A Social Network of Russian "Kompromat"
NNoname manuscript No. (will be inserted by the editor)
A Social Network of Russian “Kompromat”
Dmitry Zinoviev
Received: date / Accepted: date
Abstract “Kompromat” (the Russian word for “compromising material”) has beenefficiently used to harass Russian political and business elites since the days of theUSSR. Online crowdsourcing projects such as “RuCompromat” made it possible tocatalog and analyze kompromat using quantitative techniques—namely, social net-work analysis. In this paper, we constructed a social network of 11,000 Russian andforeign nationals affected by kompromat in Russia in 1991–2020. The network hasan excellent modular structure with 62 dense communities. One community con-tains prominent American officials, politicians, and entrepreneurs (including Presi-dent Donald Trump) and appears to concern Russia’s controversial interference inthe 2016 U.S. presidential elections. Various network centrality measures identifyseventeen most central kompromat figures, with President Vladimir Putin solidly atthe top. We further reveal four types of communities dominated by entrepreneurs,politicians, bankers, and law enforcement officials (“siloviks”), the latter disjointedfrom the first three.
Keywords kompromat · Russia · politics · social network analysis “Kompromat” is a Russian word for “compromising material.” Kompromat has beenefficiently used to harass political and business elites. It is widely considered to beessential in shoring up authoritarian durability [7]. Kompromat regimes often appearin states with low fiscal capacity and very high police capacity and harbor widespreadcriminality combined with systematic blackmail [2]. D. ZinovievSuffolk University, Boston MA 02114, USA,Tel.: +1-617-305-1985ORCID: 0000-0002-8008-0893E-mail: [email protected] a r X i v : . [ c s . S I] S e p Dmitry Zinoviev
The practice of kompromat is nothing new: it dates back to the Soviet period(Stalin, Beria, and the KGB) [6]. However, social media advances provided mas-sive and affordable opportunities for authoritarian regimes to use kompromat againstthe opposition and competing factions [9]. Simultaneously, online crowdsourcingprojects such as “RuCompromat” [11] make it possible to catalog and systematizekompromat, make it broadly available to researchers, and potentially disarm it byexposing its sources.In this paper, we explore the most extensive online collection of the post-Soviet(mostly Russian) kompromat hosted by the “RuCompromat” team from the perspec-tive of social network analysis—the process of modeling a social environment as apattern of regularities in relationships among interacting units [16]. In our case, theinteracting units are the persons subject to kompromat or related to kompromat in anyother way (e.g., reporters and politicians [5]). Their co-involvement in compromisingscenarios defines relationships among them. We identify 33 dense groups (networkcommunities) of actors: politicians, entrepreneurs, and law enforcement officials, inthe first place—associated with specific kompromat cases. Eventually, we classify thecases in four basic types that differ in the principal actors’ participation.The rest of the paper is organized as follows: In Section 2, we describe the dataset, its provenance, and its structure; in Section 3, we explain the network construc-tion and analyze it; in Section 4, we present the results and discuss them. Finally, inSection 5, we conclude and lay the ground for further work.
We collected the dataset in September 2020 from the site “RuCompromat” [11]. Thesite provides information about approximately 11 thousand persons, both Russianand foreign nationals. For each person, the cite provides references to media articles(only for the years 2013–2020) that mention potentially compromising facts aboutthe person, a list of other people mentioned together with the person, and a list oforganizations (e.g., companies, banks, and government offices) related to the person.As an example, an article called “Patrushev Jr. and the Orthodox retirees. Un-profitable Rosselkhozbank goes to the aid of “Peresvet,” published by the “Ruspress”agency [12], mentions banker Valery Meshalkin and the son of the former FSB direc-tor Nikolai Patryshev. Thus, “RuCompromat” lists the two as related.The dataset contains co-references for 11,118 persons for the time frame from1999 to 2020. We have not validated the references’ accuracy and take the “RuCom-promat” contributors’ opinions as the ground truth.
The first step in social network analysis is the construction of a social network. Thenetwork consists of 11,118 nodes representing persons, and 37,544 edges represent-ing relationships between the persons. The density of the network is ≈ . Social Network of Russian “Kompromat” 3
Fig. 1
The dense core of the “kompromat” network, with Vladimir Putin and Dmitry Medvedev at thecenter, Igor Sechin and Sergey Chemezov below, Sergey Sobyanin on the right, and Aleksandr Bastrykinabove.
