Referee Report: Schumpeterian disaggregation and integrated assessment: An endogenous, stock–flow consistent economy in disequilibrium for FRIDA v2.1

Summary

The manuscript presents an ambitious attempt to embed a Schumpeterian, stock–flow consistent, disequilibrium macro‑economy within an integrated assessment modeling (IAM) framework. The conceptual motivation is strong: IAMs typically lack explicit financial dynamics, innovation processes, and macro‑economic instability. The authors aim to fill this gap by developing a macro module capable of generating endogenous cycles, downturns, and financial stress.

However, the manuscript’s central empirical claim, that the model reproduces major historical downturns and macroeconomic cycles, is not supported by the evidence presented. Figures 5-7, which are intended to demonstrate the model’s empirical validity, instead reveal substantial mismatches with observed data. These discrepancies raise fundamental questions about the model’s suitability for long‑run IAM applications, especially given the extensive literature showing the inherent difficulty of predicting business cycles even a few years ahead.

1. Scientific Significance

The conceptual ambition is high, and the integration of Schumpeterian innovation and stock-flow consistent (SFC) accounting into an IAM is potentially valuable. However, the claimed contribution (capturing macroeconomic cycles and major downturns) is not demonstrated.

• Figure 5 shows that the model fails to reproduce the 2009 and 2015 downturns, despite these being among the most significant macroeconomic shocks of the past decades.
• The unemployment simulation moves in the opposite direction of observed unemployment in several periods.
• The simulated labor share lies outside the historical range for 1980-2000.
• Figure 6 shows no meaningful cyclical behavior, only trend‑like projections with wide confidence intervals.
• Figure 7 shows cycles only in the sense that the confidence interval is extremely wide; the median itself exhibits a downward drift.

Given these issues, the manuscript does not yet demonstrate a substantial advance in modeling science for IAMs. Many existing IAMs can reproduce long‑run trends; the novelty claimed here is not supported by the results.

2. Scientific Quality

The modeling framework is internally coherent, but the empirical validation is insufficient and, in some cases, contradictory to the claims.

2.1. Mismatch between claims and results

The manuscript repeatedly asserts that the model “reproduces major downturns” and “captures macroeconomic cycles.” However:

• The model does not reproduce the 2009 or 2015 downturns.
• The unemployment dynamics are counterfactual.
• The labor share simulation is systematically misaligned with historical data.
• No cycle‑focused diagnostics (e.g., year‑on‑year changes, HP‑filtered cycles, turning‑point analysis) are provided.

2.2. Lack of engagement with DSGE and macro‑forecasting literature

There is a large body of work such as DSGE, VAR, and macro‑forecasting studies, demonstrating that:

• Business cycles are extremely difficult to predict, even 1-3 years ahead.
• Models that match long‑run trends often fail to match short‑run dynamics.
• Predicting cycles over 50-100 years is not considered feasible in the macroeconomics literature.

The manuscript needs to acknowledge these fundamental limitations, and justify why this model should be able to do what DSGE models cannot.

2.3. Implications for IAMs

For IAMs, the question is whether this macro block provides reliable, policy‑relevant dynamics. Given the weak empirical performance, it is unclear whether the model adds value beyond simpler trend‑based representations.

3.  Presentation Quality

The manuscript is generally well written, but the presentation of results is not balanced. A more transparent discussion of model limitations would strengthen the manuscript.



4. Improvements

To move toward publication, the authors would need to:

1. Reassess the claim of capturing cycles, given the DSGE literature and the model’s performance.
2. Provide rigorous cycle‑focused validation (year‑on‑year changes, filtered cycles, turning‑point analysis).
3. Discuss why the model misses major downturns and what this implies for long‑run projections.
4. Clarify the intended use case: is the model meant to capture cycles, or long‑run structural trends?
5. Consider reframing the contribution as a conceptual integration rather than an empirically validated cycle‑generating macro module.

This is an impressive SD work worth celebrating.  A strong incrementation for similar integrated models. My main comments are about the presentation and positioning of the model so it can be of use for policymakers, industry and academia. 

1. Please elaborate in the Introduction or appendix clearly contrasting this economy module with previous versions and other climate–finance/ SFC macro‑climate models. Ideally, a small table is desired that lists key dimensions (finance, business cycles, innovation treatment, damage representation, uncertainty) and shows what is new or improved in FRIDA v2.1 relative to those models

2. Section 2 mixes motivations, theory, and trade‑offs  - I would focus a section per each.

3. In my opinion, the model is hard to make sense for a policy maker as it is now. It scares with complexity, and does not involve the user into the experiments. Also model description is quite static (looks like a validation - that you did the job good), at the same time misses holding the hand of the reader - so you can navigable the readership towards detailed reading, not skimming.

 4. You can possibly add a more detailed “model roadmap” at the start of Sec. 3: a one‑paragraph overview of each submodule (Circular Flow, Finance, Innovation, Government, Employment, GDP/Inflation) and a one‑sentence statement of its main role in the overall feedback structure

5. Please consider explicitly reference the main equations/variables from simplified stock‑and‑flow diagrams (Figures 2–5) in the text (e.g. “as shown in Fig. 3, the risk premium depends on X)

6. Elaboration of results is the weakest point of the paper. Given the complexity of the model, please consider adding the key schematic that highlights just the key feedback loops (e.g., climate → productivity → employment → transfers → debt → risk premium) - that would greatly help readers who would like to use findings from the study.

7. Please be clear on the shortcomings and elaborate limitations with more detail. Its currently unclear for me which aspects of the historical series are deliberately not reproduced apart from mentioning 2009, 2020 pandemic. Missing what range of crises this is acceptable for the model’s intended use as an exploratory tool? Say we have pandemic every 5 years from now  - what time range of the model would still be feasible for use?Moreover, I would be surprised if model covered all business cycles related to innovation the model reproduces. Say the AI that is embedded at the large scale in the society? scarcity of rare earth?  Potential wars and conflicts?

8. The current contribution is vague for a policy maker to take an action. Suggest adding one illustrative figure showing the distribution (histogram) of a key outcome (e.g. real GDP in 2100) to make the probabilistic reading more concrete. Make sure the that is improved so readers have an interactive tool to play with. 

9. I would argue there is a need to explain the story by binding variables to macro outcomes and the need for interventions, e.g.: higher STA → higher failure rates → investment slowdown → higher unemployment → more transfers → higher debt → higher risk premia → further slowdown.

10. Consider adding a short comparison of results using the best‑fit parameter set vs the median of the ensemble, to illustrate the role of uncertainty and what can be done to improve the outcome

 11.Ensure terms such as “safe loans”, “exploratory loans”, “performing/nonperforming loans”, and “risk premium” are defined once and used consistently.  Same with other terms that have synonyms in the paper.

12. For Figures 6–8 please consider explicitly stating in each caption what the lines and shaded areas represent (median, 67%, 95%),

Good luck with the revision, I hope my comments will make the paper stronger