       17.58
       Literature
19     Games and Economic Behavior      1,291      270.67      494.48
20     Journal of the American          2,231      321.31       80.22
       Statistical Association
21     Economics Letters                4,261      389.55    1,926.21
22     European Economic Review         1,504      417.39      572.00
23     Journal of Labor Economics         555    2,386.83      222.88
24     Journal of International           989      400.76      375.90
       Economics
25     Journal of Business &              786      291.34      323.55
       Economic Statistics
26     Journal of Business                481      241.70      238.15
27     Journal of Human Resources         609      261.85      182.79
28     Econometric Theory                 789      128.46      253.62
29     Journal of Law & Economics         449      193.87      123.18
30     Journal of Money Credit and        956      262.49      406.03
       Banking
31     Journal of Mathematical            888      154.60      267.44
       Economics
32     Journal of Economic Dynamics     1,485      223.62      596.44
       & Control
33     Journal of Financial and           648      171.71      360.01
       Quantitative Analysis
34     Economic Inquiry                   951      180.50      306.32
35     American Political Science         598      134.05       39.34
       Review
36     Public Choice                    1,535      141.09      373.79
37     Journal of Economic Behavior     1,512      156.23      535.27
       & Organization
38     Journal of Development           1,059      197.26      356.39
       Economics
39     Industrial and Labor               605      126.36      148.42
       Relations Review
40     Journal of Applied                 688      149.65      305.67
       Econometrics
41     Journal of Law, Economics, &       412       96.07      125.92
       Organization
42     Brookings Papers on Economic       193       78.60       45.50
       Activity
43     International Journal of           593       99.02      164.08
       Game Theory
44     Journal of Urban Economics         935      180.10      282.70
45     Journal of Accounting and          475       76.78       97.14
       Economics
46     Journal of Industrial              495      151.06      205.38
       Economics
47     Canadian Journal of              1,109      157.26      463.04
       Economics
48     Economica                          631      134.75      270.08
49     Social Choice and Welfare          849       78.83      246.92
50     Journal of Banking and           1,849      112.01      732.80
       Finance
51     Journal of Environmental           844      273.22      254.19
       Economics and Management
52     Journal of Risk and                451       94.33      132.35
       Uncertainty
53     Oxford Economic Papers             691      140.15      266.95
54     National Tax Journal               732       93.68      139.45
55     Scandinavian Journal of            641      118.45      266.88
       Economics
56     International Journal of           953      120.04      429.75
       Industrial Organization
57     Journal of Economic History        435       63.50       52.33
58     Review of Economic Dynamics        351       59.82      170.70
59     Journal of Health Economics        853      119.46      188.51
60     Oxford Bulletin of Economics       652      135.50      265.42
       and Statistics
61     American Journal of              2,140      195.83      421.21
       Agricultural Economics
62     Journal of Economics and           394       66.08      178.08
       Management Strategy
63     Journal of International           962      104.85      412.23
       Money and Finance
64     Regional Science and Urban         645       89.16      229.29
       Economics
65     Journal of Economic Growth         126       62.77       54.32
66     Economic Theory                  1,303       42.13      547.12
67     Econometric Reviews                117       51.00       49.96
68     Review of Income and Wealth        434       41.43       95.62
69     World Development                1,655       84.01      183.20
70     Land Economics                     628      131.39      161.31
71     Applied Economics                3,195       79.76    1,079.06
72     Journal of Comparative             593       41.07      155.98
       Economics
73     Explorations in Economic           325       27.86       58.49
       History
74     Economics of Education             661       35.76      165.33
       Review
75     Economic Development and           536       59.15      121.14
       Cultural Change
76     Journal of Financial               235       29.57      122.93
       Intermediation
77     Mathematical Finance               273       37.21       47.15
78     Macroeconomic Dynamics             347       29.21      172.18
79     Labour Economics                   432       39.73      188.20
80     Journal of Population              525       34.71      188.07
       Economics
81     Journal of Risk and                513       42.88      139.60
       Insurance
82     Journal of the European            323       19.28      134.57
       Economic Association
83     International Tax and Public       379       40.66      157.94
       Finance
84     Journal of Regulatory              456       30.34      143.91
       Economics
85     World Economy                      830       26.36      128.73
86     Journal of Real Estate             547       25,53      158.90
       Finance and Economics
87     Energy Journal                     402       29.66       78.14
88     Environmental and Resource         726       57.81      246.94
       Economics
89     Journal of Productivity            375       32.13      100.89
       Analysis
90     Water Resources Research         4,928       29.26       52.15
91     Journal of Economic                604       19.94      109.03
       Psychology
92     Health Economics                   868       30.69      175.07
93     Economic History Review            259       11.08       15.64
94     Experimental Economics             138       12.29       66.63
95     Resource and Energy                302       23.76      103.10
       Economics
96     Ecological Economics             1,429       31.96      193.34
97     Southern Economic Journal        1,164       11.00      350.01
98     Insurance: Mathematics and         900       14.48       77.02
       Economics
99     Journal of Economic                119        8.56       32.42
       Geography
100    Industrial and Corporate           188        6.01       26.17
       Change
101    Journal of Common Market           294        5.89       13.68
       Studies
102    Economy and Society                246        1.10         2.96
       Total                                     28,155.80   28,155.80

