Lambert says here: “There is no such thing as civil society.” -Maggie Thatcher (Apocryphal)
Samuel Bowles, Research Professor and Behavioral Sciences Program Director, Santa Fe Institute, Wendy Carlin, Scientific Council, Paris School of Economics (PSE), External Professor, Santa Fe Institute, Professor of Economics, University College London, Written by Sahana Subrahmanyam, Stanford I am a third-year student in a doctoral course in economics at university. Her research interests lie at the intersection of gender, development, and behavioral economics. Originally published on VoxEU.
Commentary on the 2024 US election included the Democratic Party’s failure to connect with issues of community, place, family, religion, and identity. [and class–lambert]. In this column, I present evidence from research papers published in top economic journals since 1900 that economists paid some attention to these very topics half a century ago. The authors document this shift and show that it is associated with new empirical methods such as experiments, large datasets, and increased attention to social norms and strategic interactions, including the exercise of power. are.
Significant shifts in the focus of economic research over the past half century have largely gone unnamed and unnoticed. This is a shift toward so-called civil society, including businesses as organizations, families, neighborhood communities, NGOs, trade unions, social movements, identity groups, and other face-to-face settings. This term may be unfamiliar to many economists, but its origins date back to Adam Smith and the Scottish Enlightenment, and it has been widely applied in other social sciences and philosophies (Cohen and Arato 1994, Edwards 2011, Ferguson 1767, Smith 1976) [1759]).
Economic interactions in civil society, as we use the term here, have in common that the relationships are personal and permanent. As a result, identity and other preferences (for better or worse) become important motivators. In civil society, the primary mechanisms guiding economic interactions are neither government regulation nor perfect contracts or market-determined prices, but instead private authority and power, social norms, and group identity. external hostility), and reputational hierarchies. (The concept of civil society is discussed further in Bowles and Carlin 2025.)
The turn to civil society in economics that we document (Bowles et al. 2025) is consistent with two propositions. First, much of what we think of as the economy consists of imperfectly contracted, non-market interactions and exchanges, not just within firms but also in labor, credit, housing, and other markets. It means that it is. Second, in these and other situations, ethical and other-regarding preferences occupy an important place alongside self-interest in explaining behavior and supporting mutually beneficial exchanges (Arrow 1971).
The proliferation of civil society themes we are documenting in our research corpus is driven by conceptual innovations (including the information revolution and game theory) and empirical methods (experiments, other advances in causal attribution, large data sets, (including the use of ). A shift in focus has been demonstrated by Myrdal and Hayek (1974), Simon (1978), Coase (1991), Nash (1994), Sen (1998), Akerlof and Stiglitz (2001), Kahneman and Smith (2002) ), Schelling (2005), Ostrom and Williamson (2009), Hart (2016), Thaler (2017), Goldin (2023), and a few weeks ago, Acemoglu, Johnson, and Robinson (2024).
Markets, states, and civil society: Topic modeling themes in economic research
We represent civil society alongside the state and market as distinct and ideally complementary structures of social governance. Figure 1 treats the pure types of these structures as unitary vertices that represent the space for economic policies and institutions. Each of these extreme cases is characterized graphically by both a set of rules of the game (through which quotas are implemented) and a set of preferences and social norms (related to how they contribute to the governance of society as a whole). It is being The coordinates of any point on the simplex sum to 1 and represent a weight that indicates the relative importance of the state, market, and civil society (for a given point in space, the largest weight is importance).
Figure 1: Institutional and policy space centered on the ideal type of social governance, its implementation mechanism, and behavioral characteristics
Here, we document the emergence of civil society as a key theme in a corpus of papers (27,436 papers in total) published in major economics journals in the UK and US between 1900 and 2014 (Fig. 2). To represent this change, we use topic modeling. Topic modeling is a probabilistic machine learning technique that treats a corpus of observed data (documents in Figure 2) as resulting from a hidden data generation process and infers its structure (Ambrosino et al. 2018, Ash et al. 2022, Blei et al. This model provides answers to the following questions: What is the distribution of the words that make up each document in the corpus? What kind of thematic structure is most likely to generate a hypothesis?
Figure 2 Corpus: 27,436 full texts of research articles by journal and year
Using variational Bayesian methods, we generated 100 topics (vectors of words whose weights measure the relative frequency of use) from the literature by modeling the corpus in Figure 2. Topics are generated by an unsupervised learning algorithm. We named them manually based on the most weighted terms. Technical details and rationale for the key assumptions of topic modeling can be found in Bowles and Carlin (2020).
