Pegs, Floats, and Forests: A Machine Learning Revisit of Exchange Rate Regimes and Growth in Transition Economies

Marjan Petreski

arxiv: 2605.17391 · v1 · pith:6AYZYT5Knew · submitted 2026-05-17 · 💰 econ.GN · q-fin.EC

Pegs, Floats, and Forests: A Machine Learning Revisit of Exchange Rate Regimes and Growth in Transition Economies

Marjan Petreski This is my paper

Pith reviewed 2026-05-19 22:59 UTC · model grok-4.3

classification 💰 econ.GN q-fin.EC

keywords exchange rate regimeseconomic growthtransition economiesrandom forestinstitutional qualitypanel datamachine learningcredibility

0 comments

The pith

Intermediate exchange rate regimes reduce growth by 1 to 10 percentage points versus fixed pegs in transition economies, with the largest shortfalls where institutions are weakest.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper combines standard panel regressions with random forest machine learning to examine how exchange rate regimes affect economic growth across 27 transition economies from 1991 to 2019. It shows that intermediate regimes deliver consistently lower growth than fixed arrangements, while floating regimes produce smaller and mostly insignificant shortfalls. The analysis further demonstrates that this intermediate-regime penalty grows sharper precisely in countries with weaker institutions and is concentrated in the early, pre-2003 stabilization years, disappearing among EU members as institutional convergence advanced. The machine learning results non-parametrically confirm and refine the parametric estimates, highlighting institutional capacity as the key conditioner of whether exchange rate anchoring produces growth benefits.

Core claim

By pairing fixed-effects and system-GMM panel estimates with random forest variable-importance metrics on the Couharde-Grekou probabilistic classification, the study establishes that intermediate exchange rate regimes impose growth penalties of 1.0 to 10.4 percentage points relative to fixed regimes, while floating regimes show negative but largely insignificant differentials; these penalties are largest in weak-institutional settings, are concentrated before 2003, and vanish among EU members, indicating that the credibility dividend of exchange rate anchoring was tied to the formative phase of transition and to limited institutional strength.

What carries the argument

The interaction between exchange rate regime type and institutional quality, isolated non-parametrically by random forest variable importance after parametric panel controls.

If this is right

Exchange rate anchoring yields measurable growth benefits mainly during the initial stabilization phase of transition.
Countries with stronger institutions can safely adopt more flexible regimes without sacrificing growth.
The regime-growth link weakens as institutional convergence with the EU advances.
Machine learning variable importance can serve as a robustness check that enriches causal claims from conventional panel methods.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Policymakers in other emerging markets may face similar trade-offs when institutional capacity is still developing.
Future work could test whether the same institutional conditioning appears in non-transition emerging economies.
If institutions continue to improve, the historical advantage of fixed regimes may continue to erode.

Load-bearing premise

The Couharde-Grekou probabilistic synthesis correctly classifies exchange rate regimes without systematic error that would distort measured growth differences across regime types or institutional conditions.

What would settle it

Finding that, in an updated panel that includes post-2019 data or additional transition economies, intermediate regimes no longer show a growth penalty once institutional quality is controlled for, or that random forest importance rankings place institutions below other covariates.

read the original abstract

This paper combines traditional panel econometrics with random forest machine learning to revisit the relationship between exchange rate regimes and economic growth for 27 transition economies over 1991-2019. Exploiting the Couharde-Grekou (2024) probabilistic synthesis classification, the random forest approach non-parametrically confirms and sharpens what fixed-effects and system GMM estimation establish parametrically intermediate exchange rate regimes consistently underperform fixed arrangements, with growth penalties ranging from -1.0 to -10.4 percentage points, while floating regimes show negative but largely insignificant differentials. Beyond regime effects, the machine learning analysis reveals that the intermediate regime penalty is sharpest precisely where institutions are weakest - non-parametric validation that institutional capacity, not regime label alone, determines whether exchange rate anchoring pays off. The regime-growth relationship is further concentrated in the pre-2003 stabilization era and is absent among EU member economies, suggesting the growth dividend from exchange rate anchoring eroded as institutional convergence advanced. Together, these findings demonstrate how machine learning variable importance metrics can corroborate and enrich causal inference from panel methods, while supporting the view that exchange rate anchoring carried a meaningful credibility dividend during the formative phase of transition.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper combines panel methods with random forests to show intermediate exchange rate regimes carry larger growth penalties in transition economies when institutions are weak, with the effect fading after early stabilization, though the ML step adds limited independent confirmation.

