- 1 All translations from French to English are mine.
1The question of the nature of the social bond between the individuals that make up a society is a broad and recurrent subject of research in the social sciences. If interest, mutual assistance, law, sociability, and sympathy are notions that have historically shaped conceptions of the social link in economics, sociology, or political science, the notion of trust has imposed itself over the last few decades to profoundly transform the way this issue is addressed. As French sociologist Alain Caillé (1994, 3) points out, “[o]n discovering that men are only likely to remain linked insofar as they grant each other a minimum of trust, the question of the essence of the social bond becomes that of the determinants of trust”.1
2The notion of trust may even have appeared “misleadingly as a holy grail of the social sciences” (Laurent, 2012, 14), given the number of economic or social phenomena it applied to. These run from the most generic, such as exchange (Williamson, 1985), money (Aglietta et al., 2016) or the market (Tiotsop et al., 2014), to the most specific, such as corruption (Ulsaner, 2013), mafia systems (Gambetta, 1993) or even the international trafficking of pre-Columbian art (Canghiari, 2020). This proliferation of research is not so recent. More than twenty years ago, Bigley and Pearce described the studies focusing on the notion of trust as constituting an “oceanic volume of literature” (Bigley and Pearce, 1998, 406). This statement is even more relevant now that the interest in trust has grown steadily over the years. One of the main difficulties in reporting on this literature is its great diversity of methods, disciplines, and research topics.
- 2 I understand interdisciplinarity in the sense of Fontaine and Backhouse (2010), i.e., as situations (...)
3This article is a contribution to the scarce but growing literature on the use of quantitative and computational methods applied to the history of economics. The methodology on which this article is based, bibliographic coupling, offers the possibility of establishing research clusters according to their “cognitive similarity”, without having to presuppose the reason for the proximity (Claveau and Gingras, 2016, 554). Bibliometric approaches and network analysis are useful methods to assess the literature on trust as they have been shown to be powerful tools to map the development of a particular topic (Goutsmedt, 2021) and are particularly relevant to address interdisciplinary issues (Truc, 2022).2
- 3 In order to ensure the preservation of the data presented on the website, an archive version of the (...)
4After introducing the methodology and the corpus on which this study is based (Section 1), the article offers a comprehensive picture of how the concept of trust is studied in the social sciences (Section 2). The aim of this article is to provide a clear vision of what social science research on trust is. The main results are twofold. Firstly, despite an ever-increasing number of publications, and the messy first impression that this literature and its literature reviews can leave (Hardin, 2002, xxi), the fields of research on trust appear to be stable since the late 1990s. Secondly, we show that the difficulty of understanding this literature stems from the fact that the fields of research on trust do not fit into any disciplinary partition. A major conclusion can be drawn from this finding: there is no such thing as an “economics of trust”. We provide an alternative partition of this literature sometimes based on a methodology (such as experimental trust), a specific institution (such as trust in the medical field) or a particular conception (such as trust in terms of social capital). A website, named NORLoT (the acronym of the title of this article) has been created in order for researchers interested in trust to navigate more on their own (and in an intuitive way) the data and results on which this article is based.3
5This section first presents the corpus and its delimitation (1.1), then the method of bibliographic coupling (1.2), the lexicometric analysis (1.3), and finally, the data provided by the NORLoT website (1.4).4
- 5 The Core Collection is a curated collection which contains over 21,100 peer-reviewed scholarly jour (...)
- 6 Most deleted documents are reviews and reprints.
6The data used in this analysis comes from the Core Collection of the Web of Science (WoS), a website that offers subscription access to multiple databases providing comprehensive citation data for many different academic disciplines.5 The first step in the construction of the corpus was to extract all the documents with the word “trust” as a topic, i.e., documents that include the word “trust” in their title, abstract or keyword. Of the 141,580 documents available, only the category of “articles” is retained, giving a total of 101,416 articles.6 WoS associates with each article a number of tags that indicate the discipline(s) of the article and less frequently the field of research or the methodology. Since there is more than one tag per article, the sum of the percentages of all the tags for a given cluster is always greater than 100%.
7At this stage, a bibliometric analysis would not be relevant. Indeed, as the word “trust” is used too commonly, the corpus still contains many articles that use the word “trust” in a random way and not in the context of explaining a phenomenon that directly or indirectly concerns trust. We used bibliographic coupling to further delimit our corpus.
- 7 This article is made possible by an enthusiastic data community that builds and makes available ope (...)
- 8 Appendix 1 provides an illustration of the bibliographic coupling principle from the work of Clavea (...)
8The bibliographic coupling method, invented by Kessler in 1963, allows documents to be linked according to the number of references they share in their bibliography (Kessler, 1963).8 The greater the number of references shared by two articles, the greater the link between them. To normalize and weight the links, we used Salton’s cosine measure (Salton and McGill, 1983), which divides the number of references that two articles share by the square root of the product of the two articles’ bibliography size. This method, which is used by historians of economic thought such as Claveau and Gingras (2016), Goutsmedt (2021) or Truc (2021), has the advantage of considering the length of the bibliography and therefore does not give too much importance to articles with a large bibliography.