                    Cites to
Rank   Self-Cites     Other     Influence        Top Field

1         150.66      541.71      11.137
2         244.78    1,748.25       9.668
3          63.24      481.66       7.635
4          50.08      471.55       5.923
5          46.79      379.27       4.734
6         255.60      576.15       4.708    Finance
7         209.00      875.90       4.503    Theory
8         163.89      427.71       3.712    Finance
9         153.10      956.77       2.770    Econometrics
10        103.98      555.12       2.606    Macro/money
11         66.45      376.53       1.904    IO
12         38.82      689.53       1.846
13        124.04      833.89       1.764    Public
14         16.32      516.40       1.621
15         49.37      312.32       1.607    Finance
16         31.67      482.04       1.592
17         51.66      824.42       1.591
18          1.10       53.32       1.529
19         94.43      674.09       1.349    Theory
20        232.33    1,631.45       1.289    Econometrics
21        157.95    2,001.84       1.210
22         44.11      801.89       1.147
23         33.30      294.82       1.089    Labor
24         90.94      517.17       1.063    International
25         33.50      414.95       0.998    Econometrics
26         18.98      220.87       0.946
27         32.70      380.51       0.915    Labor
28         57.48      455.91       0.778    Econometrics
29         19.26      289.56       0.756
30         46.13      487.84       0.751    Macro/money
31         65.13      530.44       0.733    Theory
32         62.11      790.46       0.717    Macro/money
33         27.26      259.73       0.685    Finance
34         13.15      587.53       0.588
35         44.58      419.08       0.567    Public
36        132.74      961.47       0.535    Public
37         57.43      869.31       0.519    Theory
38         47.37      621.24       0.493    Growth/development
39         40.73      393.85       0.480    Labor
40         17.05      363.28       0.466
41         17.47      261.60       0.452
42          2.79      136.71       0.438    Macro/money
43         44.83      330.09       0.429
44        104.14      534.16       0.420
45         67.28      299.58       0.416
46         22.68      260.94       0.400    IO
47         35.67      586.29       0.398
48         13.23      342.69       0.358
49         66.52      508.56       0.342
50        125.43      959.77       0.339
51         83.08      497.73       0.325
52         40.57      265.08       0.318
53         18.05      398.01       0.313
54         59.95      468.60       0.309    Public
55         16.34      347.79       0.304
56         40.86      475.40       0.302    IO
57         21.63      328.04       0.286
58          8.14      170.16       0.283
59         63.63      566.86       0.267
60         18.32      359.26       0.267
61        244.94    1,295.85       0.260
62         11.40      202.52       0.241    IO
63         51.59      490.19       0.227    International
64         34.90      363.81       0.208
65          5.41       66.27       0.202    Growth/development
66          9.71      701.17       0.191
67          3.13       62.91       0.190
68         16.79      293.59       0.150
69         65.26    1,132.54       0.150    Growth/development
70         39.99      409.70       0.146
71        109.70    1,932.24       0.139
72         31.60      388.42       0.136
73          9.29      248.22       0.123
74         40.50      430.18       0.112
75         15.44      369.43       0.111    Growth/development
76          5.71      106.36       0.110
77         15.46      183.39       0.108
78          4.43      165.40       0.102
79          7.44      236.36       0.092    Labor
80         15.70      316.23       0.091
81         55.45      306.95       0.090
82          1.94      174.49       0.088
83         12.40      205.66       0.083
84         23.63      275.46       0.069
85         25.56      518.71       0.068    International
86         35.79      336.31       0.067
87         21.43      263.43       0.065
88         26.31      439.75       0.062
89         22.58      246.53       0.055
90        885.94    3,556.91       0.054
91         24.27      392.70       0.052
92         47.98      590.95       0.044
93         13.45      191.90       0.042
94          2.78       68.59       0.033
95          4.50      185.40       0.028
96         68.89      990.77       0.028
97          2.13      705.86       0.026
98        124.96      586.02       0.022
99          3.20       74.38       0.019
100         5.52      130.31       0.013
101         9.78      171.55       0.009    International
102         5.59      106.45       0.001
        6,222.29   53,080.91         100

