{ "release": "2026.10.27-r3", "site_date": "2026-10-27", "authors": { "fuli-yang": { "name": "Fuli Yang", "family": "Yang", "given": "Fuli", "affiliation": "GARCH Institute", "orcid": "0009-0005-4055-7561" }, "tingting-qiu": { "name": "Tingting Qiu", "family": "Qiu", "given": "Tingting", "affiliation": "GARCH Institute; Fujian Vocational College of Agriculture, Fuzhou", "orcid": "0009-0009-2442-5621" }, "yu-liu": { "name": "Yu Liu", "family": "Liu", "given": "Yu" }, "shiyu-lin": { "name": "Shiyu Lin", "family": "Lin", "given": "Shiyu", "affiliation": "GARCH Institute", "orcid": "0009-0000-2471-3644" }, "jing-huang": { "name": "Jing Huang", "given": "Jing", "family": "Huang" } }, "series": { "ai-industrial-cycles": [ "AI & Industrial Cycles", "Productive capacity, investment commitments, and risk across the infrastructure supporting artificial intelligence." ], "japan-macro-finance": [ "Japan Macro-Finance", "Monetary normalization, institutional portfolio adjustment, and yen carry unwinds." ], "distributional-geometry": [ "Distributional Geometry", "Wasserstein methods for measuring market memory, monitoring stress, and examining distributional states." ], "financial-recovery": [ "Financial Crises & Recovery", "Bank equity, credit, and the paths of economic recovery after financial crises." ], "climate-finance": [ "Climate & Financial Information", "Climate-related information, rating resolution, and the measurement of financial risk." ], "ai-financial-systems": [ "AI & Financial Systems", "AI adoption, financial decision-making, learning, market behaviour, and the economics of new technologies." ], "institutions-market-structure": [ "Institutions & Market Structure", "Trading rules, investor protection, institutional coordination, fiduciary services, and market design." ], "volatility-econometrics": [ "Volatility & Financial Econometrics", "Volatility measurement, forecasting, model evaluation, and financial risk." ], "macro-finance": [ "Macro-Finance", "Interest rates, policy uncertainty, cross-market transmission, and economic adjustment." ], "central-bank-communication": [ "Central Banks & Communication", "Monetary policy expectations, central-bank language, voting authority, and financial markets." ], "energy-compute-infrastructure": [ "Energy & Compute Infrastructure", "Compute investment, electricity contracts, financing, energy markets, and infrastructure risk." ], "crypto-roughness": [ "Crypto Roughness & Perpetuals", "Roughness measurement, funding rates, tail risk, and distributional states in crypto perpetual futures." ] }, "papers": [ { "id": "0005", "title": "The Physical Limits of Compute and Financial Risk Integration Across Seven Infrastructure Layers", "authors": [ "fuli-yang", "yu-liu", "shiyu-lin" ], "type": "Preprint", "series": "ai-industrial-cycles", "doi": "10.5281/zenodo.23644877", "date": "2026-10-26", "source": "Zenodo", "url": "https://zenodo.org/records/23644877", "access": "Restricted full text at source", "summary": "Artificial intelligence links seven infrastructure layers into a persistent, uneven equity-risk system. Using 33 firms, a fixed 28-stock comparison and strictly lagged factor adjustment through August 2026, return connectedness rises from 6.34% to 20.06% and downside connectedness from 5.27% to 21.02%. Both increases survive a six-test correction; the residual diversification ratio falls by 16.3%. A traded-factor extension through 23 October retains higher connectivity, with corrected evidence strongest for downside risk. Grid equipment dominates outward variance attribution; nuclear supply and gas remain peripheral. Five supplier–buyer power-agreement announcements and fourteen eligible tariff decisions establish observable contractual links without a uniform increase in residual dependence. Power agreements validate investment while reallocating price, volume, completion and exit exposures. Financial integration therefore spans industry boundaries, while physical links and equity covariance measure different objects. Risk management must assess the infrastructure system and its contractual risk bearers jointly.", "keywords": [ "artificial intelligence", "compute", "infrastructure", "financial risk", "integration" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "author_abstract": true, "related_sources": [], "abstract_source": "Author manuscript" }, { "id": "0004", "title": "Yen Appreciation and the Identification of Carry Unwinds in Japan", "authors": [ "fuli-yang", "yu-liu" ], "type": "Preprint", "series": "japan-macro-finance", "doi": "10.5281/zenodo.23639309", "date": "2026-10-25", "source": "Zenodo", "url": "https://zenodo.org/records/23639309", "access": "Restricted full text at source", "summary": "Yen appreciation is not a sufficient statistic for a carry unwind. Official intervention, monetary repricing, global-dollar movements and portfolio adjustment can generate similar currency outcomes while implying different funding, duration and target-asset exposures. This paper develops a joint cross-market attribution framework and tests its observable return restrictions using Japanese intervention dates, historical USD/JPY paths, JGB changes, intraday prices and yen-funded AUD, MXN and BRL cash positions. An exact decomposition separates common yen valuation from target-dollar movements and interest accrual. Across ten operation dates in five episodes, twenty-observation average cash payoffs are −1.034%, −1.234% and −1.917%, respectively. The common yen component is −1.449 percentage points, but target components and short-horizon payoff signs differ. Excluding one episode can reverse the AUD and MXN twenty-observation means. Daily and intraday evidence further shows that the return baseline and market clock determine which response is measured. The central conclusion is that exchange-rate-only attribution cannot recover a carry liquidation or a unique policy transmission channel. Joint exposure, horizon and timing restrictions are necessary. The empirical comparisons are descriptive and do not separately identify causal effects of intervention, monetary news or portfolio adjustment.", "keywords": [ "Japan", "yen appreciation", "carry unwinds", "identification" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "author_abstract": true, "related_sources": [], "abstract_source": "Author manuscript" }, { "id": "0003", "title": "Monetary Normalization and Institutional Portfolio Reallocation in Japan", "authors": [ "fuli-yang" ], "type": "Preprint", "series": "japan-macro-finance", "doi": "10.5281/zenodo.23631727", "date": "2026-10-25", "source": "Zenodo", "url": "https://zenodo.org/records/23631727", "access": "Restricted full text at source", "summary": "I show that domestic government-bond acquisition coexisted with continued overseas investment and changes in external investment form during Japan’s monetary normalization. Monthly transactions, quarterly financial accounts, domestic bond trades and company disclosures document the adjustment over 2024Q2–2026Q2. Banks recorded 45.01 trillion yen of net government-security acquisition and positive foreign-security transactions, while their net claims on the central bank fell by 48.65 trillion. Investment trusts’ net foreign-security acquisitions of 20.09 trillion comprise 12.99 trillion of proportional expansion and 7.09 trillion of additional allocation. Life insurers combined net foreign-security sales with additions to direct investment and other external claims; the narrower securities-plus-direct-investment balance depends on a concentrated acquisition quarter. Company records connect these sector patterns to investment mandates, currency contracts and bond replacement. Within one insurer, asset–liability management and balanced mandates move Japanese government bond holdings in opposite directions. Replacement trades change duration and future income objectives while recognizing disposal gains or losses. Historical yield regressions relate allocation to specified asset-yield and hedge conditions; their statistical strength varies with horizon, sample and dependence assumptions. Joint Japanese/global rate equations distinguish current-month short-debt adjustment from a delayed equity-and-fund margin. These results locate Japan’s portfolio adjustment in the institutions’ joint asset and funding decisions and in the type of external claim acquired.", "keywords": [ "Japan", "monetary normalization", "institutional investors", "portfolio reallocation" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "author_abstract": true, "related_sources": [], "abstract_source": "Author manuscript" }, { "id": "0002", "title": "The Recovery Gap: Bank Equity, Credit, and Output after Financial Crises", "authors": [ "fuli-yang" ], "type": "Preprint", "series": "financial-recovery", "doi": "10.5281/zenodo.23591522", "date": "2026-10-23", "source": "Zenodo", "url": "https://zenodo.org/records/23591522", "access": "Restricted full text at source", "summary": "Banking crises leave a persistent recovery gap between bank shareholder wealth, credit, and production. Across 119 crises in 39 countries, these dimensions occupy different positions relative to their pre-crisis levels in 57.1% of episodes at year five. In 96 crises with nearby quiet-period references in the same countries, mismatch is 58.3% after crises and 20.5% in the quiet periods. Bank equity alone remains below its earlier level while credit and output attain theirs in 32.3% and 12.5%, respectively. Among the 110 crises with complete bank and nonfinancial total returns, bank equity underperforms nonfinancial equity by 19.1 log percentage points during the crisis year and a further 27.9 points over the following five years. Restoration also depends on the initial loss distance. Among 40 crises initially damaging both bank equity and output, the 30-point difference in two-year restoration combines a −41.25-point distance contribution with a +11.25-point realized-path contribution. Requiring attainment in both final years preserves the distance disadvantage and yields a 35-point restoration gap. Regaining economic scale leaves substantial growth-path losses: within the same 31 episodes with credit and output at their old levels but bank equity below, 26 fall short of GDP growth continuation and 25 fall short of credit continuation. In the 69 episodes with unemployment data, 38 of the 57 whose GDP regains its earlier level still have higher unemployment. Economic activity resumes at its earlier scale while bank shareholder wealth and prior expansion paths remain incompletely restored.", "keywords": [ "financial crises", "bank equity", "credit", "output", "recovery" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "author_abstract": true, "related_sources": [], "abstract_source": "Author manuscript" }, { "id": "0011", "title": "Mind the Gap: LLM Extraction of Climate Scenario Financial-Impact Disclosures and Their Pricing in Hong Kong Equity and Credit Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-10-22", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7608119", "access": "Source repository access terms apply", "summary": "Climate scenario analysis is increasingly common, but its translation into issuerspecific financial effects remains uneven. We examine this reporting gap and the pricing of the resulting disclosures using 267 reports from 140 Hong Kong large firms for FY2024–FY2025. Page-cited, schema-constrained LLM extraction records financial effects, numerical forms and scenario links. Quantified disclosure rises from 13.8% to 34.3%, with 33 entrants and five exits among 130 continuously observed firms. Yet 64.2% of 416 extracted records use ratios or percentages, and 54.3% are bounds or ranges; only nine report-years in the available effect lists contain a scenario-linked currency point amount. Equity tests separate average announcement returns from response intensity and distinguish broad quantification from this more specific financial content. Broad quantification shows no stable increase in signed returns, absolute abnormal returns or abnormal volume. An exploratory specificity contrast is positive for signed returns, but rests on nine events and does not survive the four-outcome multiplicity adjustment. Onshore credit pricing exhibits a positive post-release differential: the expanded sample's year-adjusted AAA MTN premium difference is 4.50 basis points, with a wild-cluster bootstrap p-value of 0.020; issue-month controls leave the estimate positive but imprecise. The evidence establishes substantial adoption of financial quantification alongside incomplete scenario-to-financial translation and differentiated equity and credit pricing associations.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Climate scenario analysis", "Financial-impact disclosure", "IFRS S2", "Large language models", "Capital markets", "Hong Kong" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7608119" }, { "id": "0001", "title": "AI Booms and Industrial Cycles: Capacity Reallocation in the Memory Industry", "authors": [ "fuli-yang" ], "type": "Preprint", "series": "ai-industrial-cycles", "doi": "10.5281/zenodo.23593124", "date": "2026-10-21", "source": "Zenodo", "url": "https://zenodo.org/records/23593124", "access": "Restricted full text at source", "summary": "Artificial intelligence (AI) booms can generate industrial price cycles even when frontier demand remains permanently elevated. I develop a two-product model in which high-bandwidth memory and conventional dynamic random-access memory (DRAM) compete for shared productive capacity, while investment arrives through a construction queue. Frontier demand initially diverts resources from conventional production. I derive a condition under which sufficiently responsive investment and costly sustained specialization subsequently expand conventional supply above its initial level. A second result maps installed capacity, inherited allocation, and committed deliveries into the entire supply path over the next commissioning window, conditional on demand. Evidence combines 68 fiscal quarters of Micron financial data, a 14-quarter fixed-supplier DRAM panel, paired business-unit price and shipment disclosures, and dated investment and contract commitments. The paired evidence identifies recipient expansion and donor restriction; joint inference places the financial investment response around two to three quarters after revenue growth. In the normalized permanent-demand experiment, the conventional price peaks 10.22 percent above its initial level and settles 0.81 percent below it. The reversal persists with current-order feedback. A temporary demand expansion generates a larger subsequent supply release, whose amplitude depends strongly on expected duration. Product composition and committed capital jointly determine how a technology boom creates and resolves scarcity across connected markets.", "keywords": [ "artificial intelligence", "industrial cycles", "capacity allocation", "investment", "memory industry" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "author_abstract": true, "related_sources": [], "abstract_source": "Author manuscript" }, { "id": "0115", "title": "Credit Support, Financing Expansion, and Risk Allocation in AI Infrastructure", "authors": [ "fuli-yang", "jing-huang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-10-20", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7635218", "access": "Source repository access terms apply", "summary": "Credit support shapes both the financing of AI infrastructure and the distribution of its downside exposure. We combine a 30-project powered-land inventory, a purposive register of 15 financing instruments across four parent groups, executed contracts and regulatory orders with a model of financing choice. Three results follow. First, support expands debt capacity under a binding expected-loss-rate constraint. When costly external equity induces use of that capacity, absolute expected lender losses increase; partial use can combine a lower loss rate with a larger monetary loss. This mechanism extends to risk budgets that rise with exposure, while a fixed monetary budget holds expected loss constant. CoreWeave’s delayed-draw term-loan principal rises USD 3.972 billion in H1 2026, and Applied Digital replaces a USD 300 million bridge with USD 1.590 billion of project notes. Second, payment priority, customer acceptance, enforcement timing and power-service conditions allocate shortfalls and financial flexibility across equipment, campuses and electricity investment. Third, common support providers increase portfolio covariance at unchanged marginal credit quality. In an exact illustrative experiment, concentration raises 99 percent expected shortfall by 58.9 percent. A heterogeneous-claim extension identifies each provider’s conditional loss exposure as the relevant concentration weight. Credit support is a financial production input and a contingent allocation rule. Financial-stability assessment must connect funded amounts, contractual activation states and the balance sheets supplying protection.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7635218" }, { "id": "0010", "title": "Who Pays for Generative AI? Work, Household Budgets, and Subscription Demand", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-10-20", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7643078", "access": "Source repository access terms apply", "summary": "Personal demand for generative AI connects workplace adoption to household budgets. We study this connection in eleven weekly ConsumerInsight surveys from South Korea, comprising 8,800 records and a main occupational sample of 4,440. Joint work and personal use is associated with substantially more widespread willingness to allocate a monthly AI budget. After adjustment for demographics, occupation, firm size, workplace policy and use frequency, the mixed-use differences reproduce in nine later waves at 11.3 percentage points relative to no work use and 8.5 points relative to work without personal use. A seven-configuration decomposition shows a further 4.6-point difference when study joins work and personal uses. The monetary ranking also holds in the full occupational sample after covariate adjustment and every admissible completion of incomplete amounts: at KRW 10,000, the later-wave mixed-use differences are bounded below by 4.2 and 2.6 points. Affirmative subscription-expenditure records are concentrated among work users, who account for 92.2% of these records against 73.8% of the comparable core sample. A net-value framework explains how benefits across purposes enter personal budgets relative to free and organizational alternatives. The findings establish purpose breadth as a substantial dimension of AI financing demand and place household budget decisions alongside organizational adoption in explaining the formation of a paying market.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Generative AI", "Workplace adoption", "Household budgets", "Subscription demand", "Willingness to pay" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7643078" }, { "id": "0113", "title": "Carbon Border Adjustment and Corporate Green Transformation: International Market Connections and Investment Composition in China", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-10-16", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7621100", "access": "Source repository access terms apply", "summary": "How does carbon border adjustment shape corporate green transformation through firms' international market connections? We examine its investment dimension by reconstructing annual project additions from Chinese listed firms' construction-in-progress notes. The source corpus contains 714 annual reports and yields 79 firms with 443 observations over 2018–2024. We distinguish investment in firms' own operations from capacity supplying the green transition. Broad-panel exposure coefficients are imprecise. In a 22-firm foreign-revenue extension, the 2023–2024 exposure-by-foreign-revenue interaction is positive under both specifications: combined-green coefficients are 1.950 and 1.066, while green-supply coefficients are 1.941 and 1.315. Firm wild-bootstrap p-values are 0.084 and 0.271 for combined investment and 0.076 and 0.075 for supply; industry and familywise inference yield weaker evidence. Green-supply signs persist under single-firm deletion and across three monetary units. The findings locate the positive investment pattern at the intersection of coveredproduct exposure and established foreign business. Corporate green transformation around CBAM extends to the composition of capital formation, including capacity supplying transition-related goods. Project-based accounting makes this investment margin visible during policy preparation.