Paper: arXiv 2511.13384

Authors: Catalin Dumitrescu

Abstract

This study explores the potential impact of introducing a Central Bank Digital Currency (CBDC) on financial stability in an emerging dual-currency economy (Romania), where the domestic currency (RON) coexists with the euro. It develops an integrated analytical framework combining econometrics, machine learning, and behavioural modelling. CBDC adoption probabilities are estimated using XGBoost and logistic regression models trained on behavioural and macro-financial indicators rather than survey data. Liquidity stress simulations assess how banks would respond to deposit withdrawals resulting from CBDC adoption, while VAR, MSVAR, and SVAR models capture the macro-financial transmission of liquidity shocks into credit contraction and changes in monetary conditions. The findings indicate that CBDC uptake (co-circulating Digital RON and Digital EUR) would be moderate at issuance, amounting to around EUR 1 billion, primarily driven by digital readiness and trust in the central bank. The study concludes that a non-remunerated, capped CBDC, designed primarily as a means of payment rather than a store of value, can be introduced without compromising financial stability. In dual currency economies, differentiated holding limits for domestic and foreign digital currencies (e.g., Digital RON versus Digital Euro) are crucial to prevent uncontrolled euroisation and preserve monetary sovereignty. A prudent design with moderate caps, non remuneration, and macroprudential coordination can transform CBDC into a digital liquidity buffer and a complementary monetary policy instrument that enhances resilience and inclusion rather than destabilising the financial system.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced econometric and mathematical modeling including VAR, MSVAR, SVAR, and agent-based simulations, indicating high mathematical complexity. It is grounded in data-driven stress testing with calibrated parameters, behavioral modeling, and scenario analysis, showing substantial empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Assess CBDC impact on financial stability<br>in an emerging dual-currency economy (Romania)"] --> B["Integrated Analytical Framework<br>Combining Econometrics, ML & Behavioural Modelling"]
  B --> C["Data & Inputs<br>Behavioural & Macro-financial Indicators"]
  C --> D["Computational Processes<br>CBDC Adoption Probabilities (XGBoost / Logistic Regression)<br>Liquidity Stress Simulations<br>Macro-Financial Transmission (VAR / MSVAR / SVAR)"]
  D --> E["Key Findings & Outcomes<br>Moderate CBDC uptake (EUR 1bn)<br>Driven by trust & digital readiness<br>Stable design: Non-remunerated, capped, payment-focused<br>Differentiated limits preserve monetary sovereignty"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes an integrated analytical framework that combines econometrics, machine learning (XGBoost and logistic regression) and behavioural modelling to simulate CBDC adoption and liquidity stress in a dual‑currency economy.

Source sentence
It develops an integrated analytical framework combining econometrics, machine learning, and behavioural modelling. CBDC adoption probabilities are estimated using XGBoost and logistic regression models trained on behavioural and macro-financial indicators rather than survey data.

Assumes 6

  • Synthetic agents are calibrated to reflect the observed behavioural traits and demographic composition of Romania’s eligible adult population. data
    Source sentence
    These proportions were incorporated into the design of the synthetic agent to ensure that the frequency of key behavioural traits in the modelled population closely reflects observed characteristics in Romania’s financial and demographic landscape. Taken together, these calibration steps support the conclusion that the synthetic dataset is representative of the Romanian eligible adult population, estimated at
  • Trust in the central bank follows a normal distribution with mean 0.75 and standard deviation 0.1 in the uncertainty layer simulations. statistical
    Source sentence
    CBDC Stress Test in a Dual-Currency Setting | Catalin Dumitrescu 216 - Uncertainty Layer Example: Suppose trust in the central bank is typically distributed with a mean of 0.75 and a standard deviation of 0.1.
  • Holding caps for digital RON and digital euro are necessary to prevent uncontrolled euroisation and preserve monetary sovereignty. other
    Source sentence
    In dual currency economies, differentiated holding limits for domestic and foreign digital currencies (e.g., Digital RON versus Digital Euro) are crucial to prevent uncontrolled euroisation and preserve monetary sovereignty.
  • Liquidity adjustment mechanisms have fixed capacities and costs as listed, and the total withdrawal amount equals exactly RON 54.6 bn. model
    Source sentence
    Each mechanism’s “Utilised” amount is calibrated so that the total equals exactly 54,600 million RON. The “Cost (%)” represents the effective annualised cost or loss rate associated with using that source.
Show 2 more
  • The XGBoost model does not overfit because the data are clean, trend‑normalised and validated with 5‑fold cross‑validation. model
    Source sentence
    High Accuracy ≠ Overfitting - If the Data is Clean and Structured In this study, the data were meticulously pre-processed and denoised, ensuring that only pronounced, deterministic patterns were captured by the models. However, the 5-fold cross-validation used in this study enabled model evaluation across different subsets of the data.
  • Deposit structures and historical liquidity shocks observed in Romania can be used to construct realistic stress scenarios. data
    Source sentence
    Using detailed modelling frameworks based on observed deposit structures, historical liquidity shocks, and stylised intervention tools, we develop three main scenarios: (1) a domestic shock from digital RON adoption, (2) an external shock from digital euro adoption by residents, and (3) a combined stress event involving both.