TABLE A2
Influence and Division by Field, 1990-2010

Macroeconomics/Monetary Economics

Rank             Institution              Influence   Cluster

1      Federal Reserve                      6.748
2      Chicago                              5.404        F
3      Harvard                              5.001        S
4      Princeton                            4.409        S
5      MIT                                  4.280        S
6      Northwestern                         3.215        F
7      Stanford                             3.184        S
8      Columbia                             3.165        S
9      Rochester                            2.898        F
10     Pennsylvania                         2.895        F
11     Carnegie Mellon                      2.298        F
12     NYU                                  2.249        F
13     Berkeley                             1.961        S
14     Yale                                 1.797        S
15     Minnesota                            1.705        F
16     UCLA                                 1.221        F
17     Michigan                             1.206        S
18     UCSD                                 1.200
19     IMF                                  1.104
20     Virginia                             1.035

Significance of division: p = .000

Microeconomic Theory

Rank             Institution              Influence   Cluster

1      Northwestern                        5.6225        F
2      Stanford                            4.9160        S
3      Harvard                             4.5609        S
4      MIT                                 3.0671        S
5      Chicago                             2.6729        F
6      Pennsylvania                        2.6345        F
7      Berkeley                            2.6283        S
8      Hebrew                              2.3405        S
9      Princeton                           2.2235        S
10     Yale                                1.9789        F
11     Rochester                           1.6796        F
12     Caltech                             1.6352        F
13     Columbia                            1.2958        F
14     UCSD                                1.2892        S
15     Minnesota                           1.2812        F
16     NYU                                 1.2608        S
17     Carnegie Mellon                     1.2550
18     UCLA                                1.2513
19     Tel Aviv                            1.2202
20     LSE                                 1.1743

Significance of division: p = .025

Industrial Organization

Rank             Institution              Influence   Cluster

1      Stanford                            5.5790        F
2      Harvard                             5.5658        S
3      MIT                                 5.0191        S
4      Berkeley                            3.8157        S
5      Northwestern                        3.8119        F
6      Chicago                             3.4647        S
7      Princeton                           3.0156        F
8      Yale                                2.7202        S
9      Pennsylvania                        1.7994        S
10     Michigan                            1.6604        F
11     LSE                                 1.6439        S
12     UCLA                                1.4305        S
13     Columbia                            1.3893        F
14     Oxford                              1.3638        F
15     NYU                                 1.3060        F
16     Wisconsin                           1.1362        F
17     UBC                                 1.0537
18     Carnegie Mellon                     1.0098
19     Boston                              0.8990
20     Toulouse                            0.8909

Significance of division: p = .156

Econometrics

Rank             Institution              Influence   Cluster

1      Harvard                             4.0812        S
2      Yale                                4.0494        F
3      Chicago                             3.2981        F
4      Stanford                            3.0671        S
5      Wisconsin                           2.7577        S
6      UCSD                                2.5052        F
7      Berkeley                            2.4271        S
8      MIT                                 2.4249        F
9      Princeton                           2.0573        F
10     Minnesota                           1.8980        S
11     LSE                                 1.7210        F
12     Australian National University      1.4065        F
13     UCLA                                1.3731        S
14     Northwestern                        1.2735        F
15     Carnegie Mellon                     1.2575        S
16     Washington                          1.1755        S
17     Rochester                           1.1570
18     North Carolina State University     1.1018
19     Pennsylvania                        1.0981
20     Federal Reserve                     1.0876

Significance of division: p = .000

Labor Economies

Rank             Institution              Influence   Cluster

1      Chicago                             6.0248        F
2      Harvard                             5.7762        S
3      MIT                                 3.9138        S
4      Princeton                           3.7586        S
5      Michigan                            2.8488        F
6      Cornell                             2.6335        S
7      Northwestern                        2.2136        F
8      Stanford                            2.1601        F
9      Columbia                            2.0402        S
10     Berkeley                            1.9212        S
11     Wisconsin                           1.7542        F
12     Pennsylvania                        1.6850        S
13     Illinois                            1.3616        F
14     Yale                                1.3361        F
15     UCLA                                1.3359        F
16     Michigan State                      1.1199        S
17     Rand                                1.0766
18     LSE                                 1.0264
19     Federal Reserve                     0.9765
20     Rochester                           0.9166

Significance of division: p = .112

Growth and Development

Rank             Institution              Influence   Cluster

1      World Bank                          6.6918
2      Harvard                             5.2555        S
3      MIT                                 3.2398        S
4      Chicago                             2.8572        F
5      Princeton                           2.8380        F
6      Stanford                            2.2975        F
7      Berkeley                            2.1822        F
8      Pennsylvania                        2.1374        F
9      Yale                                1.8715        F
10     Oxford                              1.7516        S
11     IMF                                 1.6168
12     Columbia                            1.6053        F
13     UCLA                                1.4137        S
14     LSE                                 1.2300        S
15     Sussex                              1.1414        S
16     NYU                                 1.0910        S
17     Cornell                             1.0385        S
18     Michigan                            1.0201        F
19     Maryland                            1.0175
20     Northwestern                        0.9900