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "carbon border adjustment", "corporate green transformation", "green investment", "internationalization", "China" ], "jel": [ "Q58", "F18", "Q56" ], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-10-17", "ssrn": "7621100" }, { "id": "0112", "title": "From Prediction to Allocation: The Economic Value of Algorithmic Decisions", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-10-16", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7620802", "access": "Source repository access terms apply", "summary": "Better predictions and broader allocation flexibility create complementary economic value. I measure their returns in 25 Chicago food-inspection queues from 2013 to 2026, covering 27,637 inspections and 6,473 failures. Predictions use pre-origin information; comparisons hold observed outcomes and daily inspection slots fixed. Boosting advances failure discovery by 5.61 days and reaches the historical first-month discovery target with 25.3% fewer inspection slots. A crossed comparison within logistic regression separates information from allocation flexibility. Adding establishment and location information with ZIP-specific workloads fixed improves discovery timing by 0.03 days. Expanding allocation citywide with history scores fixed adds 1.17 days. Improving both adds 2.21 days, including a 1.01-day interaction with a 95% paired interval of 0.25 to 1.73 days. Under boosting, within-ZIP prioritisation accounts for 451 additional firstmonth discoveries and citywide flexibility adds another 390. Prediction's value depends on the decisions it can change. Information and the organisation of allocation jointly determine the economic return to algorithms.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "prediction", "resource allocation", "information value", "machine learning", "inspections", "scheduling" ], "jel": [ "D83", "D24", "C53" ], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7620802" }, { "id": "0056", "title": "The Readiness Gate in Climate Finance: Allocation Scale and the Adaptation Divide across Asia and the Pacific", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-10-12", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7600100", "access": "Source repository access terms apply", "summary": "This paper identifies the allocation architecture that converts climate need into development finance across Asia and the Pacific. It assembles 2,920 Asian Development Bank financing transactions covering 1,409 projects committed between 2019 and 2025 and matches them to lagged vulnerability, institutional readiness, financing constraints, disaster exposure, sovereign market access, and pre-sample ADB relationships. Vulnerability alone does not produce larger allocations. Readiness changes the scale at which vulnerability enters the commitment portfolio. Among recipient country-years, the gate coefficient of 0.482 decomposes exactly into a 0.179 project-count margin and a 0.303 average-project-size margin; both components are significant. Pre-2019 approval volume predicts later finance, while historical project stocks absorb more of the interaction, showing that readiness compounds through inherited portfolio relationships. The gate survives wild-cluster inference, every leave-one-country-out estimate, and controls for debt, disasters, and sovereign-rating coverage. A stacked system rejects an equal mitigation and adaptation response: the adaptation interaction is 0.337 below mitigation in PPML and 0.265 below it in the log specification. Energy carries the original-dollar scale response, while agriculture's gate is 0.602 lower. Small island developing states share the scale architecture but face sharper entry friction. Climate-finance scarcity is reproduced where vulnerable countries must convert need into signable portfolios. Closing the adaptation gap requires an operating model that finances project formation, aggregation, and implementation capacity alongside physical assets.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7600100" }, { "id": "0057", "title": "Model-Informed Gold Revisions Inflate Measured Accuracy: A Versioned Gold-Standard Protocol for LLM Extraction of ESG Data", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-10-11", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7598018", "access": "Source repository access terms apply", "summary": "Language-model extraction of ESG data is scored against human references that are often revised after the model has run, yet those revisions are seldom versioned or measured. We propose a versioned gold-standard protocol: a blind reference fixed before extraction, every later revision logged with its page evidence and trigger, identical predictions scored against every version, and a worst-case headline. We apply it to 36 stratified Hang Seng Index constituents (396 cells, eleven headline fields). Revisions prompted by model disagreement change 22 cells, 21 toward the model, and raise measured accuracy by 5.1 percentage points with the predictions unchanged. Against the blind reference, a BM25-plus-LLM pipeline scores 90.9% (firm-bootstrap 95% CI 87.1–93.9%) on the 24 firms annotated fully blind and never used in prompt development, and 90.7% on all 36 (89.6% under the pre-specified unit normaliser). A generic regex-and-table heuristic scores 35.6%. Retrieval placed the gold evidence page in the model’s context for 98% of disclosed cells, including every missed or wrong value that has a cited page. Under the written prompt, only two of the 37 errors are model errors, plus one masked by an identical gold error. The rest are gold omissions, prompt–gold specification conflicts, definitional ambiguity and unverifiable disclosure. For this reasoning model, residual error arises after retrieval, in the definition of the target; a partial comparison suggests a weaker model loses about ten points. ESG extraction accuracy is a property of a model and a reference together, and benchmarks should report both.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-10-12", "ssrn": "7598018" }, { "id": "0058", "title": "Design Abundance and Laboratory Scarcity: The Changing Production Function of AI Drug Discovery", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-10-09", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7590558", "access": "Source repository access terms apply", "summary": "AI expands candidate drug design faster than validation capacity. I develop a bottleneck-relocation framework in which experimental, manufacturing, and clinical capabilities become scarce complements to computational search. An outcome-blind matched comparison finds that 10 of 14 AI-linked assets, versus 27 of 67 contemporaneous controls, entered Phase II or later; the matched gap is 31.4 percentage points. The evidence shows that AI is already changing pharmaceutical innovation by moving advantage toward organizations controlling experimental feedback and execution.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-10-12", "ssrn": "7590558" }, { "id": "0059", "title": "The AI Capital Expenditure Cycle: Investment, Financing, and Risk Across the Compute-power Complex", "authors": [ "fuli-yang", "jing-huang", "shiyu-lin" ], "type": "Working paper", "series": "energy-compute-infrastructure", "doi": "", "date": "2026-10-07", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7574438", "access": "Source repository access terms apply", "summary": "AI infrastructure investment creates a capital cycle in which demand commitments, construction funding, physical delivery and cash recovery mature on different schedules. We measure this cycle across compute, networking, data centers, electrical equipment and power using a fixed 100-firm universe, 29 announcements and a 2015–2026 quarterly grid. Public disclosure dates connect valuation, investment, revenue, financing gaps and subsequent residual risk. Consecutive-quarter models estimate conditional risk carryover of 0.455 across 99 firms and 0.399 with dated financial controls. A joint system contains 2,459 firm-event-horizon observations and provides an exact aligned decomposition of direct and accounting-path coefficients. Fixed-scale risk estimates are negative at one and two quarters and positive at four; common support identifies a substantial composition component, and all four simultaneous sensitivity bands span zero. An evidence map joins historical investment loadings, disclosed customer-funding stocks and named relationships while retaining their economic units. Ten project cohorts and fourteen company cash quarters show how construction support precedes service conversion. Net deferred-revenue contributions equal 37.7%, 42.5% and 53.5% of recorded company capital spending at Core Scientific, IREN and Nebius over their respective observed windows. Matched site disclosures measure the progression into billing, while contractual offsets determine later fee cash. The evidence establishes customer funding as a distinct component of the observed financing interval and makes the allocation, realization and recovery of AI infrastructure capital jointly measurable.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-10-08", "ssrn": "7574438" }, { "id": "0009", "title": "Where public climate signals remain visible: issuer compression and debt-rating sorting in China", "authors": [ "fuli-yang" ], "type": "Journal article record", "series": "climate-finance", "doi": "10.1108/JFRA-08-2026-0832", "date": "2026-10-07", "source": "Emerald Publishing", "journal": "Journal of Financial Reporting and Accounting", "version": "EarlyCite version of record", "url": "https://doi.org/10.1108/JFRA-08-2026-0832", "access": "Publisher access terms apply", "summary": "An archival study of Chinese medium-term notes examines how public climate information appears at the issuer and debt-rating levels. It separates concentration in issuer grades, finer debt-grade resolution, and the effects of issuance frequency. The analysis evaluates sensitivity to proxy definitions, geographic composition, and imbalanced rating classifications. The findings motivate transparent climate-information overlays and improved links between obligors, facilities, emitters, guarantors, and securities.", "keywords": [ "climate signals", "credit ratings", "corporate bonds", "rating compression", "China", "emissions trading" ], "jel": [ "G24", "G32", "G38", "Q54" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Editorial research summary", "rights": "Copyright 2026 Emerald Publishing Limited. Licensed reuse rights only.", "related_sources": [] }, { "id": "0062", "title": "Fast-Entry Funding Pressure and Incumbent Liquidity: Evidence from the SpaceX Nasdaq-100 Inclusion", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-09-27", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7531538", "access": "Source repository access terms apply", "summary": "This paper studies the financing side of large, preannounced index additions. Space Exploration Technologies Corporation entered the Nasdaq-100 before the market open on 7 July 2026, creating a 6 July implementation event. I construct model-implied funding pressure for 93 incumbents by allocating the entrant-funding requirement across pre-event weights and scaling implied sale dollars by pre-event average daily dollar volume. A one-standard-deviation increase in pressure is associated with 12.1 percent higher implementation-day volume relative to normal. The coefficient is 8.6323 with HC3 p = 0.0000283; randomization, bootstrap, and leave-one-out inference agree. A strictly lagged specification retains an 8.3 percent effect (p = 0.0080), while the coefficient decays and becomes insignificant on the next session. I then apply the frozen construction to eleven earlier off-cycle Nasdaq-100 additions from 2023 to 2026, using QQQ holdings disclosed before each announcement and controls measured over the preceding 20 sessions. Ten of twelve event coefficients, including SpaceX, are positive. The equal-event mean is 0.0672 log points per pressure standard deviation, with an exact event-sign randomization p-value of 0.0063. A stacked regression with event and sector fixed effects yields 0.0675 (event-clustered p = 0.0199). Capacity-scaled funding pressure therefore orders incumbent turnover beyond the SpaceX episode, while SpaceX provides the cleanest fast-entry implementation test.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-30", "ssrn": "7531538" }, { "id": "0061", "title": "Regulatory Window Risk in Carbon Border Adjustment: Historical Evidence for the CBAM Quarterly-to-Weekly Transition", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-09-27", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7531578", "access": "Source repository access terms apply", "summary": "The European Union Carbon Border Adjustment Mechanism converts EU allowance auction prices into regulated import liabilities. This paper measures regulatory window risk: the pricing interval determines how far an auction-day carbon price can lie from the regulated price assigned to that date. Using 2,676 official European Energy Exchange EUA primary auctions from 2013 through 2025, I apply weekly, 14-day, monthly, and quarterly volume-weighted pricing rules to identical realized price paths. Mean quarter-level absolute basis rises monotonically from 1.927 percent under weekly pricing to 6.145 percent under quarterly pricing, and weekly pricing lowers it in all 52 quarters. The 4.218-percentage-point paired reduction is highly significant under t, signed-rank, and 200,000-draw sign-flip inference. For squared loss, lower error under a nested weekly partition follows from an exact weighted-variance decomposition; the empirical contribution is its magnitude and state dependence. Between-week variation accounts for 86.0 percent of quarterly weighted squared error on average, and the quarterly-to-weekly MAE reduction correlates 0.950 with withinquarter auction-price variation. Actual calendar weeks differ from quarter-contained weeks by at most 0.188 percentage points of quarter MAE. Exact reproduction of the first two official 2026 prices validates the regulatory construction. The 2027 shift therefore materially compresses liability basis, especially when a quarter spans different carbon-price states.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-28", "ssrn": "7531578" }, { "id": "0060", "title": "Market-Implied Policy-Outcome Dispersion and Long-Run Equity Volatility: Meeting-Level Evidence from Kalshi FOMC Contracts", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "central-bank-communication", "doi": "", "date": "2026-09-27", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7531599", "access": "Source repository access terms apply", "summary": "This paper tests whether dispersion in market-implied Federal Open Market Committee outcomes contains information about the long-run component of U.S. equity volatility. For 38 meetings, I construct the crossoutcome variance of Kalshi probabilities over days -30 to -3 and embed it in a fixed (K=4), uniform-weight GARCH-MIDAS model for 824 S&P 500 trading days. The raw loading is 0.3621 (LR = 7.9637; asymptotic p = 0.00477; 999-draw meeting randomization p = 0.027). Conditioning on the implied policy mean, volume, the listed outcome menu, and pre-meeting S&P 500 realized variance gives a loading of 0.3611 (LR = 9.2777; p = 0.00232; conditional randomization p = 0.023). The listed-support specification without realized variance gives 0.3675 and conditional randomization p = 0.018. Activating the factor three business days before, on, or one business day after the meeting yields positive loadings, entropy is insignificant, and all leave-one-meeting-out variance loadings remain positive. In a pseudo-real-time design that begins after three annual policy cycles, re-estimates at each meeting, and freezes parameters until the next meeting, Kalshi reduces mean QLIKE by 2.10 percent over 417 days; the Newey-West p-value is 0.399. The evidence identifies a robust pre-decision relation with the fitted long-run volatility state and a positive but not yet statistically established forecasting gain.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-30", "ssrn": "7531599" }, { "id": "0063", "title": "The Externality of Learning: Adaptive AI and Market Resilience", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-09-26", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7527141", "access": "Source repository access terms apply", "summary": "Adaptive financial AI creates a system-level question that single-model evaluation cannot answer: what happens when multiple agents learn from the same experience? I identify this externality with a marginal-preserving experiment that varies exact cross-agent overlap in sequential learning histories while holding each agent's marginal learning problem fixed. Learned policies enter matched limit-orderbook markets under Deep/Thin liquidity and Shock/No-shock counterfactuals. In a pre-specified 240block architecture screen, KNN5 produces a +12.25 bp-round market-resilience effect and remains significant after family-wise correction across Ridge, CART, and KNN5 (max-T adjusted p = 0.023), while Ridge and CART remain near zero. KNN5 is then frozen and run on a fresh 240-block seed domain; the second estimate has the same sign and is statistically compatible with the first. Pooling the two independent KNN samples yields +9.32 bp-rounds [2.74, 15.91], and the result survives exclusion of the shock round itself. Shared histories increase policy dependence without materially changing heldout loss, while KNN converts that dependence into common local reference sets. Common learning therefore creates joint dependence; the architecture and locality with which experience is reused determine its market pass-through under stress.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-28", "ssrn": "7527141" }, { "id": "0064", "title": "Investor Protection across Markets: Who Can Exercise Shareholder Entitlements in Corporate Spin-offs?", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-09-24", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7520379", "access": "Source repository access terms apply", "summary": "Investor protection in corporate spin-offs is often coded as a binary feature: an assured entitlement exists or it does not. In a cross-market transaction, a right may attach to one shareholder market but not another, exist legally but be unusable through the ordinary custody channel, or deliver cash rather than the spun-off security. This paper asks not only whether a safeguard exists, but on whom it is conferred and whether it can be exercised in practice. It develops a five-dimension coding framework—coverage, exercisability, delivery, shareholder voice and market channel—and applies it to a source-linked registry of Hong Kong-listed parent –SpinCo transactions. Midea–Annto and Weichai–Lovol illustrate A/H legal-eligibility asymmetry; WH– Smithfield and Sihuan–Xuanzhu illustrate infrastructure constraints. Completed distributions provide comparison cases. The evidence supports mechanism and existence claims, not population frequencies or causal effects. A procedural benchmark asks whether differential treatment is justified, disclosed, practicable, and accompanied by a security route, monetised substitute or reasoned explanation. The contribution is to recast investor protection as a question of rights incidence and institutional capacity rather than rule presence alone.