A is connected to person B, then B is also connected to A), and strongly connected(it is possible to get from any person A to any person B by following edges withoutleaving the network). The diameter of the network is 12. Fig. 1 shows the central partof the network—its dense core.Fig. 2 shows the node degree distribution in the network. The distribution has a“long tail” and can be approximated by the power law with the exponent α ≈ − . cf. Fig. 1 and Table 2.We used the Louvain community detection algorithm [1] to partition the networkinto 62 network communities: groups of persons that are more closely connected to
Dmitry Zinoviev Degree10 F r e q u e n c y Fig. 2
The node degree distribution has a “long tail” and can be approximated by the power law f d ∼ d − . for d ≥ each other than to the persons from the rest of the network. Each of the communitiesis denser than the whole network (Table 1). The sizes of the communities range from3 persons to 1,470 persons. The partition modularity [8] is 0.64 on the scale from -0.5(no community structure) to 1.0 (perfect community structure).As it is customary in social network analysis, we excluded the smallest 29 com-munities with fewer than 100 persons from the study. For each node in the remainingcommunities, we calculated the clustering coefficient and several centralities: degreecentrality, closeness centrality, betweenness centrality, and eigenvector centrality [3].The degree centrality, or simply the degree of a node, is the number of the node’sconnections. A person with a high degree centrality has been affected by more kom-promat cases.To understand closeness centrality, let us consider two randomly chosen persons:Vladimir Putin and Zhou Yongkang (a late senior leader of the Communist Party ofChina, CPC). Due to the network’s sparse nature, it is improbable that they shareda direct connecting edge. Putin is more likely connected to another person: say, XiJinping (the General Secretary of the CPC)—who, consequently, was connected toZhou. We say that Putin is two hops apart from Zhou.In a different scenario, Putin is connected to Genry Reznik (a prominent Rus-sian lawyer), who is connected to Yury Antonov (a retired activist from St. Peters-burg), who is connected to Inna Pashchenko (a WWII veteran; all three are actual Social Network of Russian “Kompromat” 5 members of the “RuCompromat” dataset). Therefore, Putin is three hops apart fromPashchenko.There is a path from Putin to the other person in both scenarios, and the length ofthe path is two and three, respectively. In general, the average length of the shortestpaths from a person to all other people in the network is called the closeness cen-trality of that person. A person with a high closeness centrality has been more likelyindirectly affected by or involved in more kompromat-related cases.Quite expectedly [15], the degree centrality is positively correlated with closenesscentrality (their correlation is r ≈ . Dmitry Zinoviev
Table 1
Mean community parameters, sorted in the decreasing order of the Betweenness centrality: Size(number of persons), Closeness centrality, Eigenvector centrality, Clustering Coefficient, and Density. Themean betweenness centrality of the top six communities is above the network average.Community label B S C E CC D1 Putin Vladimir 0.54 1,470 0.24 7.80 0.53 0.01242 Bastrykin Aleksandr 0.39 749 0.23 3.44 0.61 0.00653 Sobyanin Sergey 0.38 940 0.23 4.00 0.56 0.02084 Sechin Igor 0.35 199 0.23 2.93 0.58 0.01245 Rotenberg Arkady 0.35 382 