Significance of division: p = .058

Finance

Rank             Institution              Influence   Cluster

1      Chicago                             7.8673        S
2      Harvard                             5.1636        S
3      MIT                                 4.1508        S
4      Pennsylvania                        3.7895        F
5      NYU                                 3.6090        F
6      Stanford                            3.4444        S
7      Rochester                           3.1435        F
8      UCLA                                3.0611        F
9      Northwestern                        2.8458        S
10     Princeton                           2.2709        S
11     Columbia                            2.0991        F
12     Michigan                            1.9926        F
13     Berkeley                            1.9176        S
14     Yale                                1.6008        S
15     Federal Reserve                     1.5739
16     Cornell                             1.4666        F
17     Duke                                1.4440        F
18     Illinois                            1.4170
19     use                                 1.3899
20     Ohio State                          1.2346

Significance of division: p = .191

Public Economies

Rank             Institution              Influence   Cluster

1      Harvard                             6.1096        S
2      Stanford                            3.6090        F
3      Chicago                             3.1788        S
4      Princeton                           3.0292        F
5      MIT                                 2.9096        S
6      Michigan                            2.6438        S
7      Northwestern                        1.8147        F
8      Rochester                           1.7341        F
9      Yale                                1.7076        F
10     Berkeley                            1.6974        S
11     Carnegie Mellon                     1.6327        F
12     Pennsylvania                        1.6069        S
13     Wisconsin                           1.5837        S
14     UCLA                                1.5698        F
15     UCSD                                1.5403        F
16     LSE                                 1.3275        S
17     Maryland                            1.2446
18     Columbia                            1.1431
19     Federal Reserve                     1.1407
20     Caltech                             1.1369

Significance of division: p = .027

International Economies

Rank             Institution              Influence   Cluster

1      Harvard                             4.7528        S
2      MIT                                 4.1995        S
3      Columbia                            3.6099        S
4      Princeton                           3.5515        S
5      Chicago                             3.2578        F
6      Berkeley                            3.1833        F
7      Federal Reserve                     3.1302
8      IMF                                 2.9166
9      Stanford                            2.4534        F
10     World Bank                          2.2068
11     Pennsylvania                        2.0556        F
12     Northwestern                        1.8857        S
13     Yale                                1.8760        S
14     NYU                                 1.6134        F
15     Michigan                            1.5766        S
16     UCLA                                1.4886        S
17     UCSD                                1.4779        F
18     Rochester                           1.3049        F
19     UBC                                 1.0187        F
20     Tel Aviv                            1.0149

Significance of division: p = .184

Note: Cluster column denotes members of the strongest division
between freshwater (F) and saltwater (S) clusters for 16 most
influential academic departments by field.

REFERENCES

Amir, R., and M. Knauff. "Ranking Economics Departments Worldwide on the Basis of PhD Placement." Review of Economics and Statistics, 90(1), 2008, 185-90.

Combes, P.-P., and L. Linnemer: "Inferring Missing Citations: A Quantitative Multi-Criteria Ranking of All Journals in Economics." GREQAM Working Paper Series No. 28, 2010.

Davis, R, and G. F. Papanek. "Faculty Ratings of Major Economics Departments by Citations." American Economic Review, 74(1), 1984, 225-30.

Girvan, M., and M. E. J. Newman. "Community Structure in Social and Biological Networks." Proceedings of the National Academy of Sciences of the United States of America, 99(12), 2002, 7821-26.

Liebowitz, S. J., and J. P. Palmer. "Assessing the Relative Impacts of Economics Journals." Journal of Economic Literature, 22(1), 1984, 77-88.

Mankiw, N. G. "The Macroeconomist as Scientist and Engineer." Journal of Economic Perspectives, 20(4), 2006, 29-46.

Newman, M. E. J. "Analysis of Weighted Networks." Physical Review E, 70, 2004, 056131.

Pinski, G., and F. Narin. "Citation Influence for Journal Aggregates of Scientific Publications: Theory, with Application to the Literature of Physics." Information Processing and Management, 12(5), 1976, 297-312.

Rosvall, M., and C. T. Bergstrom. "Multilevel Compression of Random Walks on Networks Reveals Hierarchical Organization in Large Integrated Systems." PLoS One, 6(4), 2011, e18209.

Tervio, M. "Divisions within Academia: Evidence from Hiring and Placement." Review of Economics and Statistics, 93(3), 2011, 1053-62.

ALI SINA ONDER and MARKO TERVIO *

* We thank Carl Bergstrom, Ted Bergstrom, John Conley, Mario Crucini, Tuomas Pekkarinen, Martin Rosvall, Laurent Simula, Yves Zenou, and three anonymous referees for helpful comments.