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-26", "ssrn": "7520379" }, { "id": "0065", "title": "From Gatekeeping to Accountability: Institutional Tension in Hong Kong's 2026 Spin-off Reform", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-09-23", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7509761", "access": "Source repository access terms apply", "summary": "This paper studies the institutional redesign proposed by the Stock Exchange of Hong Kong on 21 September 2026 for corporate spin-offs. The proposal would permit qualifying Main Board parents to self-assess compliance with Practice Note 15, shorten the post-listing moratorium from three years to one, remove the assured-entitlement requirement, and—through the accompanying reform of notifiable transactions—raise the shareholder-approval threshold for most major transactions from 25 to 50 percent. The paper develops an institutional-substitution framework in which ex ante permission, shareholder voice, and allocation entitlements are partly replaced by board certification, enhanced disclosure, continuing-listing safeguards, and ex post enforcement. Its central tension is that these mechanisms are complementary but not interchangeable: a faster process may improve organizational flexibility while simultaneously reducing independent review and direct minority participation. The historical record supports a proportionate review. The Exchange approved 123 spin-offs during 2021–2025; more than 60 percent involved an overseas listing, more than 80 percent of those overseas cases went to Mainland exchanges, and more than 60 percent of historical cases would have qualified for self-assessment. The paper therefore focuses on the boundary conditions of reform rather than assuming either deregulation or protectionism. It is a design and institutional-analysis paper rather than a completed causal estimate. It establishes a pre-analysis framework for historical counterfactual classification of Hong Kong spin-offs, threshold stress tests, an announcementdate exposure design, and a post-implementation research protocol. The accompanying registry distinguishes the official aggregate universe from a source-linked case subset and does not report a case-level qualification share before the denominator and threshold inputs are reconciled. The analysis contributes to the corporateseparation literature by examining a regulatory regime that frequently governs subsidiary IPOs and crossborder carve-outs rather than the pro rata distributions emphasized in conventional U.S. spin-off research. Because the consultation remains prospective, the paper does not claim causal effects of the proposed rules; it identifies testable predictions, measurement requirements, and the institutional conditions under which disclosure and board accountability could credibly substitute for prior approval and minority-shareholder participation.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7509761" }, { "id": "0066", "title": "Oil-price Innovations and the Treasury Term Premium: Evidence from 2010-2026", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "macro-finance", "doi": "", "date": "2026-09-21", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7497760", "access": "Source repository access terms apply", "summary": "Oil-price innovations are positively associated with the Treasury term premium, but the composition of the associated yield movement changes across periods. Monthly local projections compare 2010–2019 with 2022–August 2026 using an activity-orthogonal oil innovation and a common full-sample scale. The contemporaneous Adrian–Crump–Moench term-premium coefficients are 8.1 and 6.3 basis points, and the corresponding Kim–Wright coefficients are 4.1 and 4.7 basis points. Expected-short-rate-path estimates are smaller and imprecise. A common-regressor decomposition locates the earlier market-yield association predominantly in breakevens and the later association predominantly in TIPS real yields. The between-window change in the breakeven-minus-real-yield coefficient is −14.5 basis points, with a calendar-aware HAC standard error of 4.3 and a block-wild-bootstrap interval of [−24.3, −5.2]. This composition change is more clearly resolved than the change in the total termpremium coefficient. The positive premium association is preserved under nominal-oil and alternative CPI treatments, although the later ACM estimate is more inference-sensitive than its earlier counterpart. The findings distinguish a recurring premium association from a changing market-yield composition and provide a component-level account of energy-related Treasury repricing. The estimates are conditional associations with an oil-price innovation, rather than causal responses to a uniquely identified supply shock.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-23", "ssrn": "7497760" }, { "id": "0067", "title": "Japan's Fiscal Duration: A Measurement Architecture for Contractual Refinancing Exposure and the BOJ Balance Sheet", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-09-20", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7494820", "access": "Source repository access terms apply", "summary": "This paper measures how quickly higher interest rates can pass through Japan’s government debt and the Bank of Japan’s balance sheet without treating those channels as one consolidated fiscalcost estimate. It separates four objects: the contractual maturity profile of the Ministry of Finance debt file, the maturity profile of fixed-rate Japanese government bonds held by the Bank of Japan, a mechanical refinancing benchmark, and the BOJ’s reported income and remittance accounts. The main contribution is quantitative: it shows how common shortcuts change the measured exposure. At 31 March 2026, the archived Ministry of Finance maturity file contains 1,104.3 trillion yen of recorded principal. Exact-date horizons place 13.4%, 30.3%, 43.6%, and 67.1% of that principal within one, three, five, and ten years. The corresponding fixed-rate BOJ holdings subtotal is 521.8 trillion yen, with 14.4%, 37.3%, 55.1%, and 83.1% maturing over the same horizons. Across 37 BOJ snapshots from March 2025 to March 2026, these maturity shares remain in narrow ranges even as fixed-rate holdings decline by 40.7 trillion yen. Under a 100-basis-point shock, multiplying the entire recorded stock by the rate change and horizon exceeds the contractual refinancing benchmark by factors of 24.7, 7.7, 4.9, and 2.9 at the four horizons. Directly subtracting BOJ holdings from an inconsistently scoped government balance produces negative amounts for 13 individual issues, while a scope-consistent book-entry balance removes every negative value. The BOJ accounting evidence leads to the same conclusion. From FY2020 through FY2025, the difference between reported JGB interest income and reserve remuneration never equals BOJ net income or the government payment. In FY2025 that local difference is -0.192 trillion yen, while BOJ net income is 1.926 trillion yen and the government payment is 1.830 trillion yen. The evidence supports a reproducible measurement architecture, but not a causal estimate or a single consolidated government-BOJ cash cost.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-22", "ssrn": "7494820" }, { "id": "0013", "title": "Consensus Amplifies Dot Plot Surprises: Ex Ante Dispersion and Treasury Yield Responses to the SEP", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "central-bank-communication", "doi": "", "date": "2026-09-15", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7445698", "access": "Source repository access terms apply", "summary": "The market impact of the Federal Reserve's dot plot depends on the expectations formed before its release. Using 43 Summary of Economic Projections meetings from September 2015 through June 2026, this paper measures surprises in the projected federal funds rate path relative to the Federal Reserve Bank of New York's pre-meeting surveys and combines them with high-frequency Treasury yield changes. Unconditionally, a 25-basis-point upward path surprise is associated with yield increases of about 4.4, 3.7, and 1.6 basis points at the two-, five-, and ten-year maturities, none of which is estimated precisely. Conditioning on the pre-meeting expectation distribution changes this picture. The same path surprise produces a substantially larger yield adjustment when pre-meeting expectations are concentrated: the interaction between the path surprise and survey dispersion is negative at all three maturities, with estimates of -424, -487, and -360 basis points per percentage point squared after controlling for revisions to inflation, unemployment, and growth projections. Evaluated at the tenth percentile of dispersion, a 25-basis-point surprise moves the twoyear yield by 18.6 basis points; at the ninetieth percentile the response is indistinguishable from zero. The interaction is negative in all 43 leave-one-meeting-out estimates and improves held-out predictive accuracy at the five- and ten-year maturities. Market disagreement, not dispersion across FOMC participants' published rate projections, generates this attenuation; the dot-dispersion interaction is positive rather than negative and is at most marginal. The effect appears in the initial release window, continues through the press conference, and accumulates over the full SEP event. Cross-panel construction and a reliability-based assessment indicate that survey measurement error attenuates rather than manufactures the result. Pre-meeting consensus is therefore a central determinant of dot plot transmission: the projected path matters not only through how far it moves, but also through how firmly the market had converged before publication.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Summary of Economic Projections", "Market disagreement", "Monetary policy surprises", "Treasury yields", "Central bank communication", "FOMC dot plot" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7445698" }, { "id": "0008", "title": "Beyond Two Barycenters: Adaptive Complexity in Distributional Regime Geometry", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "distributional-geometry", "doi": "10.2139/ssrn.7451619", "date": "2026-09-15", "manuscript_date": "2026-09-12", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7451619", "access": "Source repository access terms apply", "summary": "Distributional-regime methods often summarize market geometry with a small set of reference distributions. In the Distributional Regime Persistence (DRP) framework, the maintained representation has been a calm/stress pair of Wasserstein barycenters. This paper asks whether that two-anchor geometry is empirically sufficient. Using frozen pre-2007 calibration samples for the S&P 500, FTSE 100, Nikkei 225, and Hang Seng Index, I fit one-dimensional Wasserstein-1 k-medians representations with one to four prototypes and evaluate them strictly out of sample from January 2007 through June 2026. The analysis is repeated in raw return-distribution space and in independently median-centered, IQR-scaled shape space. Additional prototypes materially reduce out-of-sample reconstruction loss in every market and both spaces: moving from two to three prototypes lowers loss by 6.4%–9.6% in raw space and 7.9%–9.1% in shape space, while moving from three to four still lowers loss by 4.1%–6.4% and 2.8%–4.9%, respectively. Yet lower loss does not imply stable regime identity. Calibration-block bootstrap recurrence rejects several apparently useful extra prototypes, and the number of stable additional anchors differs by market and representation. Stable extra prototypes do, however, recur across markets under a frozen analogue criterion. The main result is therefore a dimensionality distinction: two anchors underfit distributional geometry, but approximation dimension exceeds stable identity dimension. Shape normalization does not eliminate the extra geometry, so the result cannot be reduced to finer volatility-scale partitioning.", "keywords": [ "Wasserstein distance", "distributional regimes", "barycenters", "model complexity", "regime identity" ], "jel": [ "C14", "C22", "C52", "C58", "G01", "G17" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "related_sources": [], "author_abstract": true, "abstract_source": "Author manuscript", "ssrn": "7451619" }, { "id": "0015", "title": "The Anatomy of Repatriation: Investor Heterogeneity in Japan’s Overseas Portfolio Adjustment", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-09-14", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7437042", "access": "Source repository access terms apply", "summary": "Claims that Japanese capital is ‘coming home’ are usually made from aggregate cross-border flows. I ask whether that aggregate sign is a sufficient statistic for repatriation. Using monthly Ministry of Finance transactions in foreign securities from January 2014 through July 2026, I separate banks, trust accounts, life insurers, and investment trusts and condition their purchases on public-data measures of short-end and long-duration Japanese government bond repricing. The internally locked primary tests compare investor-specific responses of foreign long-term-debt purchases at zero-, one-, and three-month horizons, with date-score Newey–West inference and Holm adjustment. None of the primary long-duration tests rejects equality across investors; parallel short-end tests also do not reject. The more robust fact is aggregation failure: 79.5 percent of months contain both buyers and sellers, and roughly onethird of gross core-group activity cancels within the aggregate. Protocol-designated asset-class gates locate heterogeneity elsewhere. Trust accounts differ from investment trusts in short debt after positive long-duration shocks, while banks differ from investment trusts in equities after positive short-end shocks. These gates survive yen-return and rate-differential controls, but the global-rate placebo results imply that the design cannot sharply distinguish a Japan-specific mechanism from common global-rate variation. Life insurers show an exploratory post-2022 decline in foreign-long-debt purchases that is not shared by trust accounts. I interpret the combined evidence as institution-specific asset substitution rather than synchronized repatriation. Because it comes from one official monthly transaction system and the longer-horizon tests have limited power, the design does not identify institutional motives or structural effects. Repatriation is better treated as an investor–asset–horizon classification than inferred from the sign of an aggregate flow.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Japan", "Repatriation", "Portfolio flows", "Institutional investors", "JGB yield curve", "Foreign bonds", "Asset substitution" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7437042" }, { "id": "0014", "title": "Voting Rights and the Volatility Content of Federal Reserve Communication", "authors": [ "fuli-yang", "tingting-qiu", "jing-huang" ], "type": "Working paper", "series": "central-bank-communication", "doi": "", "date": "2026-09-14", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7427340", "access": "Source repository access terms apply", "summary": "Formal voting authority can matter even when it does not visibly change what a central banker says. We exploit the mechanical annual rotation of voting rights among Federal Reserve Bank presidents and ask whether otherwise comparable public language receives a different loading in a persistent component of U.S. equity variance when it carries voting authority. The sample contains 819 eligible speeches by 21 presidential officeholders over 2015-2023 and 71 scheduled meeting cycles. Using an archived Yahoo Finance S&P 500 price tape and strictly lagged realized variance, the interaction between uncertainty/modal communication and statutory voting exposure is negative (delta = -0.461; likelihoodratio p = 0.0007). Meeting-cycle clustering yields p = 0.037, while a 9,999-draw restricted wild-score bootstrap yields p = 0.084 two-sided and p = 0.042 for the prespecified negative direction. The estimate is unchanged in a fixed equal-weight 66-day benchmark and remains negative across 44, 66, 88, and 132 return lags. A 500-draw random-group placebo reaches the observed absolute interaction in 4.6 percent of draws. The interaction is null at individual speech dates. A zero-inclusive selection audit finds no extensive-margin or timing shift, although voting officeholders give about 0.19 more speeches per cycle (wild-cluster p = 0.075). Voting status does not robustly change language under equal-weight officeholder-year inference, and rotating-district unemployment does not explain collective statements. The contribution is therefore an aggregation-specific authorization result with borderline finite-cluster inference, not a claim that rotation mechanically rewrites the Committee's statement.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "FOMC", "Voting-Right Rotation", "Central Bank Communication", "GARCH-MIDAS", "Long-Run Variance", "Textual Uncertainty" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7427340" }, { "id": "0068", "title": "Shared Learning Histories and Policy Alignment: A Marginal-Preserving Experimental Design", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-09-12", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7450838", "access": "Source repository access terms apply", "summary": "Experiments on shared learning histories must distinguish cross-agent dependence from changes in the learning task assigned to each agent. We describe a block construction that varies exact public-history overlap while preserving the marginal distribution of each agent's complete history. Independent public blocks drawn from the same law are exchanged as whole units; private innovations, the learning rule, and measurement remain fixed in distribution. An analytical counterexample shows why matching individual observations is insufficient: replacing one entry of a balanced triplet preserves every entry's marginal distribution but changes the variance of the triplet sum from zero to 2/9. The construction separates overlap within a cohort from common random numbers used to pair treatment comparisons. An illustration reuses 160 archived development cohorts of eight mechanical learners. Full overlap increases mean pairwise cosine similarity by 0.00831 above a zero-overlap baseline of 0.89602, while adjacent contrasts are uncertain. Mean benchmark losses remain closely spaced. These results are conditional on one retained development specification and are not adjusted for development selection. The contribution is an explicit design recipe and its failure conditions for sequential histories. It enables controlled study of dependence from shared histories without establishing a general alignment effect or consequences for endogenous markets.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-15", "ssrn": "7450838" }, { "id": "0018", "title": "Forecast Disagreement and Conditional Model Fragility: Evidence from Eight Equity Volatility Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-09-12", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7423900", "access": "Source repository access terms apply", "summary": "Forecast disagreement is observable before a volatility forecast is evaluated, but it is not a sufficient model-selection rule. Across eight equity indices from 2006 to 2018, a frozen rule replaces XGBoost with HAR when dispersion among EWMA, HAR, and XGBoost forecasts exceeds its expanding 90th percentile. In the S&P 500, the rule identifies unusually large relative XGBoost losses under equal-frequency and run-preserving randomization. That timing evidence survives deleting the five largest contribution days and excluding 2008, but not deleting the largest ten; 77.5% of the total gain comes from 2008. Formal Giacomini-White tests do not reject conditional predictive equality, and always using HAR has lower point-estimate loss. A synchronized eight-market bootstrap supports a positive equal-weight effect but not the joint claim that every market effect is positive. An independent SPY archive replicates at five-minute, not one-minute, sampling; 2019 is a calm-period null. A two-day confirmation rule improves gain per activation while reducing switching, although all implementation rules are post-referee diagnostics without monetized costs. Disagreement is therefore a crisis-sensitive marker of model fragility, not a generally profitable selector or evidence of a ranking reversal.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Realized Volatility", "Forecast Disagreement", "Model Governance", "HAR-RV", "Xgboost", "Conditional Predictive Ability" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7423900" }, { "id": "0017", "title": "Decision-Equivalent Volatility Forecasts: Statistical Loss, Portfolio Value, and Allocation Commonality", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-09-12", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7422079", "access": "Source repository access terms apply", "summary": "Predictive accuracy and decision quality can rank volatility forecasts differently because portfolio maps are invariant to some forecast errors and sensitive to others. This paper formalizes decision equivalence: statistically distinct volatility forecasts are decision-equivalent when they induce the same allocation under a specified rule. For global minimum variance, inverse-volatility, and equal-risk-contribution portfolios, common positive rescaling of marginal volatility forecasts is an exact equivalence transformation. A frozen replay on IVV, EFA, EEM, AGG, and GSG from August 2011 through July 2026 perturbs six prespecified error directions while holding correlations fixed. The common-scale placebo changes mean QLIKE by as much as 1.0024 while maximum weight deviation is 2.22e-16. The strongest cross-sectional distortion worsens QLIKE by 0.406 and raises annualized certainty-equivalent return by 4.19 percentage points, although no cell survives the frozen 120-test Holm family at 5%. Similar loss deterioration can produce zero, modest, or large portfolio-value changes, and the strongest privately useful distortion reduces cross-rule cosine similarity. A supplementary daily-return variance check on 178 complete months preserves the leading loss-deterioration signs. The results separate statistical loss, portfolio value, and allocation commonality as distinct evaluation dimensions.