Money, Banks and Finance in Economic Thought

Classifying economics: from codes and topics, to inside boundaries

Edwards José, Universidad Adolfo Ibanez
Noblet Guillaume, Université Paris 1
Delcey Thomas, Université Paris 1 Panthéon-Sorbonne

This project (in progress) begins with the revision of several recent classifications of economics by, namely, Ellison (2002), Kim et al. (2006), Kelly & Bruestle (2010), Hamermesh (2013), Guo et al. (2015), Claveau & Gingras (2016), and Angrist et al. (2017a-b). We emphasize on how prominent the JEL Classification System has become for identifying the different economics subfields (i.e. the inside boundaries of economics). Based on that code-system (Cherrier 2017), classifications by economists tend to identify three large areas: Microeconomic theory, Macroeconomics, and Econometrics, accounting together for over 50% of the academic literature by economists. Is that a pertinent/useful representation of economics? Our project develops from this question in two stages. We first use the EconLit database, to analyze all economics articles published in 2018 exploiting the JEL Classification System (i.e. around 25,000 articles including over 80,000 JEL-codes). However, unlike most recent classifications by economists (which tend to reduce the different codes by articles into unique labels), we take into account every single JEL-code and the relations between them. Using descriptive statistics and network analysis, our procedure allows for creating a rich sorting system, drawing from the economists’ use of JEL-codes as tools for representing their own published work. We then analyze all abstracts from the exact same articles, using a topic modeling algorithm (Latent Dirichlet Allocation, LDA). By modeling the whole dataset into 20 topics (there are also 20 general JEL-codes), we produce a new sorting system, challenging most recent classifications by economists (i.e. those using the JEL-codes). We get twofold results. First: the abstract-rhetoric of economists unveils topics, which seem misrepresented within the JEL-code-system. For instance, an “energy economics” topic shows in about 25% of all abstracts for 2018. Second: by processing our topic-model, we develop a new representation of economics. Drawing from this sorting, we identify new inside boundaries, especially within areas drawing from Microeconomic Theory and Econometrics.


Keywords: JEL Classification System, Topic Modeling, History of Contemporary Economics