9Bibliographical coupling allows the detection of clusters, i.e., sets of references that have a certain cognitive proximity without presupposing the reasons for this proximity. The set of detected clusters displayed in a network map provides a macro picture of the general structure of the corpus by highlighting the “number of communities [clusters], the density of their links (within and between communities), and the position of nodes and communities in the core/periphery structure” (Goutsmedt, 2021, 564).
10Thanks to the bibliographic coupling method and the “giant component” filter, we can delimit our corpus one last time. This filter allows the exclusion of articles that use the word “trust” outside the context of explaining social phenomena related to trust. Numerous articles about “trust regions” are removed in this way. Trust regions is a mathematical optimization tool: therefore, these articles do not share common references with any of the main component articles and are excluded from the analysis.
- 9 More precisely, a lemma does not bear the flexion marks and allows the different morphological form (...)
- 10 TXM is a free open-source textometry software for text data analysis (Heiden, 2010).
- 11 On the calculation of this index, see TXM User Manual v0.7, 2018.
11The lexicometric analysis does not allow us to restrict our corpus, but to categorize our clusters. It enables us to identify the characteristic “lemmas”—i.e., the canonical form of words—of a sub-corpus by comparing the frequency of their use with the frequency of their use in the whole corpus (Bouzereau, 2021, 61).9 The lexicometric analysis conducted in this article is based on the exploitation of the titles and abstracts of the articles via TXM.10 The lexical specificity feature of TXM provides a specificity index which indicates the degree of specificity of a lemma in a sub-corpus, or in our case, in a cluster.11
12In addition to lexicometric analysis, cluster characterization is based on a set of elements such as reading the most cited articles and the most mobilized articles in the bibliography within the cluster, observing the most represented disciplinary tags, the most present journals and the year-by-year distribution of publications in each cluster. I rely on this thick network of clues to name the different clusters. However, the naming of the clusters remains a choice. The interested reader can critically scrutinize the complete set of results and all the statistics for every cluster on the NORLoT website.
13For each cluster established and studied in the article, whether in the alluvial or in the network maps, the NORLoT website offers a page such as the one presented in the example (Figure 1).
Figure 1. Screenshot of the Relationship Marketing (2006-2020) Cluster Page
14This includes:
-
the name of the cluster;
-
a word cloud that offers a more visual representation of lexical specificities;
-
the top-10 tag partition in the cluster;
-
the top-100 articles in the cluster (based on the number of citations);
-
the top-100 most shared references within the cluster (based on the number of citations within the cluster).
15The final corpus, after applying the giant component filter, contains 47,593 articles. Figure 2 corroborates the statement that interest in trust has grown steadily in the social sciences over the years. The distribution of disciplinary tags already shows the cross-disciplinary aspect of trust research. Business is the most represented discipline (its tag is found in 15.4% of the articles). It is followed by Management (14.0%), Computer Science (13.9%), Psychology (11.7%), Information Systems (8.2%) and Economics (6.9%).
Figure 2. Distribution of the 47,593 Articles by Year of Publication, since 1990
The blue bars represent the number of articles in our corpus. The orange line represents the share of these articles in the total number of publications per year in WoS.
- 12 From 1991 onwards, the number of publications in the corpus exceeded 100 per year.
16The notion of trust began to generate a significant number of articles in the 1990s.12 To study the different fields of trust studies, it is useful to divide the corpus into three time periods. The first covers publications on the subject from 1958 to 1990 (2.1). Although this is a large time window, it represents a limited number of articles (253). The next two periods will first be studied together with an alluvial that introduces a dynamic analysis of the research fields over the period 1990-2020 (2.2). The network analysis maps will provide static but more detailed results, first for the period 1991-2005 (2.3) and then for the period 2006-2020 (2.4).
- 13 There are a number of reasons for this, the most important being the fact that it is based on the b (...)
17The division into fifteen-year periods from the 1990s onwards is based on several justifications. First, it seems to be an appropriate choice from a technical point of view: the bibliographic coupling method is not recommended for periods longer than fifteen years.13 Second, trust and its research fields are deeply impacted by the way we exchange. The emergence of the Internet generated an important shift in this respect. The third period begins in 2006, the year the Internet surpassed one billion users. It will therefore allow us to study the impact of this technology on the research fields of trust.
18Psychology is by far the most represented discipline in the tags of the period, appearing in 130 of the 253 articles. Political Science, Business, Sociology, and Economics offer much more modest contributions (23, 15, 9 and 3 articles respectively). Two research topics can be identified in this first period. The first is on interpersonal trust, while the second is on institutional trust and more precisely on people’s trust toward political institutions.
19The research on the notion of interpersonal trust was initiated by the work of social psychologist Morton Deutsch (1958; see Simon, 2007, 83; Tazdaït, 2008, 28). Deutsch was initially commissioned by the Office of Naval Research to study the conditions that would foster attitudes of trust in a small group such as air or submarine crews. He carried out an experiment inspired by the prisoner’s dilemma on a group of college students. Through this two-person non-zero-sum game, he showed that it is possible to approach the notion of trust experimentally and that there are many situations in which cooperation is empirically observed but not predicted or explained by rational choice theory (Deutsch, 1958, 278). The “birth” of the notion of trust in the social sciences is therefore strongly linked to the emergence of game theory and more generally to the scientific context of the Cold War (Erikson et al., 2013, 145-149). Deutsch laid the groundwork for a research program that aimed to measure trust, and, although it would take a few decades between his publication and the more significant development of psychology and experimental economics, Deutsch’s article is the third most frequently cited in the bibliographies of the whole 1958-2020 corpus.
