|
on Risk Management |
| By: | Somnath Chatterjee (Bank of England); David Humphry (Bank of England) |
| Abstract: | The paper presents two risk-based capital frameworks for systemically important European life insurers by drawing a distinction between solvency risk and systemic risk. Solvency risk arises when the value of a life insurer's assets falls below some threshold proportion of its liabilities. To assess solvency risk we implement the Merton-Vasicek portfolio credit risk model and determine capital adequacy of life insurers that correspond to a value-at-risk measure. We measure systemic risk as the expected capital shortfall of an insurer conditional on the overall European life insurance sector being in distress. Our results show that European life insurers have been growing in systemic risk exposure since 2007 and suggest that regulatory capital requirements should account for this. We also find evidence of interconnectedness between systemically important banks and insurance companies, as measured by the transmission of volatility shocks, which increased during periods of financial stress. |
| Keywords: | Solvency risk;systemic risk;insurers;banks;capital shortfall |
| JEL: | C61 C63 G01 G21 G28 |
| Date: | 2026–01–23 |
| URL: | https://d.repec.org/n?u=RePEc:boe:boeewp:023290 |
| By: | Lee, Woongki (Yonsei University) |
| Abstract: | Systemic risk can be measured by the covariance between an institution’s return and the sector return. This allows systemic risk to be decomposed by decomposing the sector return. When the sector return is separated into a market-driven component and a market-orthogonal residual, systemic risk splits into systematic risk, reflecting exposure to the market-driven component, and net systemic risk, reflecting exposure to the residual. We examine how these two risk dimensions relate to financial sector returns, contemporaneously and intertemporally. Using a sample including the 2008 global financial crisis, we find that including both risks explains substantial variation in financial sector returns. |
| Date: | 2026–08–08 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:4qkbe_v1 |
| By: | Wenyuan Li; Haoqi Lyu; Pengyu Wei |
| Abstract: | This article studies a dynamic corporate risk management problem by considering the decision-making of risk-averse managers who exert costly effort and select project risk. We study how a Value-at-Risk (VaR) constraint affects managerial decisions and the distribution of firm value when the manager's objective is non-concave with a fixed salary and options. By the concavification technique, we analyze the optimal terminal firm value on the concave envelope of the objective function. Applying the quantile formulation and the martingale approach, we can derive explicit solutions for optimal effort, terminal firm value, and project choice. The optimal terminal firm value can be divided into nine cases by carefully discussing the choices of VaR floor and tail probability. Compared with the benchmark case, we find that a VaR manager will smooth terminal firm value across states, reducing it in good states while supporting it in adverse states. Moreover, a VaR requirement generally improves downside protection and reduces bankruptcy probability when the VaR floor is low or moderate. However, when the VaR floor is sufficiently high, it can increase bankruptcy probability and induce gambling-for-recovery behavior in adverse states. Our sensitivity analysis indicates that greater managerial effort uniformly improves firm value. Moreover, more incentive options make managers more responsible, leading to a smoother terminal firm value across states. In contrast, a high fixed salary makes the manager less responsible and ultimately causes a more dispersed firm value. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.05623 |
| By: | Ignas Gasparavi\v{c}ius; Andrius Grigutis |
| Abstract: | We show that the key optimization results of the classical Markowitz portfolio selection theory, originally formulated for variance as the risk measure, remain available in explicit closed form under a broader class of strictly convex quadratic risk measures. The proposed framework replaces the covariance matrix with an arbitrary symmetric positive definite matrix and allows additional linear and constant terms, thereby containing various models arising in transaction cost optimization, benchmark relative optimization, covariance regularization, and factor models. Closed-form formulas are obtained for the efficient frontier, the global minimum risk portfolio, the maximum Sharpe ratio portfolio, the Capital Market Curve, the tangency portfolio, and the maximum utility portfolio. In contrast to the classical Markowitz model, the tangency portfolio does not coincide with the maximum Sharpe ratio portfolio, revealing a new geometric phenomenon. A numerical example confirms the derived formulas. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.24449 |