read the letter

The main thing to know is that this paper finds intermediate exchange rate regimes underperform fixed ones by 1 to 10 percentage points in 27 transition economies from 1991 to 2019, with the penalty largest where institutions are weakest, and the pattern clearest before 2003 and outside EU members. Floating regimes show smaller and mostly insignificant differences. The random forest step is used to highlight the institutional interaction without a preset functional form. The authors draw on the Couharde-Grekou probabilistic classification and run both fixed-effects regressions and system GMM as the core estimates. The ML results are presented as sharpening and confirming the parametric findings on the interaction term. What stands out is the time split and the EU distinction, which give some sense of when anchoring delivered a credibility boost during the initial transition phase. The dual-method approach is a reasonable way to check whether the institutional moderation holds up beyond linear assumptions. The soft spot is the random forest implementation. Standard versions treat observations as independent and do not build in country or time fixed effects or correct for regime endogeneity the way GMM does. If time-varying unobservables such as reform waves or political shifts drive both institutional quality and regime choice, the variable importance metrics could pick up correlated patterns rather than isolate a clean interaction. The paper treats the ML output as corroboration, but without panel adaptations described, it functions more as a descriptive supplement than a fully independent robustness check. This is for economists working on exchange rate policy and growth in emerging or transition settings who already know the basic literature. It refines an existing question with a modest methodological addition rather than opening new ground. The evidence is coherent enough on its own terms to merit referee time, even if the ML details will need scrutiny. I would send it for peer review.

Referee Report

2 major / 2 minor

Summary. The paper combines fixed-effects and system-GMM panel regressions with random-forest machine learning on 27 transition economies (1991-2019), using the Couharde-Grekou (2024) probabilistic regime classification. It reports that intermediate regimes carry growth penalties of -1.0 to -10.4 percentage points relative to fixed regimes, with the penalty largest where institutions are weakest; floating regimes show smaller and mostly insignificant effects. The relationship is concentrated pre-2003 and absent among EU members, with the random-forest step presented as non-parametric corroboration of the institutional interaction.

Significance. If the results hold after addressing panel-structure concerns, the manuscript contributes by documenting institutionally contingent effects of exchange-rate regimes during transition and by illustrating how machine-learning variable-importance metrics can enrich panel-based causal inference. The pre-2003 and EU heterogeneity findings provide falsifiable, context-specific predictions that strengthen the credibility-dividend interpretation.

major comments (2)

[Abstract and machine-learning results section] The random-forest analysis (Abstract and the machine-learning results section) treats observations as i.i.d. and does not incorporate country or time fixed effects, clustered standard errors, or any panel-aware adaptation. Because the headline claim that the intermediate-regime penalty is 'sharpest precisely where institutions are weakest' rests on the random forest non-parametrically isolating this interaction, the absence of controls for time-varying unobservables (e.g., reform waves correlated with both institutions and regime choice) makes the reported corroboration of the system-GMM results vulnerable to confounding.
[Data and classification section] No robustness checks are described for alternative regime classifications (e.g., Ilzetzki-Reinhart-Rogoff or de facto measures). Given that the central growth-penalty estimates and the institutional-interaction finding depend on the Couharde-Grekou (2024) probabilistic synthesis, systematic measurement error in regime assignment could bias both the parametric and non-parametric results.

minor comments (2)

[Abstract] The abstract states a penalty range of -1.0 to -10.4 percentage points but does not indicate which institutional threshold or subsample produces the upper bound; adding this detail would improve precision.
[Methods] The manuscript should report the exact random-forest hyperparameters (number of trees, depth, feature sampling) and whether observations were weighted by country or time period.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments, which highlight important issues for robustness. We respond to each major comment below and outline the revisions we will implement.

read point-by-point responses

Referee: [Abstract and machine-learning results section] The random-forest analysis (Abstract and the machine-learning results section) treats observations as i.i.d. and does not incorporate country or time fixed effects, clustered standard errors, or any panel-aware adaptation. Because the headline claim that the intermediate-regime penalty is 'sharpest precisely where institutions are weakest' rests on the random forest non-parametrically isolating this interaction, the absence of controls for time-varying unobservables (e.g., reform waves correlated with both institutions and regime choice) makes the reported corroboration of the system-GMM results vulnerable to confounding.