- 14 On the use of psychological tests in economics, see Dechaux (2017). On the differences in methodolo (...)
20The two articles most frequently found in the bibliography of the corpus were written by Rotter (1967; 1971), a psychologist who pursued the idea of measuring trust with a different methodology. In his 1967 article, Rotter sought to develop what he called an “Interpersonal Trust Scale” (Rotter, 1967, 653). His objective was to find a way to measure trust by avoiding the artificiality found in “laboratory situations” (Rotter, 1967, 52) encountered since Deutsch’s initial article (Rapaport and Orwant, 1962; Scodel, 1962). Rotter’s approach was based on the development of psychological tests that, depending on the score, assess the general level of trust that can be found in a group of people. This research, funded by a grant from the US National Institute of Mental Health, was intended to provide support for research in the areas of social psychology, personality, and physiological psychology. Rotter thus extended the use of psychological tests to the study and measurement of trust.14
21From the end of the 1970s onwards, other disciplines contributed to this interpersonal trust investigation initiated by psychologists. This was the case of some sociologists who were more interested in how trust should be conceptualized than in its measurement. Sociologist David Lewis (Lewis and Weigert, 1985) explicitly tried to build a bridge between sociological and psychological contributions on trust. He attempted to synthesize sociological contributions on the analysis of trust (Luhman, 1979; Barber, 1983), in a way that could give theoretical foundations for the experimental psychology approach which then “appear[ed] theoretically unintegrated and incomplete from the standpoint of a sociology of trust” (Lewis and Weigert, 1985, 967). His main criticism was that the experimental approach reduces trust to a purely cognitive object when it should be understood as a multidimensional social reality. This was also suggested by sociologist Susan Shapiro, who applied the concept of “embeddedness” developed a few years earlier by Granovetter (1985) to investigate the notion of trust.
22A few articles on interpersonal trust were written by management scholars. These articles deal with the impact of trust on the performance of organizations. This is notably the case of Zand (1972), the 6th most-cited article published during this period, which studied how “trust can significantly alter managerial problem-solving effectiveness” (Zand, 1972, 238). In his article, Zand conducts experiments to show that trust in a group “conveys appropriate information, permits mutuality of influence, encourages self-control, and avoids abuse of the vulnerability of others” (ibid.).
23Several articles by political scientists form the second trend in this first period. Centered around Miller’s article (1974), this group was interested in institutional trust and mainly in political trust. Miller’s article is based on the exploitation of data from a national cross-section of eligible voters for the years 1964, 1966, 1968, and 1970. It includes a battery of questions dealing with public policy on race relations, foreign affairs, and a variety of domestic problems. In this cluster, most of the studies of trust are based on surveys.
- 15 On the transition of this definition from sociology to economics, see Camilotto (2021).
24During this first period, a dichotomy between interpersonal and institutional trust emerges. The approach to interpersonal trust is at this stage dominated by psychology. Interpersonal trust can be defined as the level of expectation that an individual attributes to another individual or group of individuals that they will, in a defined context, act in a way that will be beneficial, or at least not detrimental, to him/her. This definition was originally formulated by the psychologist Erikson (1953) and popularized by Rotter’s work (1967; 1971). It then traveled from discipline to discipline, firstly to sociology (Luhmann, 1979; Gambetta, 1988), then to political science (Hardin 1993), and finally to economics (Williamson, 1993; Berg et al., 1995), where it became the commonly used definition for interpersonal trust (Simon, 2007).15 Institutional trust is defined as the trust that an individual attributes to a certain institution. During this first period, one institution was studied: government. Together, interpersonal trust and institutional trust form what is sometimes called social trust (Kwon, 2019, 13).
25Before looking in more detail at the two fifteen-year periods following the 1958-1990 period, an overview of the entire 1990-2020 period provides a first approximation of the delimitation of the research fields of trust.
Figure 3. Alluvial Diagram of 42,529 Articles Spread Over 15-Year Periods, from 1990 to 2020
- 16 On the methodology and use of the alluvial diagram, see Goutsmedt and Truc (2022).
26The alluvial diagram (Figure 3) presents the evolution of clusters over the 17 periods that cover the publications between the year 1990 and the year 2020. The first bar represents the clusters detected over the 15-year period from 1990 to 2004. The second bar represents the clusters detected over the period from 1991 to 2005, and so on until the seventeenth and last period which goes from 2006 to 2020. From one period to another, we can follow the evolution of the size of the clusters, their movement, and their emergence. 16
27The first property that emerges from this alluvial diagram is the stability of the trust literature since the first period 1990-2004. The purpose of this section is to present the main streams of trust research. To this end, we will focus less on a cluster-by-cluster analysis and more on a general presentation. On the other hand, the analysis developed for the bibliographic coupling networks will be the subject of a detailed analysis of each cluster.
- 17 Hereafter, the cluster titles are clickable links to the cluster page on the NORLoT website.