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Volatility Forecasting", "Decision Equivalence", "Predict-then-Optimize", "Portfolio Allocation", "QLIKE", "Certainty Equivalent", "Allocation Commonality" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7422079" }, { "id": "0020", "title": "From Disclosure to Resilience: Climate-Risk Transmission Frictions in China’s Corporate Bond Market", "authors": [ "fuli-yang", "shiyu-lin" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-09-11", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7409338", "access": "Source repository access terms apply", "summary": "The paper examines why climate-related information does not consistently become a traceable signal in corporate bond prices, contracts, and secondary-market resilience. It defines transmission adequacy as the ability to connect a material exposure to the correct legal obligor, align it with the information date and financial horizon, document its treatment by credit intermediaries, and observe it in financing and stress-state indicators. A structured critical review of 40 verified academic and institutional sources organizes evidence around five links: disclosure, exposure measurement, pricing and contract design, credit intermediation, and market resilience. The synthesis identifies five corresponding frictions—legal-entity and asset-location mismatch, information-date and horizon misalignment, measurement non-comparability, absorption by ratings, guarantees and ownership, and limited price discovery in thinly traded bonds. For China, the article proposes Module 1, an issuer–emitter–bond map; Module 2, bond-relevant climate disclosure; Module 3, transparent intermediary treatment; Module 4, a climate-credit stress dashboard; and Module 5, pre-specified pilots and proportional scaling. These modules target traceability rather than an administratively imposed premium and are paired with burden safeguards, implementation metrics and failure interpretations.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Climate Risk", "Corporate Bonds", "Information Disclosure", "Credit Intermediation", "Market Liquidity", "China" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7409338" }, { "id": "0019", "title": "The Scale-Shape Divide in Bitcoin and Ether Perpetuals: Roughness States, Quarticity, and Bipower Residuals", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "crypto-roughness", "doi": "", "date": "2026-09-11", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7415678", "access": "Source repository access terms apply", "summary": "This paper asks whether rolling roughness estimates that label volatility states also rank the shape of near-term tail variation. The design freezes the January 2020–July 2026 Binance BTCUSDT and ETHUSDT tape, 9,560 six-hour origins per contract, and the three functionals used in companion work. Forward 8-hour and 24-hour outcomes are reconstructed from one-minute sufficient statistics. At eight hours, a one-standard-deviation increase in Ĥᴵ ⱽ is associated with 0.186 log points more realized quarticity in Bitcoin and 0.166 in Ether; the corresponding Ĥᴾᴬ effects are 0.177 and 0.157. Origin-level 14-day block-cluster inference preserves all four effects, whereas Ĥᴾⱽ remains null. The accounting identity log RQ = 2 log RV + log NRQ locates the association: it loads through realized-variance scale. Log normalized quarticity, the raw bipower-residual share, and a conservative RQ-scaled significance-filtered residual all satisfy prespecified equivalence tests. Recursive raw-quarticity forecasts improve on both historical and state benchmarks in Bitcoin, but Ether does not reproduce the short-horizon gain; normalized-tail and residual-event results are still less stable. Roughness therefore reaches short-horizon fourth-moment scale in these contracts without furnishing a common tail-shape coordinate.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Rough Volatility", "Realized Quarticity", "Normalized Quarticity", "Bipower Residual", "Perpetual Futures", "Tail Forecasting" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7415678" }, { "id": "0069", "title": "The Missing First Rung: Aggregate Resilience and Entry-Level Compression in the AI Labor Market", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-09-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7441121", "access": "Source repository access terms apply", "summary": "Public discussion of artificial intelligence and employment has split between a jobs apocalypse and an AI jobs boom. Both claims are easier to make than to count because creation is often measured with broad infrastructure employment while displacement is restricted to explicitly attributed layoffs. I construct a symmetric occupation-industry ledger from Current Population Survey microdata covering January 2015 through August 2026. The primary universe fully excludes electric-power cells; power appears only in a separate sensitivity analysis. I also formalize a demand-expansion mechanism in which AI can increase output while reducing the junior labor required per unit of output. Aggregate results show resilience rather than collapse, but not a model-invariant boom. In an observed-use design, the post-ChatGPT aggregate high-exposure employment gap ranges from minus 0.47 to plus 0.34 percentage points across four counterfactuals. Its first-year gap lies 0.009 points above the maximum of five post hoc fully pre-pandemic placebo windows, but the five-window comparison has a minimum attainable randomization p value of 0.167; the gap also lies inside the wider distribution formed by eight pre-event breaks, three of which overlap the pandemic. The latest broad occupation-industry balance is 0.52 to 0.97 million jobs above trend, but its post-event mean ranges from minus 0.25 to plus 0.12 million. For workers age 22 to 25, high observed-exposure employment is 2.19 to 2.58 percentage points below trend across specifications; the primary estimate is minus 2.58 points. Their probability of moving from nonemployment into any job is 4.30 points below trend in the primary model and remains negative under all four counterfactuals. I read this as first-rung compression within a broader and partly model-dependent youth employment slowdown, not as a pure AI substitution effect. Ages 26 to 35 have negative average stock gaps in three of four counterfactuals and negative cell-ledger balances in all four, so they are not a clean unaffected control. The bottom line is aggregate resilience with a pronounced first-rung compression whose exact causal AI share remains unidentified.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-14", "ssrn": "7441121" }, { "id": "0024", "title": "The Accuracy-Diversity Frontier in Adaptive Volatility Forecasting", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-09-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7411359", "access": "Source repository access terms apply", "summary": "Adaptive forecasting systems are usually evaluated by predictive loss. This paper asks a different question: what happens to collective model diversity when forecasting systems retain different amounts of historical information? Using four equity markets from 2000–2026, a common out-of-sample period beginning in 2015, four forecast horizons, and heterogeneous forecasting technologies, I construct frozen, finite-memory, and expanding-memory forecasting cohorts that differ only in their updating rule. Full-history expanding adaptation produces lower QLIKE loss and lower effective model diversity than every finite-memory specification throughout the 16-cell market–horizon robustness grid. The cells are dependent and are not interpreted as independent replications; when horizons are collapsed within market, all four market blocks retain the same sign, but four blocks provide limited standalone statistical power. An equal-sample-size exclusion test shows that the contrast survives when broad-history and matched 3-, 5-, and 10-year learners use the same number of training observations, indicating that sample size alone does not account for the frontier. Around 22 rule-based market-specific stress onsets, expanding adaptation attenuates the post-stress loss of forecast diversity relative to a frozen benchmark; the result also survives re-clustering temporally overlapping cross-market events into global episodes. A volatility-targeting mapping shows that forecast architecture can propagate to the correlation of position adjustments, while the paper deliberately stops short of claiming a causal effect on observed market volatility. The evidence implies that predictive accuracy and collective diversity are distinct design objectives for adaptive financial forecasting systems.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Volatility Forecasting", "Adaptive Learning", "Model Diversity", "QLIKE", "Market Stress", "Continual Learning" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7411359" }, { "id": "0023", "title": "Volatility Persistence is not Distributional Memory: Separating Variance Memory from Empirical-distribution Memory", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "distributional-geometry", "doi": "", "date": "2026-09-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7407659", "access": "Source repository access terms apply", "summary": "Volatility persistence is commonly summarized by the persistence parameter of a conditional variance recursion or by the autocorrelation of realized volatility. Neither statistic directly answers whether an empirical return distribution remains aligned with the same distributional state. This paper separates variance memory from distributional memory using daily returns on the S&P 500, FTSE 100, Nikkei 225, and Hang Seng from 2000 through June 2026. Distributional states are defined from pre-2007 low- and high-volatility prototypes, and are studied through state duration, Wasserstein displacement, and tail-state recurrence. The analysis combines a four-specification GARCH-family grid with a nested circular-block bootstrap that re-estimates prototypes and thresholds in each replication. All sixteen volatility-model fits converge, and Student-t GJR-GARCH has the lowest BIC in every market. The state-switching pattern is positive across the first three markets under the primary nested bootstrap, but only the Nikkei 225 retains a positive lower bound under the conservative 264-day block. Wasserstein and realized-volatility half-life comparisons are formally inconclusive, while no tail-recurrence lift survives Holm correction. The evidence supports a distinction between the estimands and a model-robust but block-sensitive pattern of state turnover; it does not establish a universal or causal separation between variance persistence and distributional memory. The numerical thresholds and decision rules were not changed in response to this revision.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Volatility Persistence", "Distributional Memory", "GARCH", "GJR-GARCH", "Wasserstein Distance", "Regime Duration", "Structural Breaks", "Bootstrap Inference" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7407659" }, { "id": "0022", "title": "Scale, Shape, Location, and Speed: Decomposing Distributional Volatility Persistence", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "distributional-geometry", "doi": "", "date": "2026-09-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7407899", "access": "Source repository access terms apply", "summary": "Volatility persistence is usually represented by a scalar conditional variance, even though return distributions can remain abnormal through dispersion, center shifts, shape distortions, or continued movement. This paper develops a leakage-safe Wasserstein framework that separates these features into four rolling coordinates: scale, standardized shape, signed location, and distributional speed. Using daily S&P 500, FTSE 100, Nikkei 225, and Hang Seng returns from January 2000 to June 2026, the robust scale coordinate is strongly associated with 22-day realized volatility, whereas standardized shape is approximately orthogonal to the current volatility level. Signed location is negatively associated with realized volatility, and speed occupies an intermediate position because it measures one-step movement rather than state level. Scale is also the most synchronized coordinate across markets; location and speed are moderately synchronized, while shape is largely market-specific. Paired block-bootstrap tests, multiplicity corrections, block-length sensitivity checks, and a variance-matched heavy-tailed placebo support the separation of the coordinates. The results establish measurement nonredundancy rather than forecasting superiority or causal transmission. The framework therefore recasts volatility persistence as a multidimensional financial-risk state and provides testable inputs for subsequent state-model, recovery, and forecasting comparisons.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Volatility Persistence", "Wasserstein Distance", "Distributional Regimes", "Realized Volatility", "Market States", "Optimal Transport" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7407899" }, { "id": "0021", "title": "Unknown Is Not Zero: Heat Stress, Data-Center Power-Demand Exposure, and Utility Volatility", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "energy-compute-infrastructure", "doi": "", "date": "2026-09-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7434182", "access": "Source repository access terms apply", "summary": "This paper asks whether service-territory heat stress and disclosed data-center power-demand exposure jointly amplify the volatility of U.S. listed electric utilities. Its principal finding is methodological but empirically consequential: pre-disclosure silence is not a verified zero. The study constructs an auditable point-in-time chain from dated operating or binding customer-load megawatts to the regulated utility, the applicable EIA Form 861 peak-load denominator, a representative NOAA station, daily parent-company returns, and lag-available financial controls. Ten issuers satisfy a broad data-center four-quarter evidence rule over 2021–2025, although the measure is not uniformly AI-specific and one issuer is predominantly exposed to cryptocurrency hosting. Reclassifying 8,685 legacy zerocoded issuer-days as unknown leaves 3,855 positive-disclosure issuer-days and 212 complete portfolio observations across nine variance-recursion segments. The measurement path is non-monotone: the uncentered interaction is 0.143 in the legacy zero-coded panel and 0.060 when OTTR is excluded; after the common-positive restriction it reaches −2.000 at the coefficient bound, while centering yields an interior estimate of −0.118. Before centering, Corr(C,C×A)=0.979 and the interaction VIF is 29.4; centering lowers the VIF to 1.09. In the final Student-t GARCH-X specification, AIC and BIC select the no-interaction model and a 499-replication conditional parametric bootstrap does not reject it (LR=1.638, p=0.324). The current design therefore fails to detect heat-by-load amplification; it does not establish that the population effect is zero. H3 remains untested, and the pre-specified Markov-switching extension is not estimated. Positively, the audit identifies how exposure coding and coordinate construction can manufacture apparent interaction evidence, supplies a reusable regulatory-data architecture, and defines an advance rule under which a longer panel can retest the same hypotheses without redefining the original evidence.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Heat Stress", "Data Centers", "Artificial Intelligence Infrastructure", "Electricity Demand", "Conditional Volatility", "GARCH-X", "Measurement Audit" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7434182" }, { "id": "0025", "title": "Factor Covariance in Yield-Curve Risk Forecasts: Evidence from U.S. Treasury Yields", "authors": [ "fuli-yang", "shiyu-lin" ], "type": "Working paper", "series": "central-bank-communication", "doi": "", "date": "2026-09-06", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7400719", "access": "Source repository access terms apply", "summary": "Yield-curve risk is often assembled in two steps: estimate separate volatilities for Nelson–Siegel factors, then map those volatilities back to individual maturities. The convenient diagonal map silently treats level, slope, and curvature innovations as conditionally uncorrelated. We show that this is not a harmless simplification for U.S. Treasury yields. Negative level–slope comovement makes diagonal forecasts too large at the front end, with a much smaller effect at long maturities. In a strictly out-of-sample exercise from 2008 to 2026, rolling CCC- and DCC-GARCH forecasts that retain factor covariance reduce QLIKE loss relative to diagonal GARCH at horizons of one, five, and twenty trading days. At two years, DCC lowers mean QLIKE by 0.77, 0.82, and 0.72, respectively, and reduces mean-squared error by 35–60 percent. The advantage narrows sharply with maturity. Two-sided interval exceedances provide the same diagnosis: the diagonal model covers too often because its intervals are excessively wide, whereas covariance-aware forecasts are much closer to nominal coverage. The improvement is not universal—it weakens at longer maturities and reverses for some twenty-day forecasts during the 2022–2023 tightening—but its maturity pattern follows directly from the Nelson–Siegel loadings. Factor covariance is therefore part of the forecast object, not a dispensable refinement to marginal volatility.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Nelson–Siegel", "Yield-Curve Risk", "Multivariate GARCH", "Covariance Forecasting", "Treasury Yields" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7400719" }, { "id": "0027", "title": "Correlated Beliefs, Correlated Trades: Agentic AI and Endogenous Volatility Amplification", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-09-05", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7389159", "access": "Source repository access terms apply", "summary": "Agentic trading systems are usually evaluated by the accuracy of individual forecasts. That evaluation misses a distinct systemic-risk channel: several agents can be individually well calibrated while forming correlated posterior beliefs and submitting synchronized orders. We develop a controlled market experiment that holds marginal predictive accuracy approximately constant and varies information overlap, common posterior calibration error, and the homogeneity of risk controls. Agents report auditable posterior probabilities over calm, turbulent, and crisis states. Their orders clear in a market where one-sided flow consumes depth and affects subsequent prices and observations. The identifying comparison is therefore not AI versus non-AI. It is low versus high cross-agent belief dependence conditional on comparable individual forecast quality. The mechanism predicts a temporal sequence from posterior dependence to order synchronization, depth depletion, price impact, and tail volatility. A randomized independent-execution placebo preserves correlated beliefs while breaking the belief-to-trade timing channel. In the pilot mechanism experiment, accuracy-matched high-dependence agents produce sharply higher order synchronization, realized variance, tail returns, and a spread/depth proxy, whereas the placebo returns close to the independent benchmark. These results are mechanism validation, not evidence about any particular language model. The evidence is deliberately architectural: it establishes the mechanism in a controlled synthetic market and does not estimate the prevalence or aggregate effect of deployed commercial systems.