| By: | Lucas Carvalho |
| Abstract: | Hedge ratios, factor models and diversified portfolios all rest on an estimate of which firms move together. That estimate is not stable: firms migrate between the groupings the market treats as coherent, and when enough migrate the organizing axes of the cross-section turn. We measure the rate of that turning as the mean squared sine of the principal angles between subdominant eigenspaces of consecutive twelve-month S&P 500 correlation matrices. A typical month rewrites a fifth of the structure and carries four-fifths forward. That rate is priced: it couples to the aggregate variance risk premium at t = 5.40, no level measure correlates above 0.32, and the implied-correlation surface spans at most 6.7 percent of it. Only the persistent component is priced - the premium compensates the pace of revision, not the distance traveled. The mechanism is prepayment: implied variance rises on impact, volatility follows two to three quarters later (simulated-null p |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.20020 |
| By: | Ryu, Jaehyeon; Yu, Jisang |
| Keywords: | Risk and Uncertainty |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ags:aaea26:404414 |
| By: | Hirbod Assa |
| Abstract: | We develop natural parametric (NatPar) insurance as the natural next step from natural-catastrophe (NatCat) modelling: the same hazard-exposure-vulnerability-finance machinery, with a parametric index made contractual in place of indemnity loss adjustment. Our aim is practical - a standard approach inspired by how the catastrophe-insurance industry already operates, not another optimal-contract criterion. This delivers two payoffs. First, it fixes how reporting is formulated: NatPar contracts are reported in the native NatCat language (annual average loss, EP/AEP/OEP curves, return-period levels), complemented with two-sided basis-exceedance diagnostics (BEP+/-) elevated to the central status the EP curve holds for losses - the canonical distributional view of basis risk, not a supplementary number. Second, the same standard shows how the tail is reallocated between insuree and insurer. A frost case study yields the central result: it is about time, not average. Holding a contract AAL-neutral, the bounded payout cannot follow the unbounded exposure tail, so equalising the mean separates over- and under-payment across return periods: the insuree gains at short horizons while the insurer sheds the deep tail past a crossover of several decades. This reverses under tail dependence - when regions reach extremes jointly, bounded payouts stack and the insurer reabsorbs the deep tail. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.24871 |
| By: | Timothy Besley; Peter John Lambert; Isabelle Michalski-Roland; John Van Reenen |
| Abstract: | This paper examines the impact of credit frictions arising from firm-level default risk on aggregate economic performance. We build a micro-to-macro model with heterogeneous firms and sector-specific production functions, showing that perceived default risk is a sufficient statistic for credit frictions. Using UK administrative data (2004-2019) matched to S&P risk measures, counterfactual estimates reveal that relaxing frictions raises output by 25% and wages by 23%. Ignoring equilibrium wage adjustments overstates output gains, while fixed-capital misallocation approaches understate them. Most gains reflect aggregate capital accumulation. Credit frictions remain above pre-crisis levels, reshape firm size dynamics, increase misallocation across firms, and dampen productivity growth over time. |
| Keywords: | productivity, default risk, credit frictions, misallocation |
| Date: | 2026–07–29 |
| URL: | https://d.repec.org/n?u=RePEc:cep:cepdps:dp2201 |
| By: | Kräussl, Roman; Oladiran, Tobi; Stefanova, Denitsa |
| Abstract: | We examine whether the uncertainty related to environmental, social, and governance (ESG) regulation developments is reflected in asset prices. We proxy the sensitivity of firms to ESG regulation uncertainty by the disparity across the components of their ESG ratings. Firms with high ESG disparity have a higher option-implied cost of protection against downside tail risk. The impact of the misalignment across the different dimensions of the ESG score is distinct from that of the ESG score level itself. Aggregate downside risk bears a negative price for firms with low ESG disparity. |
| JEL: | G12 G18 G32 |
| Date: | 2024–07 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19292 |
| By: | Mohammad Ghaderi; Sébastien Plante; Nikolai Roussanov; Sang Byung Seo |