28If the notion of Social Trust does not appear as such in the alluvial diagram, this is because it covers three different literatures that intersect frequently—to such an extent that, depending on the period, they can appear as one single literature. Social trust brings together articles dealing with the Experimental approach to trust, articles that address trust through the notion of Social Capital and finally articles that focus on Political trust.17 Taken together, Economics is the most represented discipline in these three literatures, being tagged in 16.6% of the articles, followed by Political Science (14%), Psychology (13%) and Sociology (8.9%).
29Inter·Organizational trust appears as a cluster from the period 1995-2009. It brings together two initially distinct but related literatures: Organizational Trust literature, which focuses on the management of trust within these organizations, and Interorganizational Trust literature, dedicated to the role that trust can play in interorganizational exchanges. Management is the discipline that stands out in these clusters (35.1% of the articles). Nevertheless, Inter·Organizational Trust remains an interdisciplinary field of research with important contributions from Business (21.9%) and Psychology (16.2%).
30Since the period 1995-2009, some of the articles of the Organizational Trust cluster form a cluster that grows in importance over time: the Online Trust: User cluster. This cluster is the most tangible evidence of the impact of the internet on the literature on trust, as it focuses on trusting behaviours online. Five tags stand out in this cluster: Computer Science (28.7%), Business (25.9%), Information Science & Library Science (21.0%), and Management (20.6%).
31The Relationship Marketing cluster gathers articles that focus on trust in marketing relationships, i.e., trust between buyer and seller. This cluster is largely composed of Business articles (50.5%) and to a lesser extent Management articles (31.5%).
32Since 72.9% of its articles contain the Computer Science tag, this is the least interdisciplinary cluster of the alluvial diagram. Research in this cluster focuses on the technological means of generating trust online. This includes broad topics such as online reputation systems, blockchain and more generally “trust management for internet things” (Yan et al., 2014).
33The Trust and Risk cluster deals specifically with risk analysis. It is a field that applies to broad subjects that are found in specific lemmas in the cluster such as “risk”, “food”, “climate”, “water” or “vaccine”. This plurality of subjects is echoed in the partitioning of the tags. The most present tags are quite vast and concern research subjects rather than a particular discipline, such as Environmental Studies (15.8%), Environmental Sciences (14.5%), or Environmental & Occupational Health (11.3%).
34This cluster focuses on the notion of trust in the medical environment. Environmental & Occupational Health is the most represented tag (18.7%), followed by Medicine (13.0%).
35This cluster focuses on the psychological factors that impact on trust in personal relationships. Psychology is also the most present tag in the various publications (with 37.5% appearances).
36The alluvial diagram allows the depiction of a certain dynamic in our analysis by observing the evolution of clusters from period to period. It also allows to have a very general vision of the different fields of research on trust. Using network analysis maps is useful to refine our analysis and characterization of the clusters.
37A point on the disciplinary partition of the period (Figure 4) allows to grasp the extent of the research on trust during this period. Psychology, which was found in almost half of the contributions over the period 1958-1990, is now present in 15.4% of the publications. Two categories are now more represented, Business (17.2%) and Management (17.0%). Many other disciplines are also involved in the discussions such as Computer Science (13%), Environmental & Occupational Health (7.3%), Economics (6.5%), Sociology (5.8%) or Medicine (4.9%).
Figure 4. Bibliographic Coupling Network of the 4778 Articles on the Trust Literature, from 1991 to 2005
- 18 The visualization of the networks and the constitution of the clusters are done on Gephi, a free so (...)
Link on my GitHub to this map.18
38To explore the map, we will first study the three clusters that form a relatively dense group: Organizational Trust, Interorganizational Trust and Relationship Marketing. These three clusters account for 40.6% of the articles of this period. We will then study Social Capital and Political Trust with Experimental Trust, two relatively close clusters. Finally, we will study the more autarkic clusters: Medical Trust, Trust and Risk, and Online Trust: Technology.
39In this cluster, the largest in volume over the period (820 articles), the articles focus on what creates, maintains, or destroys trust relationships within organizations and on the effects of trust in those organizations. It is a cluster in which a wide range of disciplines contribute in a substantial way, such as Management (39.1%), Psychology (35.2%), Business (25.3%) or Computer Science (17.6%).
- 19 However, in order to interpret this type of result, it is important to bear in mind that being the (...)
40The article by Mayer et al. (1995) is foundational for this field of research. It is both the article with the most citations in the cluster and the one that is most mobilized in the references within this cluster.19 Published in the Academy of Management Review, this article laid the foundations for an analysis of organizational trust. By distinguishing the notion of trust from other related notions such as cooperation, confidence or predictability, the authors offer a precise conceptual definition. Trust is addressed, in the tradition of interpersonal trust, as:
[t]he willingness of a party to be vulnerable to the actions of another party based on the expectation that the other will perform a particular action important to the trustor, irrespective of the ability to monitor or control that other party. (Mayer, et al., 1995, 712)
41As a sign that the field of research was still being structured at that time, theoretical grounding in this article was largely based on work in psychology (Johnson-George and Swap, 1982; Kee and Knox, 1970) and sociology (Gambetta, 1988; Luhmann, 1988) but to a lesser extent on contributions in management (Zand, 1972), albeit the most represented discipline tag in the cluster.