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Agentic AI", "Systemic Risk", "Belief Dependence", "Market Microstructure", "Order Synchronization", "Endogenous Volatility", "Model Risk" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7389159" }, { "id": "0026", "title": "Forgetting and Tail Instability in Adaptive Volatility Forecasting", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-09-05", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7411400", "access": "Source repository access terms apply", "summary": "Continual learning requires a choice about how quickly old information should lose influence. This paper studies the tail-risk consequences of that choice in adaptive volatility forecasting. Using daily historical replay for four equity markets, I estimate a recursive HAR-type forecasting architecture with forgetting factors from 1.0000 to 0.9950, holding the model and information set fixed while varying only the rate at which past observations are discounted. Point estimates for mild forgetting are sometimes favorable after stress onsets, but event-level confidence intervals include zero; no general relearning benefit is established. The robust result is instead a nonlinear tail cost. Using each market’s no-forgetting 99th-percentile QLIKE as a fixed descriptive catastrophic-failure threshold, the full-sample failure rate rises from about 1.03% with no forgetting to 2.03% at λ=0.995. A market-fixed-effect logit estimates a positive full-sample forgetting gradient (coefficient 0.0129 per basis point of 1−λ, p=0.001). A prospective threshold exercise preserves the direction but yields a smaller endpoint increase, from 1.40% to 1.80%, so the magnitude is threshold-construction sensitive. Aggressive forgetting also lengthens failure episodes in the primary classification, although bootstrap uncertainty and the prospective-duration check do not independently resolve persistence. Stress-window amplification is economically large in some markets but not precisely estimated as a universal interaction. A transparent 10/22/44-day HAR-RLS horizon check retains a pooled tail-failure penalty at aggressive forgetting, while also showing market-horizon heterogeneity. The paper therefore characterizes a market-dependent adaptationinstability region rather than a universal critical forgetting factor.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Continual learning", "Forgetting factor", "Volatility forecasting", "Tail risk", "QLIKE", "Adaptive instability" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7411400" }, { "id": "0016", "title": "Measurement Sensitivity of Apparent Volatility Roughness in Perpetual Futures", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "crypto-roughness", "doi": "", "date": "2026-09-05", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7393578", "access": "Source repository access terms apply", "summary": "Empirical associations involving estimated volatility roughness may depend on the construction of the measured path. This study compares an increment-variance slope on log realized variance, the same slope after pre-averaging, and a normalized p-variation diagnostic using Bitcoin and Ether perpetual-futures data from January 2020 through July 2026. The observation window and funding-rate variable are held fixed, while common-sample regressions separate changes in measurement from changes in sample composition. Paired coefficient tests preserve the six-hour calendar when the p-variation diagnostic is unavailable. After correcting spot-price alignment in the basis control, both slope estimators retain negative funding associations in levels and with month fixed effects. The p-variation association is negative for Bitcoin in the level regression but is not distinguishable from zero with month effects for either asset. Holm-adjusted comparisons resolve p-variation versus both slope estimators for Bitcoin and versus the pre-averaged slope for Ether. Other contrasts remain uncertain. The diagnostic is admissible in 83.7% and 81.1% of windows, respectively. Dynamic and non-overlapping-block results depend on the specification. These findings establish a conditional measurement comparison rather than a universal funding–roughness relation, an estimate of latent spot regularity, or evidence of trading profitability.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Apparent roughness", "Realized volatility", "Pre-averaging", "Normalized p-variation", "Funding rates", "Perpetual futures" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-14", "ssrn": "7393578" }, { "id": "0070", "title": "Same Yen Move, Different Shock: A Mechanism-Separated Design for FX Intervention, Monetary Repricing, and Carry Unwinds", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-09-04", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7409138", "access": "Source repository access terms apply", "summary": "This design paper starts from a simple empirical difficulty: the same yen appreciation can be produced by very different shocks. Market commentary often compresses official intervention, a carry-trade unwind, JGB repricing, and a broad dollar move into one narrative. That compression is precisely what the paper asks the data to resist. It separates confirmed yen-buying intervention, intervention anticipation without an executed transaction, Bank of Japan monetary-policy repricing, and global dollar or risk shocks. The design combines official intervention records with high-frequency USD/JPY prices, JGB yields across the curve, interest-rate expectations, option-implied risk measures, and returns on representative carry targets. Its empirical question is whether these mechanisms leave different intraday signatures, cross-asset responses, and persistence profiles. Intervention entry, immediate impact, and persistence are separate estimands; a carry claim must also identify the funding leg, target asset, holding horizon, and realized return condition. The 2026 episode—including the record ¥15.3993 trillion monthly intervention total and a subsequent sharp yen appreciation without confirmed intervention—provides the motivation. The historical date–amount audit is complete, but causal inference remains conditional on execution-time measurement and the limits of the available comparison design. The first implementation supplies a 5-minute market-data diagnostic for seven 2022–2024 operation dates and a daily cash-carry panel with explicit funding and target legs. The intraday diagnostic identifies market-implied breaks, not official transaction timestamps.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-10", "ssrn": "7409138" }, { "id": "0028", "title": "Apparent Roughness and Funding-Rate Asymmetry: Evidence from Bitcoin and Ether Perpetuals", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "crypto-roughness", "doi": "", "date": "2026-09-03", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7397158", "access": "Source repository access terms apply", "summary": "Funding rates on perpetual futures are directional transfers, yet empirical work often enters them as absolute values. That convention can hide an economically relevant asymmetry. This paper returns to the frozen Binance BTCUSDT and ETHUSDT tape from 2020-2026 and separates lagged positive funding from the magnitude of negative funding, while leaving the increment-variance and pre-averaged apparent-H estimators unchanged. In the signed-basis specification, a one-standard-deviation increase in positive funding is associated with coefficients of -0.0058 and -0.0052 for Bitcoin and -0.0042 and -0.0041 for Ether; each remains significant after Holm adjustment. The corresponding negative-side 90% confidence intervals fall within approximately ±0.0013. Two one-sided tests also establish equivalence to zero at the disclosed, sensitivity-tested ±0.0015 benchmark. The asymmetry persists in raw units, under equal state weighting and conditional-state estimation, across alternative HAC bandwidths, and in a 499-replication 28-day moving-block bootstrap. Signed basis controls do not absorb it, and the 2020-2022 and 2023-2026 estimates show no formal break. By contrast, signed funding-by-settlement interactions and signed block-extreme coefficients fail to survive familywise-error correction. The evidence places the positive-funding result inside the increment-variance measurement family: it is a persistent observation-state association, rather than evidence about latent spot regularity, long-side inventory, causal leverage, transition information, or a trading rule.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Rough Volatility", "Funding Rate", "Perpetual Futures", "Signed Asymmetry", "Equivalence Testing", "Moving-Block Bootstrap" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7397158" }, { "id": "0031", "title": "Selection, Disclosure, and Persistence in Japanese Foreign-Exchange Intervention", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-09-02", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7378919", "access": "Source repository access terms apply", "summary": "Japanese foreign-exchange intervention is selected in unusual market conditions and followed by operations, information releases, and subsequent news. This paper organizes those stages using 69 pre-2026 episodes, daily USD/JPY, and a strictly lagged Japan–U.S. two-year rate-alignment measure. A frozen matched-counterfactual design fails its common-support gate: 13 of 50 primary risk sets survive the caliper, while 9 of 21 covariates remain materially imbalanced. The analysis therefore reports a selection equation and conditional deployment and directional-persistence associations rather than average causal effects. The selection equation detects no alignment association with entry (bootstrap p = 0.577). Conditional on intervention, the estimated first-day amount association is 25.2%, but its deployment-family Holm-adjusted p-value is 0.198. Alignment coefficients are 0.41, 0.60, and 0.83 log-change percentage points at 5, 10, and 20 days; the three-horizon Holm family does not reject (adjusted p = 0.109). Longer-horizon paths may combine operation, disclosure, and later policy or news responses; the daily design does not separately identify disclosure effects. A locked spring-2026 episode shows a large initial movement followed by substantial decay. The contribution is a transparent account of selection and conditional path evidence, with an explicit boundary between reported-result auditability and complete computational reproduction.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Foreign-Exchange Intervention", "Japanese Yen", "Persistence", "Policy selection", "Disclosure", "Common support" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-13", "ssrn": "7378919" }, { "id": "0030", "title": "AI Volatility Forecasts Under Stress: Performance, Tail Risk, and Ranking Reversals", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-09-02", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7373459", "access": "Source repository access terms apply", "summary": "Machine-learning volatility forecasts are usually judged by average out-of-sample accuracy, but their practical value depends on whether relative performance survives stress. This paper studies state-dependent forecast ranking rather than unconditional loss alone. Using the full 2,615-observation archived S&P 500 realized-volatility file preserved from the Oxford-Man Realized Library, I compare EWMA, HAR-RV, and XGBoost in a walk-forward design whose state thresholds use only information available through t−1. XGBoost outperforms HAR-RV in normal observations, but the ranking reverses under high volatility. The paired XGBoost-minus-HAR QLIKE differential is −0.0097 in normal states, +0.679 in Stress90, and +1.167 in Stress95; the Stress90-minus-Normal deterioration is +0.689. The pairwise reversal survives log-MSE and log-MAE, Holm adjustment, chronological subsamples, and a pre-specified top-5% one-day realizedvolatility-ratio state with 148 events. Moving-block bootstrap and leave-one-event-out diagnostics indicate that the jump-state result is not driven by any single observation, although the five largest events account for about 42% of the cumulative differential. By contrast, the primary Q50-to-Q90 transition definition identifies only eight events and remains descriptive under the pre-specified power gate. The evidence reveals a forecasting-risk paradox: a model can look competitive under unconditional average-error criteria while becoming materially worse when volatility is high or rising abruptly. The results motivate evaluating AI volatility forecasts by state-conditional ranking stability, not average accuracy alone.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Realized Volatility", "Machine Learning", "Xgboost", "HAR-RV", "Stress Forecasting", "Ranking Reversal" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7373459" }, { "id": "0029", "title": "Exchange-rate Intervention and the Timing of Volatility: Evidence from Japan", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-09-02", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7381318", "access": "Source repository access terms apply", "summary": "Using 69 Japanese intervention episodes, this paper separates event-day USD/JPY squared-return risk from subsequent windows. Frozen multiplicity-adjusted tests detect neither pooled level differences nor delayed risk transfer; their intervals permit positive transfer. The pooled twenty-day mean log ratio satisfies a pre-specified equivalence criterion. Four 2022–2024 yen-purchase episodes drive a descriptive historical contrast that the spring-2026 holdout does not reproduce. These selected-event estimates are noncausal. The findings distinguish directional currency adjustment from horizon-specific risk stabilization and clarify what intervention outcomes can be inferred from daily data.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Foreign-Exchange Intervention", "Japanese Yen", "Volatility Timing", "Event Study", "Equivalence Testing", "Policy Regimes" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-13", "ssrn": "7381318" }, { "id": "0071", "title": "The Carry-unwind Shortcut: JGB Curve Location, Volatility States, and Global FX Repricing", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-09-01", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7385740", "access": "Source repository access terms apply", "summary": "Market commentary often compresses a rise in Japanese government bond yields into a single causal chain: yen funding becomes more expensive, carry trades unwind, and global risk assets reprice. This paper tests the empirical content of that label. Using daily Japanese yield-curve innovations, a 12-currency exposure panel, and a separately constructed non-yen high-minus-low carry factor from 2013 through July 2026, the design separates short-end from long-duration shocks, positive from negative repricing, and high- from normal-volatility states. The carry-unwind signature is sharply conditional. In the currency panel, positive short-end shocks produce a stronger five-day loss gradient among economies with larger pre-determined yen-credit exposure only in high-volatility states; the interaction survives Holm correction (p = 0.018), and the positive-minus-negative high-volatility contrast also survives (p = 0.010). In the separately constructed carry factor, the total slope on positive short-end repricing is negative at one and five days in high-volatility states, with Holm-adjusted p-values of 0.016 and 0.018. Negative short-end and long-duration shocks are close to zero. Event paths are directionally consistent but weaker: funding-type shocks are followed by a 1.51% twenty-day carry loss versus 0.31% after duration-type shocks, with a suggestive difference (p = 0.064). A 90th-percentile volatility threshold and release-lag checks preserve the main ordering. The evidence supports a state-contingent classification, not a generic causal claim: carry unwind is not a synonym for rising JGB yields.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-02", "ssrn": "7385740" }, { "id": "0074", "title": "Conditional Portability of Geometric Stress Coordinates: A Frozen Eight-Market Test of Geometric Tail Routing", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "distributional-geometry", "doi": "", "date": "2026-08-31", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7377778", "access": "Source repository access terms apply", "summary": "The results identify a conditional cross-market portability pattern for an online Wasserstein stress coordinate. Eight equity indices are evaluated strictly out of sample from January 2007 through June 2026 under a design frozen before outcome access: three volatility filters, three tail probabilities, four memory rules, 22-trading-day updates, and a single authorized run. The primary estimand is the FZ0 loss differential between Soft-DRP routing and matched filtered historical simulation. Under the prespecified hierarchical model, the Europe-domain gate passes in five realized-volatility cells: all four European markets have favorable signs, pooled FZ0 differences remain negative after Holm correction, and Soft-DRP does not add a DQ rejection in those cells. This is a domain-level forecasting result. The same run places its scope at RV22: no EWMA(0.94) or frozen-GARCH Europe-domain cell passes the conditional gate, the non-Europe domain does not pass, and wider heterogeneity stress tests remove support for the same European conditional claim. Universal portability is not supported because every cross-market prediction interval has a positive upper endpoint. Absolute calibration remains weak, with DQ rejection in 258 of 288 Soft-DRP cells. DRP6 therefore gives geometric measurement a conditional, domain-specific tail-forecasting role and identifies the configuration in which that role appears.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-25", "ssrn": "7377778" }, { "id": "0073", "title": "Apparent Volatility Roughness and Perpetual-Futures Funding: Persistent Association, Weak Transition Information", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "crypto-roughness", "doi": "", "date": "2026-08-31", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7378018", "access": "Source repository access terms apply", "summary": "A low Hurst estimate from realized volatility can describe latent variance, measurement error, or both. This paper asks a narrower empirical question: does the apparent roughness of realized volatility align with a directly observed leverage-related state in cryptocurrency perpetual futures? Using public Binance Vision data for BTCUSDT and ETHUSDT from January 2020 through July 2026, I estimate a rolling increment-variance exponent from hourly realized variance and match it to perpetual-futures funding and the perpetual-spot basis. In six-hour update samples, a one-standard-deviation increase in absolute funding is associated with a 0.010 decrease in apparent H for Bitcoin and a 0.006 decrease for Ether. The association survives month fixed effects and alternative 1-, 5-, and 15-minute realized-variance meshes. It does not survive the inclusion of lagged apparent H, and funding does not explain changes in apparent H. Non-overlapping 14-day blocks retain the relation for Bitcoin during extreme funding episodes but not for Ether. Recursive forecasts show no daily gain; an Ether one-hour gain is statistically detectable but economically small and does not generalize. The evidence therefore supports a persistent state association, not a transition signal, structural estimate of spot regularity, or trading rule. The distinction clarifies what a rolling roughness estimate can and cannot identify in noisy high-frequency markets.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-02", "ssrn": "7378018" }, { "id": "0072", "title": "Dynamic Persistence in Yen Market Operations: Pre-Intervention Rate Conditions and Exchange-Rate Reversal", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-08-31", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7380619", "access": "Source repository access terms apply", "summary": "Japanese foreign-exchange intervention is often evaluated by the exchange-rate movement on the intervention day. This paper argues that the relevant policy outcome is the persistence of the intendeddirection response after the initial operation. Using the official intervention chronology, daily USD/JPY observations, Japanese and U.S. short-rate measures, and strictly lagged market controls, the analysis constructs 69 pre-2026 intervention episodes and separates first-day response, continuation, decay, and reversal. Episodes preceded by an aligned short-end yield-curve state show mean intended-direction responses of 0.78 percentage points on the intervention day and 2.06 percentage points at the 20-day horizon, compared with 0.26 and -0.05 percentage points for misaligned episodes. Conditional regressions associate a one-standard-deviation increase in alignment with responses of 0.41, 0.60, and 0.83 percentage points at 5, 10, and 20 days. Raw wild-bootstrap p-values are near conventional thresholds, but Holm adjustment across the frozen horizon family yields 0.109. Alignment is also associated with a 25.2% larger first-day operation and lower continuation odds, although the two-outcome family does not reject. The estimates therefore support a positive state-dependent persistence gradient, not a confirmed causal effect of intervention. The findings show why an immediate currency movement should not be treated as evidence of durable policy effectiveness.