Authors: We recognize that the random forest treats observations as independent and does not explicitly model panel dependence. The primary identification relies on the fixed-effects and system-GMM specifications, which already incorporate country and time effects. Nevertheless, to strengthen the non-parametric corroboration, we will revise the machine-learning section by including country and year indicators as explicit features in the random forest, report results with clustered standard errors at the country level, and explore a panel-adapted variant (e.g., via within-transformation or mixed-effects random forest). These changes will directly address potential confounding from time-varying unobservables while preserving the non-parametric nature of the exercise. revision: yes
Referee: [Data and classification section] No robustness checks are described for alternative regime classifications (e.g., Ilzetzki-Reinhart-Rogoff or de facto measures). Given that the central growth-penalty estimates and the institutional-interaction finding depend on the Couharde-Grekou (2024) probabilistic synthesis, systematic measurement error in regime assignment could bias both the parametric and non-parametric results.

Authors: We agree that reliance on a single classification warrants explicit robustness checks. In the revised manuscript we will add a new subsection that re-estimates both the panel regressions and the random-forest analysis using the Ilzetzki-Reinhart-Rogoff de facto classification and at least one additional standard measure. We will report whether the magnitude and institutional contingency of the intermediate-regime penalty remain qualitatively unchanged, thereby mitigating concerns about measurement error in regime assignment. revision: yes

Circularity Check

0 steps flagged

No significant circularity; core results from external classification and standard panel methods, ML used only for confirmation.

full rationale

The paper's derivation relies on the external Couharde-Grekou (2024) probabilistic synthesis for assigning exchange rate regimes, followed by independent fixed-effects and system GMM panel estimations to quantify growth penalties. The random forest step is explicitly described as non-parametrically confirming and sharpening those parametric results rather than generating them by construction or fitting inputs that are then relabeled as predictions. No self-definitional loops, load-bearing self-citations, ansatz smuggling, or renaming of known results appear in the abstract or described chain. The analysis remains self-contained against external benchmarks with independent econometric content.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the accuracy of the external regime classification and the assumption that random forest importance metrics can isolate causal moderation by institutions. No free parameters are explicitly fitted in the abstract description; the work is empirical rather than axiomatic.

axioms (1)

domain assumption The Couharde-Grekou probabilistic synthesis provides an exogenous and unbiased measure of de-facto exchange rate regimes.
Invoked when the paper exploits this classification to define the regime variable for both panel and random forest analyses.

pith-pipeline@v0.9.0 · 5743 in / 1503 out tokens · 27105 ms · 2026-05-19T22:59:10.646754+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

random forest estimation is applied to identify which country characteristics most powerfully condition the ERR-growth relationship, without imposing functional form or pre-specifying interactions. The key output for our purposes is the variable importance measure: the mean decrease in prediction accuracy when each variable is randomly permuted across all trees.
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The intermediate regime penalty is sharpest precisely where institutions are weakest — non-parametric validation that institutional capacity, not regime label alone, determines whether exchange rate anchoring pays off.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

11 extracted references · 11 canonical work pages

[1]

Aghion, P., Bacchetta, P., Ranciere, R., and Rogoff, K. (2009). Exchange Rate Volatility and Productivity Growth: The Role of Financial Development. Journal of Monetary Economics, 56(4), 494–513. Ashour, M. and Chen Yong, C. (2018). The Impact of Exchange Rate Regimes on Economic Growth: Empirical Study of a Set of Developing Countries during the Period 1...

work page 2009
[2]

Breiman, L. (2001). Random forests. Machine Learning, 45, 5–32. Bubula, A. and Otker-Robe, I. (2002). The Evolution of Exchange Rate Regimes Since 1990: Evidence from de Facto Policies. IMF Working Paper No. 02/155, International Monetary Fund, Washington, DC. Calvo, G. and Reinhart, C.M. (2002). Fear of Floating. Quarterly Journal of Economics, 117(2), 3...

work page 2001
[3]