Onder: Department of Economics, University of Bayreuth, 95447 Bayreuth, Germany. Phone 0049 921 556085, Fax 0049 921 556081, E-mail ali-sina.oender@ uni-bayreuth.de

Tervio: Department of Economics, School of Business, Aalto University, 00076 Aalto, Finland. Phone 358 40 353 8342, E-mail marko.tervio@aalto.fi

(1.) For the list of journals and their summary statistics, see Table A1 in the Appendix.

(2.) For cited articles with multiple authors, only the affiliation of the contact author is available.

(3.) It would be ideal to also divide citations for multiauthor documents proportionally between the authors, but observing only on the contact author affiliation precludes this.

(4.) Note that we cannot distinguish between authors citing themselves and authors citing their peers at the same institution, because we only have data on contact author affiliation.

(5.) Newman (2004) shows that this method, although originally defined for binary networks, is also suitable for weighted networks.

(6.) We will consider the possibility of an arbitrary number of unevenly sized clusters when we apply two alternative clustering methods in the next section.

(7.) There could, in principle, be several maximizers, but this never occurs in our data.

(8.) Davis and Papanek (1984) provide an early study of department rankings based on citation counts. For rankings of academic journals using network influence, see Liebowitz and Palmer (1984) and Eigenfactor.org. Amir and Knauff (2008) and Tervid (2011) apply this method to data on PhD placement/faculty hiring data.

(9.) In Tervio (2011), the "top" was defined by PhD placement instead of citations, but using the exact same set of top 16 U.S. departments as there results in exactly the same clusters here.

(10.) Table A2 in the Appendix provides more detail on the influence and cluster membership of the top departments in each field.

TABLE 1
Unit Citations from Row to Column Department for the Top 20
Academic Departments, 1990-2010

                                                        Carnegie
                  Minnesota   Rochester   Penn   NYU     Mellon

Minnesota           44.4         5.3      8.9    5.7      5.3
Rochester            4.6        44.2      7.9    5.4      5.6
Penn                 9.6        13.4      90.2   12.7     9.4
NYU                  7.9        15.2      18.5   72.7     8.1
Carnegie Mellon      2.6         3.6      5.8    4.1      26.2
Northwestern         8.3        15.3      20.7   10.5     9.7
UCLA                 6.9         8.7      13.6   9.1      5.8
Cornell              6.4         6.4      9.4    7.0      4.8
Wisconsin            5.3         7.2      13.0   7.4      6.7
Chicago              6.2        12.7      18.2   11.1     9.3
Michigan             4.7        11.0      14.4   7.7      3.7
UCSD                 2.3         6.4      6.3    3.3      3.3
Yale                 5.6         5.6      11.5   5.7      4.3
Stanford             7.1        12.2      16.8   9.1      10.2
Columbia             4.8        10.0      13.9   9.3      6.7
LSE                  3.9         6.8      9.6    7.7      4.2
Harvard              8.2        17.2      24.4   15.5     9.2
Princeton            4.4         8.8      9.2    6.1      5.4
MIT                  5.4        10.6      17.8   7.4      5.0
UC Berkeley          6.4         7.0      16.0   9.7      6.5

                  Northwestern   UCLA   Cornell   Wisconsin   Chicago

Minnesota             9.9        6.4      4.3        5.6       15.1
Rochester             11.7       4.8      2.5        2.7       16.7
Penn                  22.2       10.4     6.7       10.5       40.5
NYU                   20.6       12.3     6.7        8.4       34.8
Carnegie Mellon       5.9        3.8      4.1        2.5       12.2
Northwestern         105.1       11.3     7.0       10.4       36.5
UCLA                  15.2       64.4     5.0        6.7       28.1
Cornell               11.0       7.8     64.9        8.2       18.8
Wisconsin             14.3       8.3      6.0       67.8       21.0
Chicago               24.3       14.0     7.9        9.8       135.8
Michigan              12.8       8.6      5.6        9.0       25.5
UCSD                  8.5        4.3      2.0        4.8       11.4
Yale                  12.9       7.5      3.5        6.2       18.0
Stanford              24.9       12.9     6.3        7.1       32.0
Columbia              14.8       8.2      4.5        7.0       29.2
LSE                   10.7       4.6      3.9        6.3       18.0
Harvard               25.5       17.2     8.8       10.2       55.2
Princeton             15.3       5.3      5.7        7.1       22.9
MIT                   16.2       9.7      6.3        7.5       35.8
UC Berkeley           17.7       9.8      5.9        7.5       27.1