| Abstract: | Do corporate bond investors earn compensation for bearing credit risk? We construct a new historical corporate bond database spanning 128 years to estimate a corporate bond counterpart to the equity risk premium. Combining hand-collected archival data with modern sources, we assemble a panel of over 100, 000 bonds and 7 million observations. While recent samples suggest corporate bond excess returns largely reflect the term premium, our long sample reveals a sizable and statistically significant credit risk premium. Credit spreads predict future corporate bond returns and macroeconomic aggregates, though their ability to forecast business cycle fluctuations weakens when prewar data are included. |
| JEL: | G1 G12 N21 N22 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35578 |
| By: | Enrico Minnella (Bank of England); Ana Pereira (Bank of England); Eugen Tereanu (European Central Bank and Joint Vienna Institute) |
| Abstract: | This paper develops an agent-based framework (DeTail) to assess the state-contingent tail effects of releasable macroprudential capital buffers. The model features heterogeneous firms, households, and banks, and a single central bank, all interacting in a fully integrated, stock-flow consistent framework which generates endogenous credit cycles. Using this approach, we evaluate how time-varying capital requirements affect the time-varying distributions of credit growth, firm and household default rates, and bank losses along the credit cycle. Policy experiments show that releasing capital buffers during economic downturns preserves credit supply by improving risky (lower-tail) credit outcomes, reduces both households and firms defaults, and supports macro-financial resilience by limiting tail bank losses. At the same time, capital buffer accumulation during upturns imposes minimal costs and does not significantly constrain lending. These findings support the active use of releasable buffers to mitigate systemic risk and smooth credit cycles without weakening the banking system. |
| Keywords: | Agent-based modelling;macroprudential policy;macro-financial linkages;credit cycles;bank resilience;tail risk;state-dependent effects. |
| JEL: | C63 E44 E58 G28 |
| Date: | 2026–07–24 |
| URL: | https://d.repec.org/n?u=RePEc:boe:boeewp:023320 |
| By: | Yang, Liu; Yi, Fujin |
| Keywords: | Risk and Uncertainty |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ags:aaea26:404416 |
| By: | Putri, Alya; Indrajaya, Danang |
| Abstract: | Financial risk management (FRM) has become a strategic function within state-owned energy enterprises exposed to volatile commodity prices, exchange rate fluctuations, and interest rate movements. While a substantial body of literature examines the determinants and firm-level consequences of financial risk exposure and hedging behavior, comparatively little empirical or practice-based work documents how such risk-monitoring functions are operationally structured and executed within large state-owned enterprises (SOEs) in emerging markets. This study addresses that gap through a process-based case study of the Market Research sub-function within the Financial Risk & Insurance Center (FRIC) of PT Pertamina (Persero), Indonesia's national energy company. Using Business Process Modeling (BPM) with BPMN notation and a SWOT analysis supported by Internal Factor Analysis Summary (IFAS) and External Factor Analysis Summary (EFAS) matrices, this study maps the daily workflow of market monitoring and Morning Market Flash Report preparation, identifies operational strengths, weaknesses, opportunities, and threats, and formulates strategic recommendations. The findings indicate that the process exhibits strong internal verification controls (a multi-tier review chain and the use of integrated tools such as Bloomberg Terminal and the FRAME application) but remains constrained by manual data-entry processes and tight pre-market deadlines. Positioned in Quadrant I (Growth/SO) of the SWOT matrix, the function is well placed to pursue an aggressive strategy centered on automation and standardization. The study contributes a practice-grounded account of financial risk monitoring operations that complements the predominantly econometric literature on FRM determinants and offers actionable recommendations for both corporate practice and academic curricula in financial risk education. |
| Date: | 2026–08–15 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:5jxw6_v1 |
| By: | Jorjin F. Godoy (Bangko Sentral ng Pilipinas) |