Dean, J.W. (2003). Exchange Rate Regimes in Central and Eastern European Transition Economies. Simon Fraser University, Burnaby, British Columbia. De Grauwe, P. and Schnabl, G. (2004). Exchange Rate Regimes and Macroeconomic Stability in Central and Eastern Europe. CESifo Working Paper No

work page 2003
[4]

and Zilberfarb, B

Dellas, H. and Zilberfarb, B. (1993). Real Exchange Rate Volatility and International Trade: A Reexamination of the Theory. Southern Economic Journal, 59(4), 641–

work page 1993
[5]

Dixit, A. (1989). Hysteresis, Import Penetration, and Exchange Rate Pass-Through. Quarterly Journal of Economics, 104(2), 205–228. Domac, I., Peters, K., and Yuzefovich, Y. (2001). Does the Exchange Rate Regime Affect Macroeconomic Performance? Evidence from Transition Economies. Policy Research Working Paper No. 2642, World Bank, Washington, DC. 28 Dubas...

work page 1989
[6]

and Kretschmann, M

Harms, P. and Kretschmann, M. (2009). Words, Deeds and Outcomes: A Survey on the Growth Effects of Exchange Rate Regimes. Journal of Economic Surveys, 23(1), 139–164. Husain, A., Mody, A., and Rogoff, K.S. (2005). Exchange Rate Regime Durability and Performance in Developing Versus Advanced Economies. Journal of Monetary Economics, 52(1), 35–64. Ilzetzki,...

work page 2009
[7]

Khan, M., Kebewar, M., and Nenovsky, N

Elsevier, Amsterdam, 91–145. Khan, M., Kebewar, M., and Nenovsky, N. (2013). Inflation Uncertainty, Output Growth Uncertainty and Macroeconomic Performance: Comparing Alternative Exchange Rate Regimes in Eastern Europe. MPRA Paper No. 45523, University Library of Munich. Kuokštis, V., Asali, M. and Spurga, S.A., (2025). How labor market institutions influ...

work page 2013
[8]

and Sturzenegger, F

Levy-Yeyati, E. and Sturzenegger, F. (2005). Classifying Exchange Rate Regimes: Deeds vs. Words. European Economic Review, 49(6), 1603–1635. Moreno, R. (2000). Pegging and Macroeconomic Performance in East Asia. ASEAN Economic Bulletin, 18(1), 48–63. Moreno, R. (2001). Pegging and Stabilization Policy in Developing Countries. Economic Review of the Federa...

work page 2005
[9]

Rogoff, K., Husain, A., Mody, A., Brooks, R., and Oomes, N. (2003). Evolution and Performance of Exchange Rate Regimes. IMF Working Paper No. 03/243, International Monetary Fund, Washington, DC. Sachs, J.D. (1996). Economic Transition and the Exchange Rate Regime. American Economic Review, 86(2), 147–152. Schonlau, M., & Zou, R. Y. (2020). The random fore...

work page 2003
[10]

Variables’ names, descriptions and sources Variable Description Source Dependent variable GDP per capita growth Annual growth rate of GDP per capita (%) World Development Indicators Growth regression controls Initial GDP (log) Log of GDP per capita in 1990; captures conditional convergence World Development Indicators Government consumption General govern...

work page 1990
[11]

Certainty index = max(P(Fixed), P(Intermediate), P(Floating))

Exchange Rate Regime Distribution in Transition Economies, 1991–2019: probabilities and classification certainty Group / Sub-period P(Fixed) P(Intermediate) P(Floating) Certainty index N All transition economies 1991–1999 0.173 0.501 0.326 0.859 149 2000–2007 0.344 0.335 0.321 0.883 211 2008–2013 0.349 0.353 0.298 0.908 156 2014–2019 0.408 0.393 0.198 0.9...

work page 1991

[1] [1]

Aghion, P., Bacchetta, P., Ranciere, R., and Rogoff, K. (2009). Exchange Rate Volatility and Productivity Growth: The Role of Financial Development. Journal of Monetary Economics, 56(4), 494–513. Ashour, M. and Chen Yong, C. (2018). The Impact of Exchange Rate Regimes on Economic Growth: Empirical Study of a Set of Developing Countries during the Period 1...

work page 2009

[2] [2]