                  Michigan   UCSD   Yale   Stanford   Columbia   LSE

Minnesota           3.4      2.8    6.0      11.2       4.9      2.8
Rochester           6.1      4.1    8.3      11.0       3.6      3.0
Penn                9.8      4.8    15.0     23.3       14.1     5.9
NYU                 8.1      7.4    11.5     22.3       16.0     7.3
Carnegie Mellon     3.7      1.7    5.9      11.3       3.9      1.5
Northwestern        9.6      8.3    14.4     29.8       10.6     7.1
UCLA                7.4      5.6    10.1     21.9       9.5      3.5
Cornell             8.5      5.8    10.6     14.9       6.4      5.8
Wisconsin           9.7      7.3    9.9      16.7       7.7      5.5
Chicago             11.7     8.9    18.5     26.1       10.7     5.8
Michigan            64.5     6.6    10.5     18.6       8.6      4.5
UCSD                4.9      40.6   8.2      9.9        3.3      3.0
Yale                6.5      5.2    80.0     17.4       8.4      7.0
Stanford            11.3     8.2    17.4    123.3       13.8     5.8
Columbia            8.9      5.8    13.5     24.1       67.1     5.0
LSE                 4.9      7.1    11.4     15.0       6.3      65.4
Harvard             18.3     9.9    24.0     38.4       19.8     10.9
Princeton           6.6      6.6    13.6     21.8       9.9      7.4
MIT                 11.1     8.4    16.3     25.9       12.1     8.4
UC Berkeley         10.8     7.1    18.1     34.7       11.7     5.3

                  Harvard   Princeton   MIT   UC Berkeley

Minnesota          14.0        6.9     11.4     7.1
Rochester          12.8        6.9      8.7     4.6
Penn               36.0       19.9     26.2     13.7
NYU                32.5       19.5     24.0     13.4
Carnegie Mellon    10.0        5.1      7.3     4.2
Northwestern       32.0       20.7     29.5     15.2
UCLA               27.6       13.5     19.3     13.2
Cornell            18.8       13.3     15.4     11.6
Wisconsin          21.4       15.2     18.0     17.3
Chicago            46.8       22.7     35.0     15.3
Michigan           34.0       16.4     11.8     11:4
UCSD               13.9       10.1      9.9     7.5
Yale               22.6       13.6     16.8     10.9
Stanford           41.8       19.9     31.6     24.5
Columbia           36.3       22.0     24.5     13.3
LSE                24.3       16.8     23.5     10.8
Harvard            224.6      35.3     59.6     30.0
Princeton          30.3       69.0     27.0     11.5
MIT                50.5       27.6     127.8    18.5
UC Berkeley        45.1       24.0     36.4    120.1

Notes: Shading depicts deviations from expected citations patterns
in the absence of clustering (excluding self-citations). Solid
cells depict citations above and striped cells below expected
intensity. Darker shades depict stronger deviations. Institutions
are ordered by the strength of their connection to the saltwater
cluster. Table available in color in the online version of this
article.

TABLE 2
Summary Statistics and Main Results for Top 50 Institutions, 1990-2010

                                                          Cites to
Institution          Cites In    Cites Out   Self-Cites    Unknown

1 Harvard            2,482.93      888.20       224.62      601.18
2 Chicago            2,042.52      582.97       135.78      349.25
3 MIT                1,941.23      570.79       127.81      360.39
4 Stanford           1,652.42      609.57       123.34      414.09
5 Princeton          1.512.75      434.00        68.98      230.03
6 Northwestern       1,303.40      570.31       105.09      259.60
7 Berkeley           1,248.10      662.44       120.08      454.49
8 Pennsylvania       1,126.74      588.49        90.16      295.35
9 Yale               1.072.01      393.06        80.00      251.94
10 Federal Reserve   1,093.26    1,053.16       234.83      393.01
11 Columbia            867.85      492.04        67.13      269.83
12 Rochester           852.98      268.14        44.17      126.69
13 Michigan            805.30      481.96        64.46      277.58
14 NYU                 821.24      566.06        72.70      232.24
15 UCLA                730.26      426.52        64.43      256.05
16 Wisconsin           732.28      522.95        67.82      318.22
17 LSE                 753.41      450.93        65.38      240.68
18 UCSD                694.68      256.96        40.56      127.47
19 Carnegie Mellon     567.71      200.28        26.16      123.56
20 Minnesota           545.43      303.93        44.38      216.68
21 Cornell             562.31      434.20        64.89      310.91
22 World Bank          545.06      469.50       120.39      346.12
23 Illinois            489.97      464.64        63.36      308.01
24 Duke                417.44      385.02        47.53      212.45
25 Maryland            477.03      417.22        57.42      252.35
26 UBC                 496.13      370.38        48.97      203.64
27 Hebrew              395.34      201.61        43.34      144.05
28 Oxford              437.86      329.36        46.90      181.75
29 Tel Aviv            365.07      215.30        31.53      103.17
30 Boston U            322.11      239.46        24.87      113.67
31 Toronto             338.08      345.74        35.95      175.31
32 UC Davis            335.19      319.12        46.27      218.62
33 Ohio State          332.69      339.89        35.45      186.66
34 Texas-Austin        339.10      382.01        37.85      224.14
35 USC                 294.50      254.58        25.37      139.05
36 Washington          304.01      230.31        24.12      129.58
37 Virginia            298.93      195.71        19.38      108.90
38 Penn State          300.33      304.04        30.63      170.33
39 IMF                 304.92      353.73        45.87      143.39
40 Michigan State      301.68      294.24        35.04      175.72
41 Caltech             238.43      129.08        22.22       71.71
42 Indiana             280.42      244.60        23.71      135.69
43 Iowa                245.24      178.83        16.01       89.16
44 ANU                 267.16      205.56        25.26      133.18
45 UNC                 236.50      291.13        31.72      180.15
46 Brown               226.44      205.73        23.66       99.61
47 Florida             242.49      196.30        21.05      117.65
48 UCL                 234.70      213.26        22.93       82.82
49 Arizona             239.10      217.00        34.02      155.98
50 Cambridge           246.18      190.01        28.51      113.48
Others (1,142       21,747.74   35,768.34    3,128.56    19,970.10
  institutions)
All                 54,708.65   54,708.65    6,130.65    30,795.70