42Among the articles with the most citations, the one by McAllister (1995) sheds an interesting light on the structuring of this research field. In his article, McAllister conducted tests on 194 managers and other professionals to understand the “nature and functioning of interpersonal trust” and the extent to which these trust relationships impact “individual and organizational effectiveness” (McAllister, 1995, 24). As with Mayer et al. (1995), McAllister’s theoretical framework is rooted in previous works outside of management: “Theoretical foundations were drawn from the sociological literature on trust and the social-psychological literature on trust in close relationships” (McAllister, 1995, 24).
43Near the cluster of Organizational Trust, Interorganizational Trust is the second most important cluster of the period with 667 articles. As in the Organizational Trust cluster, Management is the most present discipline (in 40.0% of the articles), but the other disciplines are quite different with Business (29.7%), Economics (14.7%) or Sociology (12.1%). This cluster focuses on trust relationships between organizations.
- 20 The notion of “embeddedness” appears as a specific lemma of the Interorganizational Trust cluster a (...)
44At this time, one of the recurring questions in this field of research concerned the relationship between interpersonal trust and interorganizational trust, both of which are “related but distinct” (Zaheer et al., 1998, 141). Hence the recurrent use of the notion of embeddedness, a concept developed by Polanyi (1983), then redefined by Granovetter (1985), for whom economic action is “embedded” within networks of personal relationships (Laville, 2008).20
45The issue of organizational trust is raised in a variety of organizational contexts. For Zaheer et al. (1998), the question arises as to trust relationships in buyer-supplier dyads. They conclude that a high degree of trust would help to lower the costs of negotiation and reduce conflict. Many articles focus on the role that law and justice can play in these interorganizational trust relationships (Robinson and Rousseau, 1994; Aryee et al., 2002; Sunshine and Tyler, 2003). This is the case of Poppo and Zenger (2002), who address the place of contractual commitments in trust. In this article, they attempt to empirically address a common belief in the trust literature: “complex contracts in interorganizational exchanges are substitutes for trust” (Poppo and Zenger, 707). Their study aims to show that these two elements could be better considered as complements.
46The last cluster forming the central and dense area, the Relationship Marketing cluster, is the fifth most important cluster of the period with 451 articles. The four most cited articles focus on the same subject: the role that trust plays in the buyer-seller relationship (Ganesan, 1994; Doney and Cannon, 1997; Garbarino and Johnson, 1999; Chaudhuri and Holbrook, 2001). It is a much less interdisciplinary cluster, in that Business is present in a very large part of the contributions (71.2%).
47The most cited article in the cluster is again the one that is most used in the references of this cluster. This is the article by Morgan and Hunt (1994), which lays the foundations for the study of trust in marketing. The authors develop what they call the Commitment-Trust Theory that opposes both political science and economics views on trust. According to them, political scientists tend to be interested in how trust is influenced by power relationships, and economists conceptualize trust with models based on pure and perfect competition. In contrast, the authors wish to develop a “cooperative theory of pure and perfect cooperation” (Morgan and Hunt, 20). This paradigm shift would be, for the authors, the new cornerstone for addressing, in theory and practice, the establishment, development, and maintenance of successful relational exchanges. The authors conceptualize trust as existing “when one party has confidence in an exchange partner’s reliability and integrity” (ibid., 23). This is an original definition which takes into account the moral sense of the person with whom one is interacting.
48This cluster is a bit peculiar as it appears to be split in two. The first part, on the left of the map, groups together publications mainly from Political Sciences (37.9%) and is concerned with political trust: “A basic evaluative orientation toward the government founded on how well the government is operating according to people’s normative expectations.” (Hetherington, 1998, 791)
- 21 The political trust and social capital approaches will appear in separate clusters in the following (...)
49The second part is more interdisciplinary, with the contribution of Sociology (15.2%), and deals with trust via the notion of social capital (Knack and Keefer, 1997; Kawachi, 1997; Sampson et al., 2002).21 At the end of the 1980s, and then in the mid-1990s, two publications allowed the notion of social capital to achieve a certain prominence: Coleman’s (1988) “Social Capital in the Creation of Human Capital” and Putnam’s (1995) “Bowling Alone: America’s Declining Social Capital”. They are respectively fifth and first most mobilized references within this cluster. According to these two studies, social capital is generally defined as a type of resource embedded in relationships between individuals that facilitates cooperative and collaborative actions within society (Coleman, 1988, 98; Putnam, 1993, 67).
50The social capital approach can be used to address both institutional trust (Mishler and Rose, 2001; Brehm and Rahn, 1997)—which is why the cluster is relatively close to the Political Trust cluster—and interpersonal trust (Knack and Keefer, 1997; Paxton, 1999) with articles that are closer to the Experimental Trust cluster. This approach applies to a variety of subjects, such as the impact of social capital on health (Kawachi et al., 1999), well-being (Helliwell and Putnam, 2004), or innovations (Dakhli and De Clercq, 2004).
51With only 244 articles, the experimental trust cluster is only the 8th cluster in terms of volume. Economics is the most prominent discipline in this cluster (43% of the articles). Psychology, which had initiated the experimental approach to trust in the previous period, is also present (25.8%).