Data

  • Synthetic population of 10,000 European agents synthetic public
    Romanian eligible adult population (~7 million)
    Source sentence
    Behavioural Consistency Rules for Synthetic Agent Construction This subsection provides a detailed account of how empirically observed relationships among key behavioural enablers were translated into deterministic rules to guide the creation of a dataset of 10,000 synthetic European agents.
  • Behavioural and macro‑financial indicators for XGBoost training macro public
    Source sentence
    CBDC adoption probabilities are estimated using XGBoost and logistic regression models trained on behavioural and macro-financial indicators rather than survey data.
  • European Central Bank (ECB) SPACE, HFCS, Eurostat digital skills datasets · ECB / Eurostat macro public
    Source sentence
    Data Sources and Empirical Basis The rules outlined below are based on the following primary data sources: • European Central Bank (2024) – Study on Payment Attitudes and the Consumer Experience (SPACE). • European Central Bank Household Finance and Consumption Survey (HFCS), Wave 4 (2023). • Eurostat digital skills,
  • Observed deposit structures and historical liquidity shocks in Romania market-prices public
    Romanian banking system
    Source sentence
    Using detailed modelling frameworks based on observed deposit structures, historical liquidity shocks, and stylised intervention tools, we develop three main scenarios: (1) a domestic shock from digital RON adoption, (2) an external shock from digital euro adoption by residents, and (3) a combined stress event involving both.

Costs

Unclear - sources disagree The platform classifier and the paper-text extraction give different answers, so no value is asserted.

Platform classifier: Not applicable

Paper text extraction: Modeled - Each liquidity source has an associated annualised cost percentage as shown in the cost column of the liquidity adjustment mix.

Source sentence (the authors' words)
The “Cost (%)” represents the effective annualised cost or loss rate associated with using that source.

Finds 5

  • CBDC uptake at issuance is projected to be moderate, amounting to around EUR 1 billion, driven mainly by digital readiness and trust in the central bank. predictive-accuracy
    Source sentence
    The findings indicate that CBDC uptake (co-circulating Digital RON and Digital EUR) would be moderate at issuance, amounting to around EUR 1 billion, primarily driven by digital readiness and trust in the central bank.
  • The XGBoost adoption model achieves a classification accuracy of 99.4 % with a Mean Squared Error of 0.006. performance
    Source sentence
    Methodological Strengths of the Model The XGBoost model used in the CBDC adoption simulation framework achieved high predictive accuracy, with a classification accuracy of 99.4% and a Mean Squared Error (MSE) of 0.006.
  • A liquidity outflow of RON 54.6 bn is comparable to the entire banking system’s excess liquidity in calm periods, implying an overnight shift from surplus to deep deficit. statistical
    Source sentence
    In our scenario, the scale of the outflow (RON 54.6 billion) is roughly comparable to the entire banking system’s excess liquidity during calmer periods, implying that banks would shift from surplus to a deep liquidity deficit overnight.
  • Differentiated holding limits for digital RON versus digital euro are crucial to prevent uncontrolled euroisation and preserve monetary sovereignty. theoretical
    Source sentence
    In dual currency economies, differentiated holding limits for domestic and foreign digital currencies (e.g., Digital RON versus Digital Euro) are crucial to prevent uncontrolled euroisation and preserve monetary sovereignty.
  • The synthetic dataset is judged representative of the Romanian eligible adult population of approximately seven million individuals. other
    Source sentence
    Taken together, these calibration steps support the conclusion that the synthetic dataset is representative of the Romanian eligible adult population, estimated at approximately seven million individuals.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 21 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (27)

  1. Poverty Impedes Cognitive Function (2013) — link
  2. Habit and intention in everyday life: The multiple processes by which past behavior predicts future behavior. (1998) — link
  3. MONETARY TRANSMISSION MECHANISM IN CENTRAL AND EASTERN EUROPE: SURVEYING THE SURVEYABLE (2008) — link
  4. The Essential Tension (1977) — link
  5. Network Externality: An Uncommon Tragedy (1994) — link
  6. A new approach to decomposition of economic time series into permanent and transitory components with particular attention to measurement of the ‘business cycle’ (1981) — link
  7. A cross-country analysis of herd behavior in Europe (2014) — link
  8. Predicting Behavior From Actions in the Past: Repeated Decision Making or a Matter of Habit? (1998) — link
and 19 more
  1. Does changing behavioral intentions engender behavior change? A meta-analysis of the experimental evidence. (2006) — link
  2. Contingent Valuation: From Dubious to Hopeless (2012) — link
  3. Sign Restrictions, Structural Vector Autoregressions, and Useful Prior Information (2015) — link
  4. How Habit Limits the Predictive Power of Intention: The Case of Information Systems Continuance1 (2007) — link
  5. Understanding overconfidence: Theories of intelligence, preferential attention, and distorted self-assessment (2015) — link
  6. Fear Of Appreciation (2007) — link
  7. Underestimated Habits: Hypothetical Choice Design in Consumer Research (2017) — link
  8. Structural Vector Autoregressive Analysis (2017) — link
  9. On the equivalence of private and public money (2019) — link
  10. Should Central Banks Issue Digital Currency? (2019) — link
  11. A short comment on statistical versus mathematical modelling (2019) — link
  12. Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy (2005) — link
  13. Assessing the Impact of Central Bank Digital Currency on Private Banks (2020) — link
  14. Demand–Deposit Contracts and the Probability of Bank Runs (2005) — link
  15. The Rise of Digital Money (2021) — link
  16. Time Series Analysis (1994) — link
  17. Fear of Floating (2002) — link
  18. What Triggers Consumer Adoption of Central Bank Digital Currency? (2023) — link
  19. Fintech Adoption Factors: A Study on an Educated Romanian Population (2023) — link

0 of 27 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.