                     Unique                Relative
Institution          Authors   Influence     Salt

1 Harvard               583       5.126      0.651
2 Chicago               368       4.292     -0.221
3 MIT                   295       4.005      1.042
4 Stanford              441       3.516      0.126
5 Princeton             224        3030      0.851
6 Northwestern          321       2.752     -1.147
7 Berkeley              480       2.501      1.352
8 Pennsylvania          343       2.340     -1.555
9 Yale                  277       2.225      0.059
10 Federal Reserve      677       1.965     -1.508
11 Columbia             338       1.729      0.196
12 Rochester            169       1.703     -1.982
13 Michigan             366       1.613     -0.538
14 NYU                  293       1.547     -1.419
15 UCLA                 284       1.527     -0.986
16 Wisconsin            352       1.393     -0.481
17 LSE                  305       1.283      0.625
18 UCSD                 135       1.246     -0.239
19 Carnegie Mellon      165       1.150     -1.347
20 Minnesota            249       1.084     -2.107
21 Cornell              337       1.059     -0.586
22 World Bank           407       0.963      0.932
23 Illinois             326       0.940     -1.048
24 Duke                 263       0.854     -1.271
25 Maryland             251       0.832     -0.020
26 UBC                  217       0.826      0.414
27 Hebrew               135       0.782     -0.084
28 Oxford               268       0.731      1.548
29 Tel Aviv             108       0.705     -0.775
30 Boston U             149       0.642      0.147
31 Toronto              223       0.637     -0.619
32 UC Davis             214       0.609     -0.039
33 Ohio State           230       0.603     -1.035
34 Texas-Austin         276       0.576     -1.168
35 USC                  164       0.571     -1.175
36 Washington           174       0.562     -0.789
37 Virginia             144       0.543     -0.530
38 Penn State           209       0.542     -1.565
39 IMF                  291       0.512      0.760
40 Michigan State       201       0.508      0.009
41 Caltech               73       0.501     -1.439
42 Indiana              173       0.480     -1.157
43 Iowa                 129       0.478     -2.399
44 ANU                  151       0.442      0.125
45 UNC                  245       0.436     -1.195
46 Brown                 91       0.428     -0.626
47 Florida              155       0.424     -1.417
48 UCL                  114       0.421      0.683
49 Arizona              192       0.412     -0.977
50 Cambridge            173       0.406      1.265
Others (1,142        29,934      35.546      0.020
  institutions)
All                  42,682         100      0.000

                            Strongest Division for

Institution          Top 24   Top 20   Top 16   Top 12

1 Harvard              S        S        S        S
2 Chicago              F        F        F        F
3 MIT                  S        S        S        S
4 Stanford             S        S        S        S
5 Princeton            S        S        S        S
6 Northwestern         F        F        F        F
7 Berkeley             S        S        S        S
8 Pennsylvania         F        F        F        F
9 Yale                 S        S        S        F
10 Federal Reserve
11 Columbia            S        S        S        S
12 Rochester           F        F        F        F
13 Michigan            S        S        F        F
14 NYU                 F        F        F
15 UCLA                F        F        F
16 Wisconsin           F        F        F
17 LSE                 S        S        S
18 UCSD                S        S
19 Carnegie Mellon     F        F
20 Minnesota           F        F
21 Cornell             F        F
22 World Bank
23 Illinois            F
24 Duke                F
25 Maryland            S
26 UBC                 S
27 Hebrew
28 Oxford
29 Tel Aviv
30 Boston U
31 Toronto
32 UC Davis
33 Ohio State
34 Texas-Austin
35 USC
36 Washington
37 Virginia
38 Penn State
39 IMF
40 Michigan State
41 Caltech
42 Indiana
43 Iowa
44 ANU
45 UNC
46 Brown
47 Florida
48 UCL
49 Arizona
50 Cambridge
Others (1,142
  institutions)
All

Notes: Articles from all sample journals from 1990 to 2010.
Nonacademic institutions in italics.