The lens provided by the topic models trained on this corpus records the changing balance of attention given to themes related to the state, market, and civil society, dividing a subset of 100 topics into one “ group into “metatopics”. The procedure for assigning a topic to one of three metatopics determines whether the most weighted words within a topic are commonly associated with any of the motives or mechanisms characteristic of the three vertices. That was it. Topics whose main terms were unrelated to the three systems of social governance (including topics related to mathematical or econometric methods shown in Figure 3) were not assigned to the apex.
As expected, the subtopics that make up the metatopic are clustered in the data. Across the documents in the corpus, if subtopics i, j, and h are two subtopics of the same metatopic (i and j) and a subtopic that is not in the same metatopic (h), then the weights of i and j are positively correlated. On the other hand, the weights of i and j are negatively (weakly) correlated with the weight of h (Appendix 2). Figure 3, which presents our results, confirms the growth and decline of mathematical topics and the established increased focus on empirical methods since mid-century (Angrist et al. 2017).
Figure 3 Topics in mathematical theory and econometrics since 1900
Note: The weights shown are the relative frequencies of the works in the corpus of each period sampled based on the topics shown. The early period weights reflect the early stages of mathematical and econometric methods at the time. Topic numbers shown in parentheses are optional.
A word cloud showing the frequency of words in the 100 topics and the 10 papers that featured the most weighted terms in each of the three metatopics is shown online in Appendix 3. Figure 4, which shows a word cloud of these metatopics, illustrates the following idea. Its contents. For example, important market players such as businesses, banks, and consumers are prominent in market metatopics.
Figure 4 Word cloud showing the most frequently used terms in the three metatopics, from left to right: state, market, civil society
To illustrate, examples of papers that focus on the topics that make up the three metatopics are: For markets, see Bain (1970) “Changes in Concentration in U.S. Manufacturing, 1954–1966” and Klemperer (1992) “Equilibrium Product Row”; for civil society see Fudenberg and Tirole (1990) “Agency “Moral hazard and renegotiation in contracts” and Fehr et al. (2007) “Fairness and contract design”; for states, see Saez (2001) “Using Elasticities to Derive Optimal Income Tax Rates” and Feldstein (1978) “Welfare Costs of Capital Income Taxation.”
Transformation to civil society
Our topic model allows us to locate specific texts in the state space of Figure 1, including the entire studied corpus of different time periods shown in Figure 5.
Figure 5 Economics research: shift to civil society themes since the 1970s
From 1900 to 1970, despite the growing economic importance of the state, there was a significant movement in economic research from state-related topics to market topics. This change was slowed, but not reversed, by the war economy of the early 1940s. The period after 1970 saw an equally large and uninterrupted shift towards civil society topics, primarily at the expense of market topics.
Figure 6 (top panel) shows that before 1970, the transition from near the apex of the state to the apex of the market resulted in a dramatic decline in the weight of the themes we labeled “public regulation” and “international trade organizations.” It shows that it is due to what you did. The weight of two topics related to market structure has increased significantly. Figure 6 (bottom panel) shows that the transition to civil society over the past half century has been primarily a reversal of an earlier increase in the weight of these two market structure topics, as well as asymmetric information, strategic interactions, and asymmetry. This indicates that this is due to an increase in topics related to Behavioral foundations of economics.
Figure 6 Components of the shift from state studies to market topics before 1970 (top panel) and to civil society topics after 1969 (bottom panel)
discussion
The evidence in Figure 3 suggests that the transition from states to markets is associated with an increased use of mathematics, especially constrained optimization by (mostly) non-strategic actors. This amounted to a change in method, namely Alfred Marshall’s mathematical formulation, rather than a novelty in the type of economic behavior studied.
Along with advances in quantitative methods, changes in methods, including limited information and models of strategic interactions, also characterized the subsequent shift towards civil society themes in research. However, unlike earlier methodological changes, research after 1970 was also characterized by an expansion of the concept of economic behavior studied.
Since the late 1960s, this has been done by applying traditional economic tools to an expanded set of themes, such as Gary Becker’s (1974, 1968) work on families and crime, and by drawing from other disciplines. It happened by applying new insights. The work of George Akerlof (1982) and co-authors (1982) on gift exchange and cognitive dissonance, and the work of Oliver Hart (1989, 1995) in traditional economics, About authority and the exercise of power by owners and managers.
References are available in original.