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-02", "ssrn": "7380619" }, { "id": "0077", "title": "Calendar Congestion and Trading Responses: Cornerstone Lock-up Expirations in Hong Kong IPOs", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-30", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7374679", "access": "Source repository access terms apply", "summary": "Contractual lock-ups make previously restricted shares eligible to trade on dates fixed months in advance. This paper asks whether the trading response to a Hong Kong IPO cornerstone release depends on other releases reaching the market on the same session. Congestion is formed from a frozen pre-outcome census of 504 eligible release events; the focal sample contains 226 complete event windows for 2022–2025 listing cohorts. The primary equal-event specification models log(1 + event-session volume relative to the preceding-20-session mean), retains 11 zero-volume sessions, and controls for market turnover, issuer size, and pre-event liquidity. The crowded-session coefficient is 0.432 (event-session clustered p = 0.010) and its interaction with log(1 + release strength) is −0.683 (p = 0.018). Event-week clustering, wildcluster bootstrap inference, month-constrained calendar randomization, alternative same-session congestion measures, placebo dates, and joint omission of the most crowded sessions support the negative interaction. The result is scale-sensitive: asinh and positive-only log outcomes agree, but an untransformed volume-ratio model and a Poisson pseudo-maximum-likelihood share-flow model do not. Holm adjustment across volume, return, and volatility interactions gives p = 0.054; neither price nor range volatility is robust. The evidence therefore describes a typical-event transformed-volume pattern, strongest among smaller releases. It does not establish aggregate share-flow pressure, price dislocation, or a causal market-capacity mechanism.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-01", "ssrn": "7374679" }, { "id": "0076", "title": "Persistence Without Predictive Advantage: Sponsor Histories and Cornerstone Adoption in Hong Kong IPOs", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-30", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7374758", "access": "Source repository access terms apply", "summary": "Do repeated IPO intermediaries exhibit stable contracting histories, and do those histories improve genuine out-of-time prediction? We study 930 Hong Kong Main Board operating-company IPOs from 2018 through August 2026, 1,524 IPO–sponsor relationships, and 113 canonical sponsor entities. Sponsor histories use only strictly earlier listing dates. At the primary five-prior-deal threshold, the team adoption score has a raw AUC of 0.621 (year-stratified 95% bootstrap interval [0.571, 0.668]; within-year permutation p = 0.0004), while sponsor rankings persist across chronological halves (Spearman ρ = 0.610, p < 0.001). The controlled linear-probability coefficient is 0.390. Its p-value is 0.016 with leadsponsor clustered inference but 0.068 under a 4,999-draw wild-cluster bootstrap. Among single-sponsor IPOs the coefficient is 0.298 (clustered p = 0.081; wild-bootstrap p = 0.147), and omitting CICC reduces the full-team coefficient to 0.263 (p = 0.114). Expanding-window forecasts provide no stable incremental advantage: adding sponsor history changes pooled AUC from 0.691 to 0.682 (difference −0.009; bootstrap interval [−0.035, 0.017]). The difference is +0.030 in 2020–2023 but −0.040 in 2024–2026; the formal history-by-post-2024 interaction is imprecise (p = 0.660). Histories also do not predict release intensity or multi-tranche structure. Sponsor histories are persistent descriptors, but their association is dependence-sensitive and their predictive content becomes redundant as cornerstone adoption diffuses. The evidence is consistent with matching, networks, or routines rather than an exogenous sponsor effect or a portable forecasting advantage.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-01", "ssrn": "7374758" }, { "id": "0075", "title": "Geometric Stress Coordinates for Conditional Tail Forecasting: Soft Distributional Routing across Equity Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "distributional-geometry", "doi": "", "date": "2026-08-30", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7375598", "access": "Source repository access terms apply", "summary": "The results show that an online Wasserstein stress coordinate contributes to one-step tail forecasting after current volatility scale is filtered. Daily returns for the S&P 500, FTSE 100, Nikkei 225, and Hang Seng from 2000 through June 2026 are evaluated strictly out of sample from January 2007. Frozen pre-2007 distributionalregime thresholds softly route observations between calm and stress empirical Value-at-Risk (VaR) and Expected Shortfall (ES) experts. Under the original 22-day realized-volatility filter, soft DRP routing improves pinball loss in all eight market-tail cells, with FTSE at 1% and HSI at 5% surviving cross-market Holm correction. The joint FZ0 loss also improves in seven of eight cells, with the FTSE 1% gain surviving correction; the same corrected FTSE 1% joint-loss result is retained under the RiskMetrics EWMA filter. The prespecified extensions show that the magnitude of the gain depends on scale filtering, market, and training memory: frozen GARCH is weaker and a 1,000-day rolling window reverses the SPX pattern. Soft routing adds comparative tail-loss information without systematically resolving absolute conditional-coverage or DQ failures. The evidence gives geometric stress coordinates a positive but conditional role in one-step VaR-ES forecasting.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-25", "ssrn": "7375598" }, { "id": "0078", "title": "Large-Load Tariff Decisions and Utility Market Responses", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "energy-compute-infrastructure", "doi": "", "date": "2026-08-29", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7367558", "access": "Source repository access terms apply", "summary": "U.S. regulators are adopting large-load tariffs to allocate the financial risk created when utilities build long-lived infrastructure for data centers and other concentrated loads. This paper studies 15 regulator-verified, exact-dated first final tariff decisions covering 13 listed utility parents and 13 states from 2022 through 2026. Customer-specific special-contract approvals are excluded from the primary sample. Expected returns are estimated over trading days [-120,-21]; the primary event window is [-1,+1]. Repeated decisions are collapsed to listed-parent means for exact sign-flip inference. The mean SPY-model cumulative abnormal return is 0.39 percentage points (p = 0.353; bootstrap 95 percent interval: -0.40 to 1.14), providing no evidence of average repricing under the primary model. The XLUadjusted [-1,+1] estimate is positive, but alternative timing windows and influence checks weaken it, making the valuation conclusion benchmark-, window-, and influence-sensitive. A transparent mean-adjusted log-volume measure rises by 0.150 log points, or approximately 16.2 percent. The parent-level result is positive (p = 0.028) but does not survive event-date aggregation (p = 0.110). The log post/pre realized-risk ratio is not distinguishable from zero. The evidence is consistent with, but does not establish, a dependence-sensitive attention response; valuation conclusions are not robust across specifications. The design cannot identify project-level incidence or downstream ratepayer effects.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-01", "ssrn": "7367558" }, { "id": "0032", "title": "Not All JGB Yield Shocks are Carry Shocks: Curve Location and the Global Transmission of Japan’s Rate Normalization", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "japan-macro-finance", "doi": "", "date": "2026-08-29", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7363659", "access": "Source repository access terms apply", "summary": "Rising Japanese government bond (JGB) yields are often treated as direct evidence that yen-funded carry trades are becoming uneconomic and that Japanese capital will return from foreign assets. That interpretation conflates curve location, relative currency sensitivity, and aggregate balance-sheet adjustment. This paper separates a daily short-end proxy from an orthogonalized long-duration innovation over January 2013-July 2026 and estimates transmission through predetermined BIS yen-credit and IMF portfolio-holdings gradients. The primary sample contains eight practically tradable currencies; four managed, restricted, or disrupted currencies enter only an extended sample. Hierarchical interaction models include the time-varying exposure levels, and 999 moving-block replications re-estimate both stages of long-duration inference. The primary linear short-end gradients are statistically equivalent to economically negligible effects at H0, H1, and H5. Directional exposure and average-carry estimates remain inconclusive. Positive-shock high-volatility amplification survives an expanded fifteen-test Holm family at H1, while the direct positive-minus-negative volatility contrast survives its three-horizon family at H5. These results do not coincide at a common horizon and therefore cannot satisfy the paper's joint carry-unwind rule. The sole full-family rejection is also sensitive to expanding the family from 18 to 19 tests. Long-duration gradients and weekly local projections through eight weeks do not confirm portfolio substitution or later repatriation. The evidence supports limited historical state dependence, but neither a systemic carry unwind nor confirmed capital return.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Bank of Japan", "JGB Yield Curve", "Yen Carry Trade", "Global Liquidity", "Portfolio Rebalancing", "International Spillovers" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7363659" }, { "id": "0079", "title": "The Half-Life of a Futures Contract: Technological Obsolescence and Vintage Risk in Compute Derivatives", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-27", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7359042", "access": "Source repository access terms apply", "summary": "A futures contract has a legal maturity, but it can lose economic relevance before that date. This paper develops the concept of contract half-life for derivatives written on rapidly evolving technological vintages. Contract half-life is the time required for a hardware-specific contract's economic relevance - measured through commercial exposure, trading liquidity, and price-discovery importance - to fall to onehalf of its reference level. The distinction matters for compute derivatives because the announced H100 and B200 futures are tied to named GPU generations whose commercial importance can migrate faster than their rental prices disappear. A simple vintage-migration model separates rental-price depreciation from benchmark-relevance depreciation and shows how innovation intensity, maturity, and liquidity network effects jointly determine economic expiry. The model yields four testable propositions: activity should migrate to a new vintage faster than the old rental index declines; distant maturities of an oldvintage contract should lose viability first; faster innovation should shorten contract half-life; and rolling or performance-adjusted benchmarks should extend durability while increasing conversion-model and governance risk. The paper proposes observable half-life measures based on open interest, volume, information shares, and commercial benchmark adoption, and explains why a still-published settlement index is insufficient evidence that a contract remains economically alive. Because these propositions require post-launch data, the paper also uses three completed benchmark transitions - the shift of a gasoline contract to an RBOB basis, the ultra-low-sulfur respecification of heating oil, and the inclusion of WTI Midland in Dated Brent - plus unsuccessful DRAM-futures proposals as identification templates. These analogues distinguish benchmark replacement, internal respecification, constituent revision, and launch failure; they are not estimates or formal bounds on compute-contract half-life.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-28", "ssrn": "7359042" }, { "id": "0082", "title": "From Guaranteed Allocation to Deferred Float: A Contract-level Census and Event Evidence on Cornerstone Unlocks in Hong Kong IPOs, 2018-2026", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-24", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7343039", "access": "Source repository access terms apply", "summary": "Cornerstone investors in Hong Kong initial public offerings receive guaranteed allocations in exchange for a disclosed six-month lock-up, so each tranche converts a known share quantity into a contractually dated expansion of tradable supply. This paper constructs and audits the complete population of these contracts — 930 Main Board operating IPOs from January 2018 through August 2026, of which 583 (62.7%) carry cornerstone tranches mapping into 596 dated release events — and measures unlock outcomes on 231 fully audited events of the 2022–2026 cohorts, net of a systematic screen for voluntary lock-up extensions. A pre-outcome estimation gate, built on reconciling public float to immediately tradable float through allotment results, over-allotment exercises, and corporate actions, identifies 504 clean occurred events. First, deferred-supply intensity has risen: the annual median release ratio — newly eligible cornerstone shares over immediately tradable pre-release float — roughly doubled, from about 0.25 in 2018–2019 to 0.52 –0.60 in 2024–2026 (trend p < 10⁻¹⁰); both margins moved materially, with the numerator — cornerstone books relative to offer size — showing the larger change. Second, release days sharply amplify trading activity — mean volume over pre-release float roughly quadruples (p < 10⁻⁵) — and roughly double the probability of extreme daily moves. Third, prices show a small negative release-day displacement relative to each stock’s own pre-release drift (−0.58 percentage points, at the 5% significance boundary); equivalence within ±1 percentage point cannot be established, and the surrounding declines begin weeks before eligibility, consistent with younglisting drift and/or anticipatory pricing.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-07", "ssrn": "7343039" }, { "id": "0081", "title": "THE 40-50-10 IPO: Regulatory Allocation Floors and Contractual Corner Solutions in Hong Kong", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-24", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7343758", "access": "Source repository access terms apply", "summary": "Hong Kong’s August 2025 IPO reform introduced a minimum 40% non-cornerstone bookbuilding allocation and a choice between a clawback-based public tranche and a no-clawback alternative. Using prospectus-level coding for all 161 operating IPOs in the first post-reform year, together with exact final allocation compositions for 146 issues, we document rapid coordination on two nested corners. Mechanism B became the de facto standard outside the specialist-technology regime: 139 of 141 non-Rule 18C offerings selected it, and every Mechanism B issuer began with the 10% minimum public tranche. Among Mechanism B offerings with cornerstone investors, 32 of 112 chose an initial 40–50–10 split across non-cornerstone bookbuilding, cornerstone, and public tranches. Its incidence rises from 10.7% in the smallest proceeds quartile to 60.7% in the largest. The evidence shows how a regulatory menu can generate a market standard and, within it, a size-related contractual corner. The design is descriptive rather than a sharp regression discontinuity.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-07", "ssrn": "7343758" }, { "id": "0080", "title": "Beyond the Coupon: Climate Risk and the Quantity-Maturity Design of Chinese Interbank Bonds", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-08-24", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7345743", "access": "Source repository access terms apply", "summary": "This paper examines whether climate risk reaches the primary bond market through contract margins that coupon regressions do not observe: the amount placed, the initial maturity, the use of perpetual structures, and the decision to issue at all. The sample contains 1,468 Chinese interbank medium-term notes issued by 165 listed legal entities during 2010–2025. Three distinct exposures—registered-location weather, national emissions-trading-system (ETS) coverage matched by legal entity and date, and climate-risk language in the latest available annual report—enter issuer and issue-month fixed-effects models. None of the nine principal transaction-level coefficients survives issuer-clustered wild-bootstrap inference and multiplicity correction. The nulls are not equally informative: equivalence tests exclude effects beyond ±10% only for the language measure in the issue-size and maturity equations, while most remaining intervals are too wide to establish economically small effects. ETS estimates depend on changes among only 17 transaction-level issuers. In the zero-inclusive annual panel, issuance rises before firms enter ETS scope as well as afterward, ruling out a causal policy interpretation. The contribution is therefore a mapped identification boundary: observable climate exposure does not generate a common within-issuer shift across these contract margins, but the design also shows precisely where measurement, support, and statistical power prevent stronger conclusions.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-09-04", "ssrn": "7345743" }, { "id": "0033", "title": "The Inference Problem in Financial AI: Forecasting, Portfolios, Market Quality, and Systemic Resilience", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-24", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7328678", "access": "Source repository access terms apply", "summary": "Financial AI is commonly evaluated at the model or forecasting level, while its economically consequential effects arise through portfolio construction, market interaction, and stress-state behavior. Evidence at these layers is not interchangeable. This paper develops a cross-layer inference framework that separates five economic objects—AI capability, volatility forecasting, portfolio performance, market quality, and systemic resilience—and treats each transition as a distinct empirical question. Recent evidence reveals three recurring divergences: greater general model capability does not guarantee superior volatility forecasts; attractive AIgenerated portfolios can arise without superior forecast skill; and lower price variation can coexist with deteriorating liquidity. The framework shows why these divergences are economically meaningful rather than merely terminological: optimization can alter the value of forecasts, aggregation can transform private decisions into market-wide order flow, and liquidity or funding constraints can change normal-market relationships under stress. A structured evidence audit further reveals an identification gradient, with comparatively dense predictive evidence and much thinner evidence linking adaptive AI to systemic outcomes during real market stress. The paper does not presume that AI is stabilizing or destabilizing. It specifies where inference stops and what additional evidence is required before an upstream performance gain can support a downstream financial conclusion.