0.23 3.99 0.57 0.00356 Kostin Andrey 0.33 233 0.23 3.30 0.64 0.00757 Chemezov Sergey 0.31 322 0.23 3.83 0.59 0.02068 Nabiullina Elvira 0.30 390 0.22 2.70 0.58 0.01449 Shuvalov Igor 0.29 176 0.22 1.97 0.62 0.015210 Luzhkov Yury 0.29 298 0.22 2.12 0.59 0.030911 Rogozin Dmitry 0.28 293 0.22 3.11 0.59 0.022812 Poltavchenko Georgy 0.28 230 0.21 2.43 0.61 0.011813 Gref German 0.27 353 0.22 2.06 0.61 0.012314 Kerimov Suleyman 0.27 225 0.23 2.65 0.63 0.034115 Chayka Yury 0.27 338 0.22 2.75 0.60 0.045116 Mutko Vitaly 0.27 121 0.22 2.09 0.63 0.012917 Abramovich Roman 0.26 465 0.23 3.21 0.61 0.024218 Ismailov Telman 0.25 105 0.21 2.34 0.63 0.018519 Tolokonsky Viktor 0.25 264 0.21 2.08 0.65 0.010820 Poroshenko Petr 0.25 385 0.23 2.74 0.63 0.363621 Trump Donald 0.24 289 0.22 2.87 0.59 0.013022 Kadyrov Ramzan 0.24 202 0.23 3.95 0.64 0.016323 Ivanov Sergey 0.24 230 0.22 2.98 0.62 0.043424 Vekselberg Viktor 0.23 331 0.22 1.92 0.65 0.015425 Khodorkovsky Mikhail 0.23 142 0.22 3.91 0.65 0.013926 Avetisyan Artem 0.23 100 0.22 2.58 0.62 0.015127 Miller Aleksey 0.23 107 0.21 2.06 0.64 0.036528 Yakunin Vladimir 0.22 340 0.23 4.59 0.59 0.017429 Chayka Igor 0.21 307 0.23 2.92 0.61 0.012330 Mordashev Aleksey 0.21 156 0.22 1.58 0.64 0.019431 Mikhaylov Sergey 0.20 182 0.22 2.18 0.61 0.031632 Prigozhin Evgeny 0.20 152 0.22 1.67 0.65 0.023533 Radaev Valery 0.19 130 0.22 2.01 0.66 0.0214Overall 0.32 n/a 0.23 3.56 0.60 0.0006 bank; 2), Andrey Kostin (President of VTB Bank; 2), Arkady Rotenberg (2), SergeySobyanin (2), and Dmitry Zakharchenko (a former anti-corruption official convictedof bribery; 2). They represent almost all principal groups of kompromat stakeholders:business, politics, law enforcement, banking, government, and press—with a notableexception of the criminal underworld.
In the rest of the paper, we investigate the composition of the communities, theirrelationships, and kompromat cases’ typology.
Social Network of Russian “Kompromat” 7 Eigenvector centrality5.4 × 10 C l u s t e r i n g c o e ff i c i e n t Fig. 3
Mean eigenvector centrality and clustering coefficient for the major 33 communities (see Table 1).The size of a dot represents the number of persons in the community.
Table 2
Top ten persons, sorted in the decreasing order of the betweenness centrality, closeness centrality,eigenvector centrality, and degree. The dagger † marks the persons who appear in more than one column.Betweenness Closeness Eigenvector Degree(Typicality) (Indirect involvement) (Toxicity) (Direct involvement)Putin † V. Putin V. † Putin V. † Putin V. † Sechin I. † Medvedev D. † Medvedev D. † Sechin I. † Bastrykin A. † Kostin A. † Sechin I. † Medvedev D. † Mevedev D. † Sechin I. † Chemezov S. † Chemezov S. † Chemezov S. † Zakharchenko D. † Rotenberg A. † Sobyanin S. † Sobyanin S. † Klimenko G. Navalny A. † Bastrykin A. † Zakharchenko D. † Kamenschchik D. Timchenko G. Navalny A. † Navalny A. † Mints B. Bastrykin A. † Rotenberg A. † Gref G. † Chemezov S. † Deripaska O. † Deripaska O. † Kostin A. † Oreshkin M. Kostin A. † Gref G. † Fig. 4 presents a birds-eye view of the original kompromat network, called aninduced network. Each node in the induced network stands for a community in theoriginal network. Node size represents the number of persons in the community. Theinduced network edges are weighted; their weight (thickness) represents the num-ber of individual edges in the original network. The communities are named afterthe most prominent persons—the persons with the highest betweenness centrality.