Influence's in the network of citations is calculated after
dropping self-citations by institutions from the data.

"Relative salt" measures the propensity to cite members of
saltwater cluster relative to freshwater cluster (with clusters
defined for top 20).

TABLE 3
Division by Field, 1990-2010

                           Modified Q              Within-Cluster Bias

Field              p Value   Actual   [P.sub.95]   Actual   [P.sub.95]

Macro/monetary       .000    1.350        0.877     32.6         11.5
Micro theory         .025    0.870        0.837     13.8         10.8
Industrial           .156    0.716        0.769     11.2         12.8
  organization
Econometrics         .000    1.480        1.032     40.8         14.0
Labor                .112    0.774        0.811     17.3          7.8
Growth/              .058    1.139        1.149     24.0         13.7
  development
Finance              .191    0.455        0.499      2.8         10.2
Public               .027    1.063        1.026     31.7         15.9
International        .184    1.161        1.263     18.8         13.1

All 102 journals     .000    0.642        0.400     16.1          8.5

Field              Citations   Articles

Macro/monetary        82,995      3,764
Micro theory          90,430      5,455
Industrial            51,305      2,608
  organization
Econometrics         104,810      5,527
Labor                 44,372      2,201
Growth/               61,463      3,376
  development
Finance              104,398      3,801
Public                76,361      4,365
International         62,415      3,075

All 102 journals   1,662,212     91,635

TABLE 4
Cluster Membership of Top Departments by Field, 1990-2010

                   Macro/                    Micro
                   Money     Econometrics    Theory     Public

Harvard              S            S            S          S
Chicago              F            F            F          S
MIT                  S            F            S          S
Stanford             S            S            S          F
Princeton            S            F            S          F
Northwestern         F            F            F          F
Berkeley             S            S            S          S
Pennsylvania         F                         F          S
Yale                 S            F            F          F
Columbia             S                         F
Rochester            F                         F          F
Michigan             S                                    S
NYU                  F                         S
UCLA                 F            S                       F
Wisconsin                         S                       S
LSE                               F                       S
UCSD                              F            S          F
Carnegie Mellon      F            S                       F
Minnesota            F            S            F
Cornell
                  p = .000     p = .000     p = .025   p = .027

                    Growth/               Industrial
                  Development    Labor    Organization   International

Harvard                S           S           S               S
Chicago                F           F           S               F
MIT                    S           S           S               S
Stanford               F           F           F               F
Princeton              F           S           F               S
Northwestern                       F           F               S
Berkeley               F           S           S               F
Pennsylvania           F           S           S               F
Yale                   F           F           S               S
Columbia               F           S           F               S
Rochester                                                      F
Michigan               F           F           F               S
NYU                    S                       F               F
UCLA                   S           F           S               S
Wisconsin                          F           F
LSE                    S                       S
UCSD                                                           F
Carnegie Mellon
Minnesota
Cornell                S           S
                   p = .058     p= .112     p = .156       p = .184

                                          All
                                        without
                  Finance    General     Macro     Overall

Harvard              S          S          S          S
Chicago              S          S          F          F
MIT                  S          S          S          S
Stanford             S          S          S          S
Princeton            S          S          S          S
Northwestern         S          F          F          F
Berkeley             S          S          S          S
Pennsylvania         F          F          F          F
Yale                 S          F          F          S
Columbia             F          S          S          S
Rochester            F          F          F          F
Michigan             F          F          S          F
NYU                  F          F          F          F
UCLA                 F          S          S          F
Wisconsin                       F          F          F
LSE                                                   S
UCSD                                       F          S
Carnegie Mellon                 F                     F
Minnesota                                             F
Cornell              F                                F
                  p = .191   p = .003   p = .003   p = .000

Notes: We list top 20 departments from the overall ranking and find
the division among top 16 departments in each field (see Table A2
for field rankings). S is defined as the cluster with Harvard in
it.

Each field is made of citations going out from top four field
journals (see Table A1 for journal rankings).

There are departments that are ranked top 16 in some fields,
although they are not placed in top 20 in the overall ranking.
These are left out of this table. Cluster membership is in bold if
it is the same as in the full data for top 20 departments.

p Values (last row) give the statistical significance of the
division.

General: citations going out from top five general interest
journals (Econometrica, AER, JPE, QJE, REStud).

All without macro: top four journals from all fields put together
excluding macro/money journals. This group consists of 32 journals
in total.