- 22 The first game to test trust is usually considered to be the ultimatum game initiated by Siegel and (...)
- 23 To be more precise about the measures: the trust measure is the result of the quotient of the sum s (...)
52The most important article of this cluster in this period was written by Berg et al. (1995), who offer a type of game known as the investment game, which was later renamed the trust game.22 The experimental protocol proposed by Berg et al. (1995) is both the most cited article in the cluster and the most used reference within the cluster. In this game, two groups of people are separated in two different rooms A and B. In stage one, participants are given a $10 show-up fee. While subjects in room B pocket their show-up fees, subjects in room A must decide how much of their $10 to send to an anonymous counterpart in room B. The money received by subjects in room B is multiplied by 3, so if a subject in room A decides to send their full $10, a subject in room B will receive $30. In the second phase, it is up to subjects in room B to make a choice. They can send more or less of the money back to the subjects in room A (the ones who initially sent the money) and keep the rest, or they can keep the whole amount. This game “provides a role for the use of trust in achieving a joint improvement to the subgame perfect outcome” (Berg et al., 1995, 125). In this experimental protocol, the amount sent by subjects in room A measures trust, while the amount sent by subjects in room B measures trustworthiness.23 This experimental setup became popular because it offers results that run counter to the predictions of game theory. The players in room B have no monetary incentive to send money back to the players in room A. Anticipating this, the players in room A should keep the entire amount for themselves. Yet the experimental results show robustly and consistently that trust appears between players.
53Of the last three relatively isolated clusters, the medical trust cluster is the largest (570 articles). This cluster has a specific object of study: trust in the relationship between health professionals (or health institutions) and their patients. Medicine is the most present discipline (in 31.4% of the articles), followed by neighbouring disciplines such as Health Care Sciences (28.6%) or Environmental & Occupational Health (21%).
- 24 On the relationship between trust and vulnerability, see Wiesemann (2017).
54Although the subject is specific, the study of trust in these contributions has some conceptual ambitions. One of the most quoted articles in the cluster calls for an “explicit conceptual framework for trust” (Hall et al., 2001, 632) and is devoted exclusively to the definition of trust and its possible measurement. While it is quite commonly assumed that increased vulnerability should produce less trust (Pellegrino et al., 1992), medicine has shown that this relationship is not so straightforward.24 In fact, the opposite can happen when vulnerability leads patients to see professionals as demigods, imbued with superhuman powers (Katz, 1984). The medical field is therefore an important example of cases where “the greater the sense of vulnerability, the higher the potential for trust” (Hall et al. 2001, 615). This is a cluster in which trust is approached and measured either through experiments (Anderson and Dedrick, 1990) or through surveys (Corbie-Smith et al., 2002).
55This cluster concern risk analysis (390 articles). Risk analysis focuses mainly on “understanding the root causes of social conflict” (Slovic, 1991, 17). It is applied in many different domains: nuclear risk (Slovic et al., 1991), food safety (Frewer et al., 1996; Krystalli and Chryssohoidis, 2005) or media credibility (Tsfati and Cappella, 2003). The disciplinary tags are not particularly informative about the disciplinary distribution of this cluster since the first one is the tag “Social Sciences” (25.6%), and the others are more about research topics rather than disciplines, such as “Environmental & Occupational Health” (26.5%). In this respect, risk analysis appears to be a quasi-discipline. From an institutional point of view, risk analysis also has its own journals: 8 of the 10 most cited articles in this cluster come from the same journal (Risk Analysis), a journal of the Risk Analysis Society, an academic society created in 1980.
56Online Trust: Technology is the least interdisciplinary cluster of the period with 91.6% of the articles related to Computer Science. Another particularity of this cluster comes from the temporal distribution of its articles: 75.0% of the 379 articles were published between 2003 and 2005. This is due to the growing importance of the Internet, and what Dellarocas (2003, 1407) calls “The Digitization of Word-Of-Mouth”. Indeed, the goal of most contributions within the cluster is to build technology that fosters trust and cooperation in online marketplaces.
57There are many developments between this period and the previous one. Concerning interpersonal trust, economics joined psychology in the experimental approach, and the publication of the trust game by Berg et al. (1995) opened up a significant field of research in economics. Interpersonal trust was also the subject of a new type of analysis via the notion of social capital. Finally, a particular type of interpersonal trust, namely interpersonal trust within organizations, generated a significant number of articles. Regarding institutional trust, politics and government were no longer the only institutions studied. Many articles focus on the trust of individuals in companies or brands, in the medical sector or in online platforms.
58The fields of research on trust are relatively stable over the period 1990-2020 (Figure 5). The study of the second period, however, allows us to refine the analysis even further for two reasons. Firstly, the research fields that emerged in the previous period appear more structured and distinct. Secondly, the appearance of meta-analyses and surveys makes it easier to grasp the most salient research subjects, methods, and results of each cluster.
Figure 5. Bibliographic Coupling Network of the 42,353 Articles on the Trust Literature, from 2006 to 2020
Link on my GitHub to this map, click on the download button and open the file on Gephi.