Dmitry Zinoviev
Nabiullina ElviraSobyanin SergeyKerimov SuleymanVekselberg Viktor Putin VladimirBastrykin Aleksandr Sechin IgorYakunin VladimirTrump DonaldRadaev Valery Mordashov AlekseyGref GermanRotenberg ArkadyPrigozhin EvgenyMiller AlekseyChemezov SergeyKadyrov Ramzan Ivanov SergeyAbramovich RomanKarelin Aleksandr Poroshenko Petr Rogozin DmitryAvetisyan ArtemTolokonsky ViktorLushkov YuryChayka Igor Mutko VitalyPoltavchenko GeorgyChayka Yury Kostin AndreyKhodorkovsky MikhailShuvalov IgorMikhaylov SergeyIsmailov Telman Ionin Andrey Nikiforov SergeyRaykin KonstantinFedorov Evgeny Kogan VladimirMolokoedov AndreyKorshunov Oleg Svitova Natalya Novikov ArkadyKhakim Azat Nikitin Aleksandr Fursenko SergeyOlersky ViktorStreltsov Fedor Vafin AdelPetrov OlegBuderin Ilya Zhelyabovsky YuryAlekseenko AndreyStavsky Sergey Fardiev IlshatKamalov KhadzhimuradPrel' ArturTyurina Irina Borodin MikhailBeloborodov AlekseyKondratenkov AleksandrShatokhin Evgeny Nabiullina ElviraSobyanin SergeyKerimov SuleymanVekselberg Viktor Putin VladimirBastrykin Aleksandr Sechin IgorYakunin VladimirTrump DonaldRadaev Valery Mordashov AlekseyGref GermanRotenberg ArkadyPrigozhin EvgenyMiller AlekseyChemezov SergeyKadyrov Ramzan Ivanov SergeyAbramovich RomanKarelin Aleksandr Poroshenko Petr Rogozin DmitryAvetisyan ArtemTolokonsky ViktorLushkov YuryChayka Igor Mutko VitalyPoltavchenko GeorgyChayka Yury Kostin AndreyKhodorkovsky MikhailShuvalov IgorMikhaylov SergeyIsmailov Telman Ionin Andrey Nikiforov SergeyRaykin KonstantinFedorov Evgeny Kogan VladimirMolokoedov AndreyKorshunov Oleg Svitova Natalya Novikov ArkadyKhakim Azat Nikitin Aleksandr Fursenko SergeyOlersky ViktorStreltsov Fedor Vafin AdelPetrov OlegBuderin Ilya Zhelyabovsky YuryAlekseenko AndreyStavsky Sergey Fardiev IlshatKamalov KhadzhimuradPrel' ArturTyurina Irina Borodin MikhailBeloborodov AlekseyKondratenkov AleksandrShatokhin Evgeny
Fig. 4
An induced kompromat network: a birds-eye view of the original network in which the nodesrepresent communities.
The node color represents the mean betweenness centrality of the community nodes:darker nodes have a higher centrality. (One exception is the node labeled “TrumpDonald”—it is painted pink because it contains disproportionally many foreign na-tionals.)Table 3 additionally shows the top five most prominent persons in each commu-nity (some names have been spelled in Russian and English in the original dataset,which resulted in duplications; also, Igor Slyunyaev changed his unpleasantly sound-ing name to Igor Albin and was recorded in the same community under two names).The community
Social Network of Russian “Kompromat” 9
Table 3
Top five most prominent persons (in terms of betweenness centrality) in each community (seealso Table 1). The dagger † denotes English-language duplicates. The star (cid:63) denotes foreign nationals. Theequal sign = denotes the same person known under two different names.Community label Other persons1 Putin V. Medvedev D., Navalny A., Deripaska O., Timchenko G.2 Bastrykin A. Zakharchenko D., Sugrobov D., Titov B., Chubays A.3 Sobyanin S. Magomedov Z., Gutseriev M., Mints B., Ananyev A.4 Sechin I. Sechina M., Leontyev M., Rakhmanov A., Avdolyan A.5 Rotenberg A. Peskov D., Rotenberg B., Lisin V., Rotenberg I.6 Kostin A. Evtushenkov V., Rakhimov U., Gagiev A., Khamitov R.7 Chemezov S. Prokhorov M., Manturov D., Khudaynatov E., Ignatova E.8 Nabiullina E. Tokarev N., Bedzhamov G., Pugachev S., Markus L.9 Shuvalov I. Shuvalov I. † , Kesaev I., Levchenko S., Filev V.10 Luzhkov Yu. Chayka A., Yurevich M., Dubrovsky B., Rashnikov V.11 Rogozin D. Lavrov S., Patrushev N., Komarov I., Vinokurov A.12 Poltavchenko G. Beglov A., Albin I. = , Slyunyaev I. = , Oganesyan M.13 Gref G. Usmanov A., Mamut A., Durov P., Dmitriev V.14 Kerimov S. Matvienko V., Turchak A., Golodets O., Petrenko S.15 Chayka Yu. Shoygu S., Serdyukov A., Vasilyeva E., Ivanov T.16 Mutko V. Rashkin V., Ablyazov M. (cid:63) , Rodchenkov G., McLaren R. (cid:63)
17 Abramovich R. Vorobyev A., Potanin V., Sobchak K., Lebedev V.18 Ismailov T. Usoyan A., Mitrofanov A., Dzhaniev R., Varshavsky A.19 Tolokonsky V. Morozov S., Shantsev V., Khinshteyn A., Nazarov V.20 Poroshenko P. (cid:63)
Yanukovich V. (cid:63) , Surkov V., Belykh N., Kolomoysky I. (cid:63)
21 Trump D. (cid:63)