Breiman, L. (2001). Random forests. Machine Learning, 45, 5–32. Bubula, A. and Otker-Robe, I. (2002). The Evolution of Exchange Rate Regimes Since 1990: Evidence from de Facto Policies. IMF Working Paper No. 02/155, International Monetary Fund, Washington, DC. Calvo, G. and Reinhart, C.M. (2002). Fear of Floating. Quarterly Journal of Economics, 117(2), 3...

work page 2001

[3] [3]

Dean, J.W. (2003). Exchange Rate Regimes in Central and Eastern European Transition Economies. Simon Fraser University, Burnaby, British Columbia. De Grauwe, P. and Schnabl, G. (2004). Exchange Rate Regimes and Macroeconomic Stability in Central and Eastern Europe. CESifo Working Paper No

work page 2003

[4] [4]

and Zilberfarb, B

Dellas, H. and Zilberfarb, B. (1993). Real Exchange Rate Volatility and International Trade: A Reexamination of the Theory. Southern Economic Journal, 59(4), 641–

work page 1993

[5] [5]

Dixit, A. (1989). Hysteresis, Import Penetration, and Exchange Rate Pass-Through. Quarterly Journal of Economics, 104(2), 205–228. Domac, I., Peters, K., and Yuzefovich, Y. (2001). Does the Exchange Rate Regime Affect Macroeconomic Performance? Evidence from Transition Economies. Policy Research Working Paper No. 2642, World Bank, Washington, DC. 28 Dubas...

work page 1989

[6] [6]

and Kretschmann, M

Harms, P. and Kretschmann, M. (2009). Words, Deeds and Outcomes: A Survey on the Growth Effects of Exchange Rate Regimes. Journal of Economic Surveys, 23(1), 139–164. Husain, A., Mody, A., and Rogoff, K.S. (2005). Exchange Rate Regime Durability and Performance in Developing Versus Advanced Economies. Journal of Monetary Economics, 52(1), 35–64. Ilzetzki,...

work page 2009

[7] [7]

Khan, M., Kebewar, M., and Nenovsky, N

Elsevier, Amsterdam, 91–145. Khan, M., Kebewar, M., and Nenovsky, N. (2013). Inflation Uncertainty, Output Growth Uncertainty and Macroeconomic Performance: Comparing Alternative Exchange Rate Regimes in Eastern Europe. MPRA Paper No. 45523, University Library of Munich. Kuokštis, V., Asali, M. and Spurga, S.A., (2025). How labor market institutions influ...

work page 2013

[8] [8]

and Sturzenegger, F

Levy-Yeyati, E. and Sturzenegger, F. (2005). Classifying Exchange Rate Regimes: Deeds vs. Words. European Economic Review, 49(6), 1603–1635. Moreno, R. (2000). Pegging and Macroeconomic Performance in East Asia. ASEAN Economic Bulletin, 18(1), 48–63. Moreno, R. (2001). Pegging and Stabilization Policy in Developing Countries. Economic Review of the Federa...

work page 2005

[9] [9]

Rogoff, K., Husain, A., Mody, A., Brooks, R., and Oomes, N. (2003). Evolution and Performance of Exchange Rate Regimes. IMF Working Paper No. 03/243, International Monetary Fund, Washington, DC. Sachs, J.D. (1996). Economic Transition and the Exchange Rate Regime. American Economic Review, 86(2), 147–152. Schonlau, M., & Zou, R. Y. (2020). The random fore...

work page 2003

[10] [10]

Variables’ names, descriptions and sources Variable Description Source Dependent variable GDP per capita growth Annual growth rate of GDP per capita (%) World Development Indicators Growth regression controls Initial GDP (log) Log of GDP per capita in 1990; captures conditional convergence World Development Indicators Government consumption General govern...

work page 1990

[11] [11]

Certainty index = max(P(Fixed), P(Intermediate), P(Floating))

Exchange Rate Regime Distribution in Transition Economies, 1991–2019: probabilities and classification certainty Group / Sub-period P(Fixed) P(Intermediate) P(Floating) Certainty index N All transition economies 1991–1999 0.173 0.501 0.326 0.859 149 2000–2007 0.344 0.335 0.321 0.883 211 2008–2013 0.349 0.353 0.298 0.908 156 2014–2019 0.408 0.393 0.198 0.9...

work page 1991