In the realm of economics research, the volume of scholarly work being produced and consumed is continuously expanding. With the proliferation of working-paper repositories and advancements in machine-learning tools such as large language models (LLMs), the process of drafting and revising manuscripts has become more efficient [Ludwig et al., 2025, Humlum and Vestergaard, 2025, Feyzollahi and Rafizadeh, 2025]. However, as the output of economic research grows, the challenge shifts from generating papers to organizing and interpreting vast bodies of literature. This involves deciphering what each paper claims and how these claims interconnect to form mechanisms. It also requires identifying genuine innovations and understanding how incentives influence the production and dissemination of evidence. While there have been advancements in technology that aid in this process, such as LLMs for drafting and revising manuscripts, the primary mode of scholarly communication still relies on narrative prose. Claims are articulated in diverse languages and spread across various sections within papers. This makes it challenging to systematically compare them when manual synthesis is impractical. To address this issue, a novel solution is proposed in the form of paper-level claim graphs. These structured representations map each paper into a directed network comprising standardized economic concepts linked by stated relationships. Nodes correspond to economic concepts aligned with a common ontology (JEL codes), while edges represent connections between source and sink concepts as articulated by authors. Importantly, these edges are annotated with information regarding evidentiary support – whether a relationship is backed by credible causal inference designs or relies on theoretical reasoning, descriptive evidence or correlational analysis. By distilling papers down to their core arguments in a consistent format that can be compared across documents, claim graphs enable scalable comparison, aggregation and analysis. The introduction of this standardized representation holds promise for facilitating various applications beyond those explored in this study. For instance, it could support evidence synthesis at the claim level across different literatures; facilitate structured search and discovery by linking questions, concepts and evidentiary support; as well as provide foundational elements for tracing how research claims are summarized and reframed in policy discussions or media discourse. This innovative approach bridges three distinct strands of literature that are typically studied independently: it complements existing work on the "credibility revolution" within economics by documenting the increasing use of quasi-experimental and experimental designs for empirical identification; it aligns with accounts tracking methodological evolution over time; and it contributes to understanding how quasi-experimental styles have diffused across subfields within economics. In summary, the use of claim graphs in economics research has the potential to greatly enhance our ability to organize, compare, and analyze vast bodies of literature. It also has implications for other fields beyond economics, such as evidence synthesis and structured search. By focusing on measuring causality at the level of relationships rather than broad methodological tags or topic classifications commonly used in previous approaches, claim graphs offer a more nuanced and comprehensive understanding of economic research.
Summary- People are writing a lot about economics using new tools and methods.
- It's hard to understand all the information, so a new idea is to connect papers together based on their main points.
- This makes it easier to compare and analyze the information in a clear way.
- The new method can help organize and understand the literature in economics and other fields.
- It also helps track changes in how research is done.
Definitions
- Scholarly work: Serious academic writing or research done by experts.
- Economics: The study of how money, goods, and services are produced, distributed, and used.
- Repository: A place where things are stored or kept for safekeeping.
- Machine-learning tools: Computer programs that can learn from data and improve over time.
- Claims: Statements made by someone that they believe to be true but need evidence to support them.
- Incentives: Things that encourage or motivate people to do something.
Title: "Revolutionizing Economics Research: The Power of Claim Graphs"
Introduction:
In the world of economics research, the amount of scholarly work being produced and consumed is constantly expanding. With the rise of working-paper repositories and advancements in machine-learning tools, such as large language models (LLMs), the process of drafting and revising manuscripts has become more efficient. However, with this growth comes a new challenge – how to organize and interpret vast bodies of literature.
The Challenge:
As economic research output continues to increase, it becomes increasingly difficult to systematically compare papers and identify genuine innovations. This is due to the fact that claims are articulated in diverse languages and spread across various sections within papers. Manual synthesis is impractical in these cases, making it challenging to understand how different claims interconnect or what evidence supports them.
The Solution:
To address this issue, a novel solution has been proposed – paper-level claim graphs. These structured representations map each paper into a directed network comprising standardized economic concepts linked by stated relationships. Nodes correspond to economic concepts aligned with a common ontology (JEL codes), while edges represent connections between source and sink concepts as articulated by authors.
How It Works:
Claim graphs go beyond traditional methods of organizing literature based on broad methodological tags or topic classifications. Instead, they focus on measuring causality at the level of relationships between economic concepts. This allows for a more nuanced understanding of research claims and their supporting evidence.
Benefits:
By distilling papers down to their core arguments in a consistent format that can be compared across documents, claim graphs enable scalable comparison, aggregation, and analysis. They also have implications for other fields beyond economics – such as evidence synthesis and structured search – by providing foundational elements for tracing how research claims are summarized and reframed in policy discussions or media discourse.
Applications:
The use of claim graphs holds promise for facilitating various applications beyond those explored in this study. For instance:
1. Evidence Synthesis: Claim graphs can support evidence synthesis at the claim level across different literatures, allowing for a more comprehensive understanding of research findings.
2. Structured Search and Discovery: By linking questions, concepts, and evidentiary support, claim graphs can facilitate structured search and discovery in economics research.
3. Tracing Methodological Evolution: Claim graphs contribute to understanding how quasi-experimental styles have diffused across subfields within economics, providing insights into methodological evolution over time.
Conclusion:
In summary, the use of claim graphs in economics research has the potential to greatly enhance our ability to organize, compare, and analyze vast bodies of literature. It also has implications for other fields beyond economics by offering a more nuanced and comprehensive understanding of research claims and their supporting evidence. With this innovative approach bridging three distinct strands of literature – the credibility revolution within economics, methodological evolution over time, and diffusion of quasi-experimental styles – it is clear that claim graphs are revolutionizing the way we approach economic research.
As economics scales, a key bottleneck is representing what papers claim in a comparable, aggregable form. We introduce evidence-annotated claim graphs that map each paper into a directed network of standardized economic concepts (nodes) and stated relationships (edges), with each edge labeled by evidentiary basis, including whether it is supported by causal inference designs or by non-causal evidence. Using a structured multi-stage AI workflow, we construct claim graphs for 44,852 economics papers from 1980-2023. The share of causal edges rises from 7.7% in 1990 to 31.7% in 2020. Measures of causal narrative structure and causal novelty are positively associated with top-five publication and long-run citations, whereas non-causal counterparts are weakly related or negative.