| Abstract: | This study evaluates the exposure of Philippine industries and conglomerates to systemic risk using SRISK, a metric that measures expected capital shortfall during market downturns. It also examines how macroeconomic indicators—gross domestic product (GDP) growth, the policy rate, the USD–PHP exchange rate, unemployment, and inflation—influence systemic risk through a vector autoregression (VAR) model. The findings show that SRISK was three times higher during the COVID-19 pandemic than during the Global Financial Crisis, rising from 2017 and peaking in Q1 2024. Although its share has declined, the banking sector has remained the largest contributor to systemic risk, while the property sector’s risk contribution has increased since 2018. At the conglomerate level, a widening network of major groups contributes significantly to systemic risk, indicating increasing complexity and potential for contagion. Impulse responses indicate that higher GDP growth, policy rate, and inflation reduce risk in the short term, while increased unemployment and peso depreciation increase risk. Variance decomposition analysis reveals that systemic risk variability is primarily driven by shocks to GDP, followed by the policy rate, unemployment, the USD–PHP exchange rate, and inflation. For the banking sector, self-induced shocks remain the dominant influence |
| JEL: | C22 C23 C53 G20 |
| Date: | 2025–08 |
| URL: | https://d.repec.org/n?u=RePEc:bhd:dpaper:202508 |
| By: | Olivier Wang |
| Abstract: | I decompose stock returns into a duration-matched Treasury component, identified from monetary policy surprises, and a payoff component. Risk and returns rise much less with duration for stocks than for their matched Treasuries. Stock volatility is dampened by rate insurance: rates fall in bad times, so the bond inside a stock provides insurance against the stock’s payoff risk. Expected stock returns are dampened because the insurance works in reverse: rates rise in good times, so stocks’ payoff gains hedge the losses borne by investors holding net duration, notably government bonds when Ricardian equivalence fails. This framework helps reconcile positive bond premia with negative stock-bond covariance, sheds light on equity anomalies and the collapse of the value premium, implies that fiscal and monetary policy shape bond and equity premia, and motivates a two-factor model that jointly prices stocks and bonds. Rate insurance can even turn the price of long-run risk negative, explaining why long bonds beat long stocks. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35636 |
| By: | Jaehyung Choi |
| Abstract: | We develop parametric Entropic Value-at-Risk (EVaR) portfolio optimization for tempered stable L\'evy returns. We derive portfolio cumulant-generating functions and weight-dependent admissible moment-generating-function domains under two multivariate constructions: a multivariate normal tempered stable approach and an independent component factorization. These expressions allow portfolio EVaR to be evaluated from fitted asset- or component-level parameters without repeated portfolio-level distribution fitting. We construct minimum-EVaR portfolios and two entropic reward--risk portfolios. We test the portfolios in a rolling 2000 to 2026 out-of-sample U.S. sector ETF allocation. In this universe, several entropic portfolios have higher realized Sharpe ratios than their matched CVaR portfolios or standard allocation benchmarks. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.18022 |
| By: | Siyuan Sun |
| Abstract: | We present in this article a non-parametric value-at-risk (VaR+CVaR) algorithm that remains accurate for an arbitrarily large number of underlying positions. The algorithm solves the two inherent problems of VaR estimation. First, past history is not directly applicable to the future, but all predictions of the future are based on the past. Second, VaR estimation is equivalent to modeling a single corner of a high-dimensional space (the corner where all bets lose simultaneously). The algorithm only uses mathematical methods that strictly do not degrade in accuracy at high-dimensions. Historical data are then directly incorporated with all high-dimensional relationships present, without manipulation. We test the algorithm with an ensemble of 500 portfolios with random positions across 49 distinct liquid futures of different expiries (VIX, equity indexes, gov. bonds, rates, energy, metals, livestock, agriculture, and softs). All VaR estimations are performed strictly blind to the future. The median portfolio rate of loss exceeding the 99% confidence daily VaR estimate is between $1.0\pm0.1$% depending on algorithm input parameters. 68% of portfolios have a rate of loss exceeding 99% VaR between $1.0\pm0.3$%, and 95% of portfolios between $1.0\pm0.5$%. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.17481 |
| By: | Andrea De Polis (Banco de España); Leonardo Melosi (European University Institute and CEPR); Ivan Petrella (Collegio Carlo Alberto, University of Turin and CEPR) |