59The increase in both the number of articles published and clusters found is noticeable. The study of the disciplinary partition does not show a drastic change: Business is still the most represented discipline, present in 15.1% of articles. Computer Science is now second with 14.1%. Management (13.8%), Psychology (11.0%) and Economics (6.9%) follow.
60Clusters are addressed in the same way as in the previous period. Firstly, by focusing on the clusters that form a central and relatively dense group: Organizational Trust, Interorganizational Trust, Relationship Marketing and Online Trust: User. Then by the three relatively close clusters, Political Trust, Social Capital, and Experimental Trust.
61As in the previous period, the Organizational trust cluster, with 4108 articles, is the largest cluster of the period. The disciplinary partition did not really evolve from one period to the next, except for Computer Science which fell from 17.6% of contributions to 4%. However, this is not so much due to any change in the research focus of the cluster in question, but rather to the fact that the online issues, which were already present in the previous period, are now found in the Online Trust: User cluster.
62The article by Colquitt et al. (2007), the most cited article in this cluster, highlights the benefits of fostering trust in the workplace. Trust influences job performance, reduces absenteeism and employee turnover, and finally can predict counterproductive behaviours (Colquitt et al., 2007, 922). A sign of continuity between the two periods is the fact the articles by Mayer et al. (1995), Rousseau et al. (1998) and McAllister (1995) are the three most frequently used references in the cluster.
63Two observations must be made concerning this cluster’s statistics. Firstly, while it was the second in terms of volume in the previous period, it is now the 10th. This is due in particular to the fact that the contributions are now focusing on the influence of the strategic alliance on firm performance (Phelps, 2010). Indeed, this cluster is much less interdisciplinary than it was in the previous period. Still dominated by Management, the interdisciplinary questioning around the issue of embeddedness is much less present, which is coherent with the fact that Economics and Sociology are tagged in only 4.8% and 3.5% of the articles (compared to 14.7% and 12.1% previously).
64The Relationship Marketing cluster gained considerable importance during this second period and is now the second-largest cluster (4005 articles). It remains in continuity with the work of the previous period, both in the disciplines represented and in the most cited references within the cluster.
65There is now a substantial body of research on the impact of the internet on relationship marketing. It is a broad subject of research that covers a variety of topics such as consumer engagement in an online brand community environment (Brodie et al., 2011, 105), social media (Sashi, 2012; Laroche et al., 2012), or digital influencers (Jimenez-Castillo and Sanchez-Fernandez, 2019).
66This is a new cluster, which appeared in the alluvial diagram from the period 1996-2011. The Online Trust: User cluster is becoming increasingly voluminous. Over the period 2006-2020 it is the fourth-largest cluster (3507 articles). It is a relatively interdisciplinary cluster where Computer Science and Business are the most represented disciplines (31.2% and 28.5%).
67While the Relationship Marketing cluster focuses on the strategic mobilization of online activities for brands, this cluster focuses on the behaviour of individuals in relation to the new ways of exchanging that these technologies offer. Individuals’ online trust is subject to multiple determinants, such as perceived risk, security, privacy, reputation, or usefulness (Kim and Peterson, 2017, 45). The ability to generate trust plays a key role in online exchanges, as Reichheld and Schefter (2000, 107) point out: “Price does not rule the Web [market]; trust does”.
68Most of the cluster’s topics concern e-commerce, such as the role of feedback text comments in online marketplaces (Pavlou and Dimoka, 2006), Internet banking (Azouzi, 2009) or the determinants of consumer adoption of e-commerce (Pavlou and Fygenson, 2006). However, a significant part of the cluster focuses on the development of e-government defined as:
[t]he use of information and communication technologies (ICTs) and the Internet to enhance the access to and delivery of all facets of government services and operations for the benefit of citizens, businesses, employees, and other stakeholders (Thompson et al., 2008, 100).
69Economics and Psychology are still the most represented disciplines. Neuroscience is now present in 5.7% of publications and it is worth noting that two articles concerning the effect of oxytocin on trust are among the most cited articles in the cluster (Kosfeld et al., 2005; Baumgartner et al., 2008). Cited in 40.8% of the articles in the cluster, the article by Berg et al. (1995), is by far the most widely used reference in this cluster. The trust game is so popular because it offers particularly robust results for several reasons.
- 25 For instance, the double-blind treatment drastically reduced the amount of money sent in the dictat (...)
70Firstly, the trust game intentionally seeks out and even encourages selfish behaviour. By conducting the experiment in a double-blind fashion, experimenters send the message “that it is OK not to send money, and OK to keep the money received.” (Smith, 2020, 4) This is also what makes this result so powerful: trust remains, despite an experimental protocol that encourages its disappearance.25 Secondly, the experiment has been replicated many times. In 2011, Johnson and Mislin proposed a meta-analysis of 162 replications of the trust game involving over 23,000 subjects in 35 different countries.
71Political trust and the approach through the notion of social capital now appear in two separate clusters. The notion of social capital, although developed in sociology, is at this time frequently used in Economics, which is the tag most present over the period (28.3%). Sociology is still present in 17.3% of publications.
72If economics is so important in this cluster, it is because one of the recurring themes of this cluster is the impact of the level of social capital on economic performance. Social capital plays a role in economic performance mainly through the generalized trust it can generate (Bjørnskov, 2007). Nevertheless, there are debates about the direction of the relationship between social capital and trust (Ponthieux, 2006).
