Telephone service for the AEA administrative office is temporarily unavailable. For ASSA inquiries, email assa@vanderbilt.edu. For AEA inquiries, email aeainfo@vanderbilt.edu.
If you do not have access to email, please call the publications office at 412-432-2300.
« Back to Results

Economic Models, Explanations, and Performativity

Paper Session

Sunday, Jan. 3, 2027 10:15 AM - 12:15 PM (EST)

Grand Hyatt Washington
Hosted By: International Network for Economic Method
  • Chair: Guilhem Lecouteux, Université Côte d'Azur

Rolling the Dice: Performativity and Its Limits in Integrated Assessment Models

Mathieu Guigourez
,
Université Paris 1 Panthéon-Sorbonne

Abstract

A recent growing body of literature has argued that Integrated Assessment Models (IAMs) were performative; models such as DICE would act as policy advisory tools that enable the mitigation policies they aim to simulate. The paper's central aim is to examine the limits of performativity, by highlighting this discrepancy: although IAMs benefit from considerable political and institutional support, they leave a normative blind spot in their modelling practices.

On this basis, I analyse two contrasting cases: the RAINS model in late 20th-century acid rain regulation, often presented as a success of a performative model-based policymaking, and the DICE model in climate economics, one of the most widely used IAMs to compute the social cost of carbon. I argue that, in both cases, claims of performativity are underdetermined: policy outcomes cannot be straightforwardly attributed to models, even where their influence is widely acknowledged.

Nonetheless, despite their popularity and widespread use by policy experts and decision-makers, climate economists often deny any epistemic responsibility for their modelling practices. The concept of epistemic responsibility has largely gone unquestioned in the climate economics literature, and becomes even more pertinent in light of the potential performativity of IAMs: if these models can influence, at least to some extent, the mitigation scenario they aim to simulate, upon what normative basis can we account for the responsibility of the economist? Accordingly, questioning the epistemic responsibility of a climate economist amounts to questioning their modelling practices, as well as the justifications underlying them, some of which can be blameworthy from a normative perspective, yet get strong political and institutional recognition and legitimacy.

The Cumulativity of Knowledge in the Economics of Science: Modalities, Foundations, and Failures

Arnaud Niedbalec
,
Université Paris 1 Panthéon-Sorbonne
Daniel Zarama Rojas
,
University of Lorraine

Abstract

This article examines how the cumulativity of knowledge has been conceptualized in the economics of science and the economics of knowledge in the early 2000s, before questioning its limits in light of strategies of production of ignorance.
The first section situates this debate within the economic stake it serves: securing a positive relationship between the growth of knowledge, innovation, and economic growth. Cumulativity is conceived as a two-stage process — codification (Cowan & Foray, 1997) followed by diffusion (Rotolo et al., 2022) — and functions as a structuring assumption in endogenous growth models (Mankiw, Romer and Weil, 1992; Aghion and Howitt, 1992). In economics of scientific knowledge, cumulativity rests on an individual disposition toward diffusion and on shared epistemic infrastructures, revealing a tension between a stylized fact and an institutional condition to be constructed — a tension on which intellectual property regimes are grounded.

The second section offers a philosophy-of-science reading of this cumulativity process, drawing on the classical distinction (Sosa, 1980) between foundationalism (the pyramid) and coherentism (the raft). The conception of cumulativity appears largely foundationalist, yet this reading is complicated by the never-fully-completed nature of codification and by the tacit background competencies required to mobilize any codified knowledge — features that call for a complementary coherentist reading.

The third section mobilizes this framework to analyze situations where cumulativity fails, particularly strategies of ignorance production (agnotology). The case of the tobacco industry (Proctor, 2012) shows how such strategies act directly on the conditions of codification (paradigmatic fragmentation) and diffusion (strategic publication without appropriation), turning informational overabundance into an obstacle to knowledge.

Models, Realities, and Expertise in Economics: An Analysis Through the Lens of Performativity

Celso Pereira Neris Jr.
,
São Paulo State University
José Ricardo Fucidji
,
State University of Campinas

Abstract

This paper examines three themes at the intersection of contemporary economic methodology: the persistence of economic models as theoretical instruments; the role of quantification in modelling practice; and the challenges to the legitimacy of economic experts. A common thread connects these themes: the ways in which economics, as a scientific discipline, constructs and transforms social reality. Drawing on the performativity thesis and critical realism, we argue that the interpretive flexibility of economic models, when detached from empirical validation, can catalyse performative processes with serious implications for the quality of scientific enquiry and public trust in economic expertise.

First, we address economic models as tools for explanation. We contrast the view that theoretical economics is defined by an open, unifying conceptual framework with the critical claim that, without proper qualifications, such flexibility minimises the role of empirical evidence. Relying on analytical modelling to protect pre-existing frameworks rather than investigating real-world phenomena obstructs scientific progress.

Secondly, we explore the relationship between models and reality using an interactive approach to performativity. Models do not operate in isolation: when applied through economic institutions, they act as cognitive devices that influence decision-makers' views. When presented as neutral formal structures, their performative capacity is difficult to identify and contest.

Thirdly, we examine the influence of economists as experts, emphasising the risk of 'numerical fatalism': the use of quantification to make fundamentally political processes appear natural. By prioritising aggregate indicators over qualitative, local and personal experience, economic expertise has contributed to public disengagement.

Ultimately, requiring economic theory to be expressed through axiomatised mathematical models is not a neutral methodological choice, but a normative intervention with performative consequences. It is essential to restrict interpretive flexibility through firm empirical evidence and recognise that mathematical formalisation is not value-free if economics is to fulfil its societal functions responsibly.

Discrimination: Theory and Measurement

Florence Adams
,
University of Cambridge
Issa Kohler-Hausmann
,
Yale University

Abstract

The economic approach to measuring discrimination is organized around the supposition that to prove discrimination researchers must identify what caused a decision. If race or sex caused an unequal outcome, then it is discriminatory; if it did not, then it is not. This supposition conceals an ambiguity: do empirical methods identify discrimination by isolating the causal effect of race or sex exclusive or inclusive of the causal effect of credentials? The source of the ambiguity lies in the very thing econometric approaches most want to avoid making assumptions about: the psychological mechanism underlying discriminatory decisions. Most empirical papers present the decision function in reduced form and make no additional assumptions about the functional form of the various inputs. So, nothing in the formal notation or experimental setup commits researchers to an assumption we call (borrowing from Matzkin 2013) separability. Separability is an assumption about whether the inputs into a causal system operate through independent causal channels in the data-generating units (in this case human decisionmakers). Only if separability obtains can researchers detect discrimination by isolating the effect of group perception from the effect of credentials. If separability does not obtain, then economic tests of discrimination are doing something very different: they are producing evidence of discrimination by showing that outcomes caused by a decisionmaker do not match a pattern the decisionmaker is morally, legally or otherwise normatively obliged to produce. On this interpretation, the purpose of the conditioning set is cannot be to isolate the effect of race or sex; it can only be there to stipulate a normative standard for comparability. The upshot is that economists must either explicitly embrace separability and define discrimination as decisions caused exclusively by race or sex or explicitly reject it and defend their chosen comparability standard on normative grounds.

Discussant(s)
Florence Adams
,
University of Cambridge
Celso Pereira Neris Jr.
,
São Paulo State University
Mathieu Guigourez
,
Université Paris 1 Panthéon-Sorbonne
Arnaud Niedbalec
,
Université Paris 1 Panthéon-Sorbonne
JEL Classifications
  • B4 - Economic Methodology
  • D8 - Information, Knowledge, and Uncertainty