Cherkalin K., Mishustin M., Belousov A., Tkachev I.22 Kadyrov R. Ulyukaev A., Nemtsov B., Timakova N., Galchev F.23 Ivanov S. Gordeev A., Skrynnik E., Ivanov A., Korolev O.24 Vekselberg V. Golunov I., Golubev V., Varshavsky V., Blavatnik L. (cid:63)
25 Khodorkovsky M. Skripal S., Petrov A., Litvinenko V., Lebedev P.26 Avetisyan A. Calvi M. (cid:63) , Nazarbaev N. (cid:63) , Maduro N. (cid:63) , Delpal P. (cid:63)
27 Miller A. Slipenchuk M., Khlebnikov P. (cid:63) , Lurakhmaev V., Lanin M.28 Yakunin V. Belozerov O., Tikhonova E., Tikhonova K. † , Gorbuntsov G.29 Chayka I. Traber I. (cid:63) , Zhirinovsky V., Skoch A., Yarovaya I.30 Mordashov A. Shvets A., Novak A., Cyril (Patriarch), Khotimsky S.31 Mikhaylov S. Malofeev K., Strelkov I., Kaspersky E., Girkin I.32 Prigozhin E. Prigozhin E. † , Kligman I., Gerasimenko A., Uss A.33 Radaev V. Savelyev V., Lebedev A., Shishov O., Vantsev V. act as an unregistered foreign agent of Russia within the USA [14]. The community’scomposition suggests that it concerns Russia’s controversial interference in the 2016U.S. presidential elections [4].In the final stage of the analysis, we identified several types of communities,based on the affiliations of their most prominent members with one of the followingcategories: “business,” including state corporations (53 people in Table 3), “politics”(50), “law enforcement” (known in Russia as “siloviks,” or “people of force” [10];16), “banking” (15), “government officials” (13), “criminal world” (6), “press” (5),and “others” (4). Sometimes, there was more than one affiliation per person: for ex-ample, Igor Sechin, as the CEO of Rosneft’, is an entrepreneur, but since Rosneft’ isa state oil company, he is also a government official. In such cases, we selected themost notable affiliation. Table 4
Kompromat types: “business” T , “politics” T , “banking” T , and “law enforcement” T .Type T T T T Business 39 7 4 3Politics 11 29 8 2Banking 4 — 11 —Law Enforcement 1 2 2 11Government 5 2 3 3Criminal 6 — — —Press 5 — — —Other 2 — 1 1
As a result, we described each community with eight numbers—in other words,represented it as a point in 8-dimensional space, to a total of 33 points. For example,of the five most prominent persons in community k = ), politicians (T ), bankers (T ), and “siloviks” (T ). However, typesT and T also have a significant secondary population of politicians, and T ad-ditionally includes entrepreneurs. Thus, the first three types represent a corruptedsymbiosis of industrial and banking capital and political power, biased towards oneof the factions, depending on the type.The fourth type, T , comprises the “siloviks” and has few representatives from theother categories. The difference suggests that the kompromat cases involving law en-forcement officials, though not entirely isolated, differ from those that have affectedthe political and economic block. In this paper, we analyzed a social network of 11,000 Russian and foreign nationals,including politicians, entrepreneurs, bankers, law enforcement officials, and high-profile criminals, affected by kompromat: compromising materials. The data for thestudy was downloaded from “RuCompromat,” a Russian online encyclopedia of kom-promat. The network is modular and has an excellent community structure. Each
Social Network of Russian “Kompromat” 11 network community brings together persons who participated in similar kompromatcases. We calculated the network attributes (such as various centralities and clus-tering coefficient) and identified the most prominent persons in the whole networkand each community. By looking at the top community members’ affiliations withvarious socio-economic groups, we introduced a four-type taxonomy of the commu-nities. Three types represent industrial and banking capital and political power; lawenforcement officials (the “siloviks”) dominate the fourth.“RuCompromat” offers an additional level of information: organizations involvedin the kompromat cases. In the future, this information can be used to construct andanalyze a joint socio-organizational network. Since organizations are usually easierto classify than individuals, adding them to the dataset could help us automaticallyassign persons to the categories, which would improve the kompromat cases’ typol-ogy.