73As the cluster dealing with political trust has been stable since the first period, it will not be discussed in depth. Now that the social capital approach appears as an isolated cluster, the disciplinary partition is much less interdisciplinary: Political Science is present in 44.6% of publications. The other tags are much further away, Public Administration is present in 15.7% of the publications, followed by Sociology (9.9%).
74Online trust technologies have stabilized around the idea of rating:
The basic idea is to let parties rate each other, for example after the completion of a transaction, and use the aggregated ratings about a given party to derive a trust or reputation score, which can assist other parties in deciding whether or not to transact with that party in the future. (Jøsang et al., 2007, 618)
75Most articles focus on how to improve these technologies, the quality of which is based on four criteria: accuracy for long-term performance, weighting towards current behaviour, robustness against attacks and smoothness (ibid.).
76A notable change in the Trust and Risk cluster comes from the journals, which are much more diverse than in the first period. There are also new research topics related to societal developments. This is the case, for example, with articles on the public perception of scientific consensus (Kahan et al., 2010), whether it concerns global warming issues (Tranter and Booth, 2015), or health, with numerous articles on Covid-19 (Dryhurst et al., 2020; Imhoff and Lamberty, 2020).
77Regarding Medical Trust, we can underline the important role that trust seems to play in the medical process. The bond and communication between the practitioner and the patient can lead to better health. This is achieved through various channels including better access to care, better knowledge and understanding shared by the patient, better medical decisions, strengthened therapeutic alliances, increased social support, empowerment, and better emotional management. (Street et al., 2009).
- 26 This result is mainly based on the interpretation of the alluvial diagram (Figure 3). Since the lat (...)
78The analysis of the two maps confirms what appeared in the alluvial diagram and which is the first important result of our study: since the beginning of the 1990s, the fields of research on trust have formed stable sets.26 Seven different main search fields can be established: Social trust, which combines experimental trust, political trust and trust in terms of social capital (1), Inter-organizational trust, which brings together articles that focus on trust within or between organizations (2), the trust between buyer and seller that is developed in the Relationship Marketing literature (3), medical trust (4), the socio-economic risks that can arise from a lack or excess of trust (5), and finally two fields that focus on online trust, either through the prism of technology (6) or through the prism of the users (7).
- 27 The title of Evans and Krueger’s (2009) article: “The Psychology (and Economics) of Trust” illustra (...)
79To my knowledge, there is no extensive presentation of trust research in the social sciences similar to the one presented in this article. Our analysis offers two reasons as to why this is the case. Firstly, the sheer volume of contributions makes it difficult to apprehend this literature with the naked eye. Secondly, and more importantly, the research literature does not lend itself to a disciplinary partitioning. Attempts to define what might be the “economics of trust”, the “sociology of trust” or the “psychology of trust” always fail to hold these boundaries because the fields of research are, with a few exceptions, interdisciplinary.27 To prove this point, an attempt to approach the work on trust in economics through bibliographic coupling gives the results in Figure 6.
Figure 6. Alluvial Diagram of the 3247 Economics Articles Spread Over 15 Periods, from 1990 to 2020
80Three clusters already studied in the cross-disciplinary analysis can be found in the economics articles that deal with trust. Studying trust by disciplinary partition would therefore imply studying a part of a cluster without grasping its interdisciplinary aspect. From these elements another major result of our research stands out: there is no such thing as an “economics of trust” but rather articles in economics that focus on trust. Some of these articles focus on trust experimentally and engage mainly with psychology. Some others focus on relationship marketing and engage mainly with marketing and yet other articles focus on social capital and engage mainly with sociology and political science.
81As the objective was to delimit the fields of research on the notion of trust, the level of detail chosen in this article is deliberately low. This perspective fails to capture the potential different views of trust that may exist within a cluster. In the work on the notion of social capital, for example, there are at least two opposing views: on the one hand, those who, in the Bourdieusian tradition, define social capital as exogenous, and on the other, those who, in the tradition of Coleman and Putnam, consider it to be endogenous (Ponthieux, 2006, 91). There are also three different approaches to organizational trust according to Zhong et al. (2014): transaction cost economics, social embeddedness theory, and resource dependence theory.
82By zooming in on each cluster, our method could detect different approaches and help to identify the different meanings of trust. A complex task, as Hardin, one of the major authors on the notion of trust, pointed out in the early 2000s:
Unfortunately, conceptual issues in the understanding of trust are far messier and more complicated than one might hope. Clearing up these issues turns out to be a major task. (Hardin, 2002, xxi)
83By trying to clarify the structure of the research fields of trust, this article can be helpful in this major task.
I would like to thank the editors of this special issue, François Claveau, Catherine Herfeld and especially Aurélien Goutsmedt who helped me from the beginning of this work. I also thank Francesco Sergi, the managing editor of Œconomia, for his kindness as well as the two anonymous reviewers, whose helpful comments and suggestions helped improve the quality and relevance of the article. I am also grateful to Dorian Jullien, Justine Loulergue, Nicolas Brisset, Alexandre Truc, Marie Cavallo and Julie Camilotto for their assistance. All errors and omissions remain mine.