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Artificial Intelligence", "Financial Volatility", "Market Stability", "Algorithmic Markets", "Systemic Risk" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7328678" }, { "id": "0084", "title": "Hidden in the Stack: Shared AI Infrastructure, Common-Mode Failures, and Potential Operational Contagion in Global Banking", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-21", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7323798", "access": "Source repository access terms apply", "summary": "Shared technology suppliers can create common operational exposure across banks, but public evidence rarely identifies when that exposure becomes a cascade or financial contagion. This paper develops a layered, disclosure-based framework that separates four objects: common exposure, common-mode operational shock, potential operational cascade, and realised financial contagion. The main sample is the complete 2025 Financial Stability Board list of 29 global systemically important banks. The source register contains 44 public disclosures and yields 43 unique institution-supplier relationships involving 17 external suppliers through 20 August 2026. Microsoft reaches 10 banks in the all-stage network and Google eight; however, a production-only sensitivity produces an eight-eight tie, showing that deployment-stage classification changes the apparent leader. Deployment-specific look-through paths expand OpenAI's structural reach from three direct institutions to six. Yet only seven of 43 relationships state criticality, six disclose fallback or substitutability, and one reports exit testing. Missing-edge analysis further shows that only three adversarially placed Google relationships would change the leading supplier, while the same number of undisclosed relationships can move concentration in opposite directions depending on whether missingness favours incumbents or challengers. The evidence therefore supports a propagation-ready topology, not a claim of realised financial contagion. The paper's contribution is to identify the information and network conditions required to move from visible co-exposure to a defensible systemic-risk assessment.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7323798" }, { "id": "0083", "title": "Disclosure Stops Before Resilience: Public Evidence on AI Dependencies at Global Systemically Important Banks", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-21", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7326518", "access": "Source repository access terms apply", "summary": "This study examines how far public information allows external observers to move from a bank's visible artificial-intelligence (AI) use to disclosure relevant to dependency resilience. We audit all 29 banks on the Financial Stability Board's 2025 G-SIB list using a frozen twocorpus protocol. The bank, not an individual service path, is the unit of analysis. The eightelement sequence covers AI materiality, concrete use, external-stack visibility, AI-specific risk, linked dependency, common dependency, substitutability, and tested resilience. Under the strict AI-specific rule, 23 banks reach AI-risk recognition, seven reach linked dependency, and three reach common dependency; under the broader enabling-technology rule, 22 reach dependency and six reach common dependency. No bank completes either sequence through tested resilience. Two blinded independent model recodings agree on 87.4% of 348 institution-item decisions (pooled Cohen's kappa = 0.719; Gwet's AC1 = 0.771), and both independently end at zero complete tested sequences despite different intermediate counts. The evidence concerns public assessability within the specified corpus, not internal control quality or a verified service-level dependency path. The study identifies a bank-level linkage gap and proposes separate minimum disclosures for public comparability and confidential supervisory collection.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-24", "ssrn": "7326518" }, { "id": "0034", "title": "Futures Before Fungibility: Benchmark Governance and Regulatory Friction in GPU Compute Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-21", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7324139", "access": "Source repository access terms apply", "summary": "Benchmark design shapes whether a derivatives market forms and how much disruption it can withstand. We develop a participation model of compute derivatives in which benchmark granularity improves exposure matching but divides counterparties, while technological abstraction exchanges vintage mismatch for conversion uncertainty. Eight propositions with proofs establish a critical participation threshold and show that the margin above it is exactly the largest common adverse benefit shock preserving a positive equilibrium. Shared liquidity supports finer benchmarks. Governance expands formation capacity but can favor finer or coarser menus. Strong demand on one side cannot compensate for inadequate counterparty participation. Official records and 165 published rental-index observations establish the institutional setting and segmentation. Analytical proofs and assumedparameter illustrations establish the mechanisms. Benchmark architecture, governance and liquidity integration jointly determine market formation and resilience.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Compute Futures", "GPU Rental Markets", "Benchmark Design", "Basis Risk", "Technological Obsolescence", "Commodity Financialization" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-10-01", "ssrn": "7324139" }, { "id": "0085", "title": "Beyond Assets under Management: A Comparative Framework for Measuring Fiduciary Service Quality", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-20", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7320922", "access": "Source repository access terms apply", "summary": "Assets under management (AUM) is widely used to describe the reach of fiduciary institutions, yet it measures the volume placed inside an arrangement rather than the quality with which entrusted functions are performed. This structured comparative review examines fiduciary architectures in the United States, the United Kingdom, Japan, Singapore, and China across personal trusts, pensions, collective investment arrangements, corporate trusteeship, and restructuring uses. It develops a function–state–evidence framework that evaluates five core functions—asset segregation, administration, discretion, monitoring, and transition—under normal operations, conflict, stress, and termination. Evidence is organized into legal design, operating process, and realized outcomes. Seven quality dimensions follow: legal integrity, mandate fidelity, competence, operational reliability, transparency, accountability, and resolvability. The review argues against a universal scalar ranking because fiduciary mandates differ in purpose and state dependence. It instead proposes a modular dashboard with minimum gates, functionspecific indicators, and stress-state tests. The framework separates fiduciary capacity from asset volume and provides a common measurement language for regulators, beneficiaries, researchers, and service providers.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-21", "ssrn": "7320922" }, { "id": "0035", "title": "Who Absorbs Climate Risk? Institutional Transmission and Risk Sharing in Chinese and U.S. Bond Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-08-20", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7304978", "access": "Source repository access terms apply", "summary": "Bond-market climate research usually asks whether exposure changes yields. That is incomplete because a yield is only one point in a chain through which losses may be recognized, delayed, insured, guaranteed, refinanced or transferred to public balance sheets. This structured review compares Chinese corporate bonds and U.S. municipal bonds through five layers: information, pricing, contracts, liquidity and public support. U.S. municipal studies more directly identify the roles of maturity, insurance, pledged revenues, mitigation and federal assistance. Chinese studies more often document environmental or climate-related financing effects where ratings, state ownership and guarantees alter risk visibility and loss allocation. The evidence does not rank either market as more resilient; it shows that similar hazards reach investors through different obligors, information systems and loss-sharing arrangements. Companion diagnostics illustrate measurement boundaries and are not pooled cross-country estimates. An absorber is the private or public balance sheet that contractually, institutionally or de facto bears part of a climate-related loss. The review proposes an auditable architecture linking legal-entity exposure, disclosure, pricing, credit enhancement, liquidity surveillance and contingent fiscal support, plus a feasible comparative design using common standardized hazards and outcome families estimated separately within each market.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Climate Risk", "Corporate Bonds", "Municipal Bonds", "Risk Sharing", "Credit Enhancement", "Public Support" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7304978" }, { "id": "0087", "title": "Scaling into Low-Yield Trust Services: Business-Mix Transformation and Fee Compression in China's Trust Industry", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-19", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7313279", "access": "Source repository access terms apply", "summary": "China’s trust industry has returned to rapid asset growth, yet the fee income generated by that asset base has continued to contract. This article asks whether the divergence is visible within institutions and whether the post-2023 shift toward standardized asset management and asset-service mandates offers a plausible accounting explanation. I assemble a balanced panel of 12 trust companies with comparable public disclosures for 2021–2025 (60 firm-years). The sample’s year-end assets under management nearly doubled, rising by 95.8%, while fee or trust-business income fell by 42.4%; aggregate fee density declined by 70.6%, and every sample institution recorded a lower density in 2025 than in 2021 (exact two-sided sign-test p = 0.00049). In a first-difference log specification with year effects, the estimated fee-income elasticity with respect to assets is 0.572 (firm-clustered SE 0.187). Exact Rademacher wildcluster inference rejects a zero elasticity (p = 0.0396), while rejection of proportional scaling is borderline (p = 0.0566). The corresponding firm- and year-fixed-effects estimate in levels is 0.714 (SE 0.273) and does not reject unit elasticity. I therefore treat sub-proportional scaling as suggestive rather than definitive. Mechanism evidence is stronger at the accounting level: at Minmetals Trust in 2023, innovative business represented 73.6% of the two disclosed categories’ assets but 16.4% of their trustbusiness income, with an 8.82 basis-point yield versus 125.23 basis points for traditional non-standard business. Corroborating disclosures from FOTIC and China Resources Trust show that recent scale is concentrated in securities, administrative, and other non-financing mandates. The findings recast declining fee density as a business-mix consequence of regulatory transition, not as proof that larger trust companies are less profitable or that the reform caused the decline.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-28", "ssrn": "7313279" }, { "id": "0086", "title": "Likelihood Fit Is Not Directional Fit: A Time-arrow Diagnostic for GARCH Models", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-08-19", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7314379", "access": "Source repository access terms apply", "summary": "Volatility models are normally selected by likelihood, information criteria, residual diagnostics, and forecast loss. These criteria need not reveal whether a model reproduces the temporal direction of risk. We compare Gaussian GARCH, Student-t GARCH, Student-t GJR-GARCH, and Student-t EGARCH using daily S&P 500, FTSE 100, Nikkei 225, and Hang Seng returns from 2000 through June 2026. Model adequacy is evaluated against a 1–22-day time-arrow profile formed from return–variance mixed moments. EGARCH wins AIC and BIC in every market, whereas GJR-GARCH has the lowest directional RMSE in the United States, United Kingdom, and Japan. Bootstrap intervals do not resolve the GJR– EGARCH gaps in those markets, so this is a criterion-dependent point-ranking reversal, not evidence of GJR dominance. Separately, the best asymmetric model reduces full-sample RMSE by 18.6%–74.6% relative to the best symmetric benchmark, and these asymmetric-versus-symmetric intervals exclude zero in all four markets. Raw Monte Carlo diagnostics reject both asymmetric models only for the Hang Seng, while Holm and false-discovery-rate corrections appropriately weaken several model-level rejection claims. In fixed-parameter 2019–2026 and reviewer-motivated 2021–2026 evaluations, asymmetric models remain directionally closer than symmetric GARCH but are not uniformly adequate. Likelihood fit and directional fit therefore answer different questions: a model can explain conditional scale and heavy tails while attenuating how negative returns propagate into future variance.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-20", "ssrn": "7314379" }, { "id": "0049", "title": "Accurate but Correlated: Common AI Forecast Errors and Endogenous Market Volatility", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-19", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7298579", "access": "Source repository access terms apply", "summary": "Artificial intelligence can improve each institution's risk forecast while reducing the financial system's ability to diversify forecast errors. This paper separates these two objects. Individual accuracy is a marginal property of an error distribution; systemic model risk is a covariance property. With N institutions, equicorrelated log-risk forecast errors leave an aggregate variance floor of ρσ² that does not vanish as N grows. We derive the exact threshold above which an AI technology with lower individual mean-squared error nevertheless creates greater aggregate forecast risk than noisier independent legacy models. Risk-sensitive position rules then map common errors into synchronized net demand. With linear price impact, forecast-error covariance enters return variance; with GARCH updating, the initial variance increment is propagated by the multiplier (1-β)/(1-α-β). Prospectively separated simulations validate the signs under smooth volatility targeting and a discrete risk limit. Holding marginal error variance fixed, raising cross-institution error correlation from zero to 0.8 increases return variance by 3.15% under the smooth rule and 6.26% under the discrete risk limit. The stronger precommitted claim that both variance and one-percent expected shortfall rise by at least 5% is rejected. Common errors robustly amplify two-sided volatility under the model's conditions, but symmetric errors do not necessarily create comparably large downside-tail losses. Model diversity should therefore be evaluated through error covariance rather than model counts, vendor counts, or average predictive accuracy.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Artificial Intelligence", "Common Model Risk", "Forecast-Error Correlation", "GARCH", "Endogenous Volatility", "Model Diversity" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7298579" }, { "id": "0037", "title": "The Same Distribution, a Different Direction: Temporal Irreversibility in Global Equity Returns", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "distributional-geometry", "doi": "", "date": "2026-08-19", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7293060", "access": "Source repository access terms apply", "summary": "Reversing a return series leaves its empirical distribution unchanged. Its mean, variance, skewness, kurtosis, quantiles, and Wasserstein distance from the original sample are therefore identical. Yet the risk process need not be unchanged, because negative returns may precede volatility expansions more often than volatility expansions precede negative returns. We use this exact distribution-preserving transformation to study daily S&P 500, FTSE 100, Nikkei 225, and Hang Seng returns from January 2000 through June 2026. Temporal direction is measured by the difference between two mixed third moments, averaged over lags one through five. The forward-time statistic is positive in all four markets—0.339, 0.317, 0.320, and 0.364—and remains significant after Holm correction across the four primary tests. Reversing each series changes the statistic's sign while leaving marginal Wasserstein distance exactly zero. Randomizing return signs while preserving the magnitude sequence centers the statistic at zero in 2,000 replications (p = 0.0005 in every market). The direction remains positive across bandwidth, bootstrap, past-only standardization, lag, episode-deletion, subsample, and winsorization checks. Of 36 prespecified robustness tests, 31 pass a 5% Benjamini–Hochberg false-discovery-rate threshold and 13 survive the stricter Holm family-wise correction. Distributional risk measurement is therefore incomplete without an ordering-sensitive layer: the same return distribution can encode a different direction of risk propagation.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Time Irreversibility", "Asymmetric Volatility", "Leverage Effect", "Distributional Risk", "Wasserstein Distance", "Global Equity Markets" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7293060" }, { "id": "0088", "title": "Trust as Resolution Infrastructure: Value Preservation, Creditor Coordination, and Bankruptcy Service Trusts in China", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-18", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7308278", "access": "Source repository access terms apply", "summary": "Bankruptcy service trusts have become an increasingly common component of large corporate restructurings in China. Among the 144 listed companies whose reorganization plans were approved by courts between 2007 and 2025, the share of plans repaying creditors with trust beneficial interests rose from 4 percent over 2019–2022 to 36 percent in 2024 and 40 percent in 2025. Yet the economic contribution of these trusts is easily misstated: restructuring plans may treat trust units as debt satisfaction while underlying assets remain unmonetized, whereas evaluations based only on immediate cash ignore the value of avoiding forced sales. This paper first develops a two-period model of deferred disposition in which the fire-sale discount, expected asset recoverability, and governance-dependent agency costs jointly determine when a service trust preserves value; four hypotheses follow as comparative statics, including a net-effect boundary for asset complexity. It then combines a censusbased sampling frame compiled from public plan records and professional annual reports, a sourcegraded audit of 20 cases, a focal positive mechanism analysis of the Founder Group restructuring, and a structured comparison of the 2024 listed-company cohort. Of the 20 audited cases, 12 have a verified plan, seven a verified establishment, and only two provide case-level monetary evidence meeting the audit threshold: Bohai Steel reports cumulative cash distributions, while HNA Group’s trust units are reported at a Level 3 fair-value carrying amount in affiliated companies’ audited financial statements. Because realized trust distributions remain sparse, the paper specifies a package-level market-implied recovery measure and an exploratory roadmap of three prospective research designs built on the 2023 trust-business classification reform, the 2024 Supreme People’s Court and CSRC Minutes, and withincase creditor-class variation, stating explicitly the conditions under which each would support causal inference. The evidence to date supports an institutional conclusion: where viable operations can be TRUST AS RESOLUTION INFRASTRUCTURE separated from difficult but potentially recoverable assets, a service trust preserves contingent creditor rights and allows the core restructuring to close without a forced sale. Whether it raises discounted recovery is the empirical hypothesis this design makes testable and auditable.