Acknowledgment
The author is grateful to Pelin Bic¸en, Professor of Marketing at Suffolk University,and Vasily Gatov, Russian media researcher and author, for their encouragement andhelpful suggestions.
Conflict of interest
The authors declare that they have no conflict of interest.
Funding
Not applicable.
Availability of data and material
Not applicable.
Code availability
Not applicable.
References
1. Blondel, V., Guillaume, J.L., Lambiotte, R., Lefebvre, E.: Fast Unfolding of Communities in LargeNetworks. J. of Statistical Mechanics: Theory and Experiment (10), 1000 (2008)2. Choy, J.: Kompromat: A theory of blackmail as a system of governance. Journal of DevelopmentEconomics p. 102535 (2020)2 Dmitry Zinoviev3. Freeman, L.: Centrality in Social Networks: Conceptual Clarification. Social Networks , 215–239(1979)4. Hayes, N.: Trump’s kompromat (2017)5. Iakhnis, E., Badawy, A.: Networks of power: Analyzing world leaders interactions on social media.arXiv preprint arXiv:1907.11283 (2019)6. Ledeneva, A.V.: How Russia really works: The informal practices that shaped post-Soviet politics andbusiness. Cornell University Press (2006)7. Markowitz, L.: Beyond kompromat: Coercion, corruption, and deterred defection in Uzbekistan.Comparative Politics (1), 103–121 (2017)8. Newman, M.: Modularity and community structure in networks. Proceedings of the NationalAcademy of Sciences of the United States of America (23), 8577–8696 (2006)9. Pearce, K.: Democratizing kompromat: the affordances of social media for state-sponsored harass-ment. Information, Communication & Society (10), 1158–1174 (2015)10. Reddaway, P.: 2007–2008: War fades, tandem forms, Cherkesov clan dissolves, Sechinites decline;Putin generates new factional warsgeneral procuracy vs the investigations committee of the procuracy(icp), Medvedev vs Sechin. In: Russias Domestic Security Wars, pp. 89–93. Springer (2018)11. RuCompromat: Encyclopedia of the Library of Kompromat. Retrieved in 2020. URL http://rucompromat.com/
12. Ruspress: Patryshev-mladshiy i pravoslavnye pensionery. Ubytochnyy Rossel’khozbank idyot na po-moshch “Peresvetu”. Published online in Russian (2018). URL . Removed, cached copy available13. Shi, X., Adamic, L., Strauss, M.: Networks of strong ties. Physica A: Statistical Mechanics and itsApplications (1), 33–47 (2007). DOI 10.1016/j.physa.2006.11.07214. Swaine, J.: Maria Butinas alleged backer linked to Kremlin-financed bank and Putin asso-ciates. The Guardian (2018). URL
15. Valente, T., Coronges, K., Lakon, C., Costenbader, E.: How correlated are network centrality mea-sures? Connections (Toronto, Ont.) (1), 16 (2008)16. Wasserman, S., Faust, K.: Social network analysis: Methods and applications, vol. 8. Cambridgeuniversity press (1994)17. Watts, D.J., Strogatz, S.H.: Collective dynamics of “small-world” networks. Nature393