| Abstract: | Time-varying asymmetric inflation risks generate persistent stagflationary effects. A quantitative general equilibrium model with time-varying skewness in the distribution of cost-push shocks matches these effects. Central to the analysis is a representation theorem that provides a tractable characterization of a broad class of models with asymmetric shock distributions. The theorem enables a closed-form characterization of optimal monetary policy, according to which the central bank should lean against the balance of inflation risks, while rendering quantitative general-equilibrium models with time-varying risks amenable to counterfactual and scenario analysis. |
| Keywords: | Balance of risks, optimal monetary policy, asymmetric beliefs, policy trade-offs, risk-adjusted inflation targeting, geopolitical risks. |
| JEL: | E52 E31 C53 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:bde:wpaper:2626 |
| By: | Remy Levin; Daniela Vidart |
| Abstract: | We design a method for measuring the risk preferences of agents in the deep past. The method combines a structural model of crop choice as a portfolio allocation with machine-learning prediction of expected crop returns, using historic agronomic and climate data. We estimate county-level risk preferences for the United States and farmer-level preferences in Kansas from 1889 to 1929. More risk averse farmers leveraged less, were less likely to purchase novel WWI Liberty Bonds, and were more likely to participate in local risk-sharing institutions. We show that higher risk aversion predicts slower tractor adoption and farm mechanization during the 1920s. |
| JEL: | D81 G11 N51 N52 O13 Q12 Z10 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35634 |
| By: | Valentin Burban; Pavel Diev; Gilles Dufrénot; Nelson Mongeaud |
| Abstract: | This paper proposes a new empirical taxonomy of safe assets based on their safe-haven behavior during periods of global risk aversion. Our multi-criteria framework captures the persistent performance of bond securities, currencies, and alternative assets during episodes of acute risk-off sentiment, allowing us to construct a cross-asset ranking of safe-haven behavior by asset characteristics: global safe assets, credit-sensitive assets, and emerging assets. We find that sovereign bonds issued by G10 economies, including U.S. Treasuries, consistently exhibit the strongest safe-haven behavior. A limited set of corporate bond markets displays partial safe-asset characteristics, while gold is the only alternative asset that consistently scores highly across our safe-haven criteria, particularly during periods of geopolitical risk. We further show that U.S. Treasuries have exhibited weaker safe-haven properties since the pandemic, although this reflects a broader reconfiguration of global safe-asset hedging properties rather than a uniquely U.S. decline. We find that weaker safe-haven properties are associated with higher inflation, debt levels and scarcity of available assets. |
| Keywords: | Safe Assets, Safe-Haven, U.S. Treasuries, Asset Pricing, Risk Aversion. |
| JEL: | F31 E44 G01 G12 G15 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:bfr:banfra:1049 |
| By: | Zhongtian Chen; Nikolai Roussanov; Xiaoliang Wang; Dongchen Zou |
| Abstract: | We identify a strong common risk factor structure that is pervasive across corporate securities: stocks, corporate bonds, and options. The common factors are closely linked to observable factors and key economic indicators. These factors explain much of the variation in individual asset returns, although pricing errors persist. A joint mean-variance efficient portfolio across asset classes achieves a high Sharpe ratio, resulting in part from cross-market hedging of the common sources of risk. We develop a measure of market segmentation based on differences in the common factor risk premia between markets and document a significant degree of segmentation. |
| JEL: | G1 G12 G13 G17 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35579 |
| By: | Maciej Wysocki |
| Abstract: | This paper develops the first end-to-end application of cross-sectional learning-to-rank to the S&P 500 weekly options (SPXW) zero-day-to-expiration surface, integrated with margin-aware position sizing, an abstention rule driven by model uncertainty, and a strict out-of-time integrity check. A LightGBM LambdaRank ranker scores a daily nine-strategy cross-section composed of eight delta-targeted short-put positions and a \textit{SKIP} candidate, trained against a path-aware Sortino-on-bars label computed at one-minute resolution. The framework is evaluated under index-option margin requirements, a tiered fee schedule, and bid-to-mid execution assumptions across a four-window walk-forward over 2021-2024 and a strictly held-out 2025 out-of-time slice. Seven sizing methods produce out-of-time annualized Sharpe ratios between 4.31 and 5.76, with the headline method reaching a Probabilistic Sharpe Ratio of 0.964 and a sample-period maximum drawdown of -2.28%, on a single hold-out year against a walk-forward range of 1.90 to 3.11. Out of time, every method exceeds three passive benchmarks (CBOE PUT, CBOE WPUT, SPX buy-and-hold) by at least 3.84 in Sharpe ratio and five internal selection baselines by at least 3.69. A two-by-two ablation of the confidence gate against the tail-risk features places 5.05 of the 5.59 out-of-time Sharpe gap over the CBOE PUT with the ranker and the selection layer, the two risk controls adding 0.54 between them. On walk-forward, where the gate binds, neither control comes close to the headline alone and their interaction supplies most of the result. A fifteen-group feature ablation shows that removing the multiplicative regime interactions collapses walk-forward statistical confidence. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.24786 |
| By: | Miquel Noguer i Alonso |
| Abstract: | This paper develops a unified mathematical theory of implied, local, and learned volatility surfaces. Total variance $w_t(k, \tau)=\tau\sigma_t^2(k, \tau)$ is an infinite-dimensional state constrained by positivity, calendar monotonicity, and the butterfly differential inequality. We establish the topology and tangent geometry of this arbitrage set and prove that a nondegenerate Gaussian shock at an active constraint exits with probability tending to one half. Exact invariance therefore requires tangency, reflection, or confinement to an arbitrage-free manifold. We separate this static invariance problem from dynamic no-arbitrage, derive the Musiela maturity-transport identity, and identify the additional fixed-contract martingale restriction. We formulate Hilbert-space dynamics, prove an exact modal reduction with closed-form truncation error, derive Karhunen--Lo\`eve factors, identify the portfolio derivative as a vega field, and obtain the covariance-optimal hedge $\alpha^\ast=(H^\ast C H)^{-1}H^\ast C\nu$. The local-volatility chart completes the geometry: Dupire local variance is the ratio $a=\partial_\tau w/g[w]$ of the calendar and butterfly constraint functionals. Neural operators provide arbitrage-free universal approximation through simplex and cone heads. Normalizing-flow maps then add tractable conditional densities: we derive exact change-of-variables formulae for exponential local-variance flows and invertible stick-breaking price-simplex flows, while stating the quasi-invariance conditions required in genuine function space. Finally, fading-signature fields encode surface history and yield autonomous finite-dimensional controlled dynamics. The result is one framework for representation, dynamics, arbitrage, dimension reduction, likelihood-based learning, simulation, and hedging, together with a falsifiable empirical protocol. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.05198 |
| By: | Yifan Wu; Junjie Lei; Wenjie Huang |
| Abstract: | Risk-aware Q-learning (RaQL) provides a model-free, two-timescale estimator for dynamic risk objectives, but its finite-budget behavior remains fragile: fixed inner-loop hyperparameters can produce unstable value estimates, persistent Bellman residuals, and inefficient sample reuse. This paper proposes an adaptive training controller for Conditional Value-at-Risk (CVaR) RaQL and evaluates it on a daily Bitcoin trading task. The controller preserves the original CVaR estimator and Bellman fixed point; instead, it redesigns the training procedure through six coordinated mechanisms: per-cell inner-step sizing, outer-rate-matched decay synchronization, a short early correction for the VaR-like inner variable, a coverage-first-then-greedy sample allocation rule, progressive suffix aggregation of mature inner estimates, and data-driven calibration of key scales from online-observable quantities. Across 20 random seeds and 856, 000 inner-transition samples, the controller reduces the mean empirical CVaR Bellman residual by approximately 85% relative to the fixed-parameter baseline (MeanBEQ: 1.2202 to 0.1854; MeanBEV: 1.1624 to 0.0535) and maintains stability across CVaR levels, discount factors, and training budgets. On the chronological out-of-sample test set, the learned policy attains a Sharpe ratio of 0.9281 with a maximum drawdown of 6.46% after transaction costs. Although buy-and-hold yields a higher cumulative return (35.43% vs. 23.61%), the adaptive policy achieves far lower volatility (9.57% vs. 47.93%), drawdown, and CVaR loss. These results demonstrate that adaptive finite-budget training design, applied solely to the training procedure without altering the risk objective, can materially improve the reliability and risk-adjusted performance of risk-aware Q-learning in financial applications. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2608.04305 |