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-21", "ssrn": "7308278" }, { "id": "0038", "title": "Faster Is Not Calmer: Agentic AI, Reaction-Time Compression, and the Term Structure of Volatility", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-18", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7286838", "access": "Source repository access terms apply", "summary": "Artificial intelligence can shorten investors’ responses to public news without making their desired positions more alike. This paper isolates the consequences of changing when a fixed information-driven order reaches the market. Investors first complete a decision stage and then split their orders over time. With exponential decision and execution delays, aggregate arrival concentration equals 1 / [ 2 ( τ D + τ E ) ), where the parameters are the decision and execution time scales, respectively. Faster decisions therefore concentrate arrivals even with execution smoothing. For a fixed aggregate order and permanent price-impact coefficient, greater temporal concentration raises short-horizon realized variation. It also weakly increases cumulative flow beyond a fixed liquidity capacity. Temporary displacement later reverses, allowing faster price discovery and sharper event-time volatility to coexist with an unchanged long-run informational return. In the threshold model, stress depends on recovery relative to arrival speed and on capacity relative to peak flow. An adaptive-supply extension identifies how faster recovery and greater capacity can offset compression. The variance clock shifts toward the announcement in all 36 reported noise–capacity configurations. A participant-level empirical design tests the central distinction by examining whether response times become more concentrated while aggregate order size and the correlation of eventual signed position changes remain stable.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Agentic AI", "Volatility", "Market Microstructure", "Price Discovery", "Liquidity Resilience", "Algorithmic Trading", "Reaction Time" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7286838" }, { "id": "0091", "title": "Three-Layer Climate Risk Absorption in China's Corporate Bond Market: Diagnostic Evidence from Interbank Medium-Term Notes", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-08-17", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7297858", "access": "Source repository access terms apply", "summary": "We study the institutional channels through which climate risk is absorbed before reaching the primary-market coupon of Chinese interbank medium-term notes (MTNs). We develop a three-layer absorption framework\u0016credit ratings, third-party guarantees, and implicit state support\u0016and implement its diagnostic protocol on a fully public, reproducible 12-city prototype that combines 4,130 NOAA station-day weather records for 2024 with 1,667,274 ChinaBond rating-curve observations for 2020\u00152025. Three results emerge. First, rating-implied funding costs are sharply compressed at the AAA end (2.23\u00152.31%) and dispersed at the A end (8.56\u001510.90%), mirroring the issuer-level rating compression documented for the 1,468-MTN universe. Second, physical climate exposure sorts signi cantly across all three institutional layers: AAA status, state ownership, and guarantee ags each separate climate risk at the 5% level, and all three comparisons survive Benjamini\u0015Hochberg falsediscovery-rate correction. Third, a one-standard-deviation increase in physical climate risk is associated with a 3.01-percentage-point higher rating-implied coupon (R2 = 0.67; HC1 p < 0.001) that is robust across seven pre-speci ed variations of the extreme-day thresholds, the yield tenor, the polar-city sample, mid-tier rating assignments, and CRI weights, and the guarantee and rating interactions carry the predicted negative absorption signs. A 5,000-draw placebo permutation con rms that the interaction sign template is not an artifact of arbitrary ag assignment, while making explicit that con rmatory interaction inference requires the 1,468-MTN companion sample. The results locate the climate-pricing null of primary markets in the rating\u0015guarantee\u0015 ownership architecture rather than in investor inattention, with direct implications for disclosure-based climate policy.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-19", "ssrn": "7297858" }, { "id": "0090", "title": "Trading Through the Shock: Physical Climate Events and Transaction Continuity in China's Corporate Bond Market", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-08-17", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7299058", "access": "Source repository access terms apply", "summary": "Physical climate risk can raise expected credit losses without necessarily causing an immediate breakdown in market functioning. This paper studies whether localized extreme-heat and extreme-rain episodes disrupt transaction continuity in China's exchange-traded corporate bond market. Public daily histories for 513 bonds issued by 93 exactly matched listed legal entities are linked to 1,011 issuerspecific weather events across 33 registered-office locations. The original matched event design finds no common deterioration in zero trading, traded-bond share, or reported volume; pooled five-day onesided harm limits are +2.1 percentage points, -1.7 percentage points, and -20.8%, respectively. I add a transaction-latency module that distinguishes a recently observed price from subsequent trading resumption. Among bonds quoted within the preceding 60 market sessions, the difference in 60session resumption probability is +0.1 percentage points against same-date location controls, -1.1 points against an adjacent-year same-bond placebo, and -0.2 points when both adjacent years are averaged. The corresponding one-sided lower harm limits are -3.4, -3.2, and -3.5 points, excluding a decline larger than the existing five-point monitoring candidate in all three designs. Five- and twentysession estimates are less precise. Bounded price-adjustment diagnostics are sensitive to time trend and do not establish prompt price discovery. The evidence supports persistent transaction function in a selected segment, while motivating joint regulatory monitoring of zero trading, price age, and time to resumption.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-24", "ssrn": "7299058" }, { "id": "0089", "title": "From City Proxies to Legal Entities: How Measurement Choices Change Climate-risk Inference in Chinese Corporate Bonds", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-08-17", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7301319", "access": "Source repository access terms apply", "summary": "Empirical climate-finance results often travel from geographic proxies to firm-level claims without a direct audit of what changes during that transport. This paper conducts such an audit for Chinese corporate bonds. Its starting point is a deliberately constructed twelve-city diagnostic, not a bondpricing estimate: 2024 weather indicators and assigned credit categories generate a positive association with a rating-implied three-year coupon. The design is then transported to a frozen universe of 1,468 Chinese interbank medium-term notes issued by 165 listed legal entities during 2010–2025. Four bridges are examined separately: geographic support and aggregation, constructed versus observed outcomes, benchmark-yield decomposition, and weather-source completeness. Under harmonized outcomes and exposures, city-year, issuer-year, and bond-level estimates are all negative and statistically indistinguishable from zero. In the five-city common-support sample, the observed-coupon coefficient is statistically detectable under restricted wild-cluster inference (p=0.0225), whereas the corresponding issue-premium estimate is only marginal (p=0.0710). Complete ERA5 temperature data preserve a positive association inside the prototype. Separately, a literal city-specific all-day precipitation-percentile count has no cross-city variation under complete daily coverage. The evidence shows that the prototype and full-sample results are not contradictory: they answer different estimands. City proxies are useful diagnostic devices, but their coefficients should not be transported as legal-entity bond-pricing estimates without explicit outcome, support, and data-source bridges.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-19", "ssrn": "7301319" }, { "id": "0093", "title": "From Black-Box AI to GARCH: Forecasting Gains and the Limits of Structural Translation", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-08-15", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7289361", "access": "Source repository access terms apply", "summary": "Machine-learning models often improve volatility forecasts, but it is unclear whether their predictive information can be translated into stable economic mechanisms. This paper treats that distinction as an empirical identification problem. A nonlinear AI teacher forecasts daily range-based variance for four equity indices, while constrained state-dependent GJR-GARCH students attempt to absorb the teacher's information through interpretable ARCH, persistence, and leverage channels. The design separates predictive qualification, information transfer, and structural identification. Boosting teachers deliver double-digit out-of-sample QLIKE gains in all eight market-proxy comparisons, all favored by HAC-robust Diebold–Mariano tests. The ranking survives temporal partitions, alternative losses, moving-block inference, a feed-forward neural teacher, and a fully trained GRU teacher. Yet teacher guidance adds at most 0.11% over an otherwise identical linear-gate student. A medium-complexity nonlinear gate produces essentially no retention under the primary Parkinson target and only small, market-specific retention under Garman–Klass. Separately, ordered Markov-switching GJR anchors fail the predeclared identification gate in every market. The evidence therefore establishes a student-class-specific structural translation gap: AI predicts volatility more accurately, but its gain is only weakly and inconsistently retained by the tested two-anchor GJR gates and is not uniquely represented by identified regime anchors. GARCH is not obsolete; it remains the transparent benchmark that reveals the boundary between prediction and explanation.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-17", "ssrn": "7289361" }, { "id": "0092", "title": "When the Scenario Outlives the Bond: Measuring Horizon Bias in Climate Credit Stress Tests", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-08-15", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7289938", "access": "Source repository access terms apply", "summary": "Climate stress tests translate scenario paths into portfolio losses, but the scenario horizon can outlive the instruments being stressed. This paper isolates one implementation choice: whether sector-level shocks are censored at each security's contractual exit. The frozen universe contains 822 Chinese interbank medium-term notes issued by 116 listed legal entities and outstanding at year-end 2025. Holding the NGFS CLIMACRED path, legalissuer mapping, and portfolio weights fixed, a horizon-blind maximum-loss rule exceeds its maturity-aware counterpart by 0.081–0.253 percentage points across three short-term scenarios. The gap is concentrated in bonds exiting within two years, remains positive in all 5,000 issuer resamples and leave-one-issuer-out exercises, survives a conservative sectorremapping envelope, and can reorder sector rankings. Mean and cumulative variants preserve the positive maximum-loss direction. A terminal-year statistic, which compares different calendar years, is positive in two scenarios and negative in DIRE; the set-inclusion property guaranteeing a non-negative maximum-loss gap does not extend to that statistic. Legalmaturity and first-call variants form a separate modeling margin. For regulators, the result shows why runoff and reinvestment conventions should be explicit in climate stress WHEN THE SCENARIO OUTLIVES THE BOND templates. This is a measurement contrast, not a forecast, causal estimate, realized loss, or claim about market mispricing.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-19", "ssrn": "7289938" }, { "id": "0095", "title": "Many Voices, One Model: Synthetic Consensus and the Illusion of Independent Board Judgment", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7278279", "access": "Source repository access terms apply", "summary": "Boards derive value from plural judgment, yet conventional independence measures observe directors rather than the information infrastructures that shape their decisions. This paper develops a theory of synthetic consensus: agreement generated partly by shared large-language-model advice or common blind spots rather than independent evidence production. A common-signal model distinguishes procedural, model, and cognitive independence and shows how shared AI advice can correlate director errors while leaving formal board structure unchanged. A structured computational mechanism audit of five platform observations, six synthetic governance scenarios, and two prompt conditions yields 60 assessments. All actions were within researcher-frozen acceptable sets and all 12 tasks produced unanimous directional recommendations. Conservative coding identified one universal shared omission and one red-team threshold shared omission. Red-team prompting recovered four of 13 neutral omissions, introduced two new omissions, changed no action, and raised mean confidence from 88.07 to 89.93, or 1.87 points. The audit provides computational evidence for the mechanism at the advice layer; human uptake is a distinct external-validity question. The paper derives an independence-first governance architecture based on private pre-AI judgments, source-level provenance, evidence partitioning, adversarial verification, minority-signal preservation, and dependence-aware corroboration.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-19", "ssrn": "7278279" }, { "id": "0094", "title": "From AI Spending Guidance to Grid Commitment: Disclosure Attrition and Tariff-Duration Mismatch", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "energy-compute-infrastructure", "doi": "", "date": "2026-08-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7278420", "access": "Source repository access terms apply", "summary": "An 80% or 90% minimum bill is not necessarily 80% or 90% protection over the life of a grid asset. This paper documents two linked losses in the conversion of AI investment narratives into infrastructure commitment: disclosure attrition upstream and protection-duration attrition downstream. We combine two independently assembled, hand-audited U.S. datasets covering 40 hyperscaler earnings events and 27 large-load tariff chains. Cash capital expenditure is observable in all 40 events, but numeric forward guidance appears in 14, comparable same-target-year revisions in nine, and a blind model supports positive AI attribution in only nine of the 14 author-coded guidance observations; no event reports a standardized AI-specific payoff. Among five evidence-complete tariff architectures, stated minimumpayment floors are 80% or 90%, yet their narrow discounted 30-year coverage ranges from 0.388 to 0.900 at a 7% discount rate. Excluding the single 30-year term option, coverage ranges from 0.388 to 0.773 because support expires before the assumed recovery horizon. The datasets are not transactionmatched and support no causal claim. Their joint contribution is a duration-aware measurement result: better disclosure can improve planning without creating security, while strong headline protection can leave substantial tail exposure.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-17", "ssrn": "7278420" }, { "id": "0048", "title": "Regime Splits are not Interaction Tests: Spurious Amplification in Mixed-frequency Volatility Models", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-08-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7254039", "access": "Source repository access terms apply", "summary": "Applied volatility research routinely infers state dependence by estimating a model separately in subsamples defined by a conditioning variable and comparing slopes. This paper shows that in mixedfrequency volatility models such a comparison is not an interaction test, for two independent reasons. First, the conditioning variable typically enters the long-run variance component directly; omitting it makes each group-specific slope a pseudo-true quasi-maximum-likelihood parameter rather than a structural marginal effect. The resulting distortion does not diminish with sample size. Second, selecting non-contiguous daily observations by a monthly partition breaks the GJR-GARCH recursion at every omitted block, so the split-sample objective is not the likelihood of a valid GARCH process. We apply the argument to daily commodity futures returns, the New York Fed Global Supply Chain Pressure Index, and the Caldara-Iacoviello geopolitical-risk index. The consequences are not secondorder. For copper, the conventional split reports a supply-chain slope 0.069 higher in high-geopoliticalrisk months, but that positive difference exists only under the concatenation implementation used in applied work: re-initialising the recursion at each block gives − 0.165 and filtering on the full daily sample gives − 0.134 . The reported amplification is a sign flip away from what coherent implementations deliver. In a unified specification with both direct effects and their monthly product under a shared Beta weight, the interaction is significantly negative for copper (θ^ 3 = − 0.1002, p = 0.005) and cocoa ( − 0.4635, p < 0.001). A first-order decomposition attributes the split difference to the omitted direct effect: across commodities the observed difference and its predicted omitted-variable component correlate at 0.997, and the largest, most threshold-stable apparent amplification occurs in soybeans, the one market where supply-chain pressure has no detectable effect on long-run volatility at all. In 1,000-replication placebo simulations with the true interaction fixed at zero throughout, the naive split-based Wald test rejects at 14 to 25 percent against a nominal 5 percent, and its rejection rate is nearly flat in the true interaction, whereas the unified likelihood-ratio test reaches 86 percent power at the estimated parameters. The same argument applies to splits on uncertainty, financial stress, liquidity, and policy states.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "GARCH-MIDAS", "Sample Splitting", "Interaction Effects", "Pseudo-True Parameters", "Commodity Volatility", "Geopolitical Risk" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7254039" }, { "id": "0047", "title": "Green ETF Returns across the Conditional Distribution: Tail Asymmetry, Low-Frequency Market Stress, and Dynamic Dependence", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-08-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7253218", "access": "Source repository access terms apply", "summary": "We revisit return-distribution and volatility differences between a broad U.S. ESG ETF and the market, and between clean- and fossil-energy ETFs, using a fully reproducible daily sample from January 2017 through July 2026. The design separates two economically different spreads: ESGU minus SPY and ICLN minus XLE. Conditional quantile regressions show near-zero median spreads but substantial dispersion in both tails. Importantly, negative lower-tail estimates denote relative underperformance, not downside protection. A custom GARCH-MIDAS model with twelve strictly lagged monthly VIX observations yields positive long-run market-stress loadings for all four ETFs; all solutions are internal, although ICLN is comparatively ill-conditioned. Formal dynamic-correlation tests around 24 February 2022 show opposite shifts: ESGU-SPY dependence rises, whereas ICLN-XLE dependence falls. These findings reject a single, universal “green premium” narrative. Broad ESG tilting behaves differently from thematic clean-energy exposure, and conditional-distribution evidence cannot by itself identify an ESGpricing mechanism. The paper also documents how sign errors, unavailable ESG-score histories, and boundary solutions can generate misleading interpretations in small-sample sustainable-finance applications.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "ESG ETFs", "Clean Energy", "Quantile Regression", "GARCH-MIDAS", "Dynamic Conditional Correlation", "Market Stress" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7253218" }, { "id": "0046", "title": "AI Investment Is not an Accounting Variable: Evidence from Hyperscaler Disclosures", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7251624", "access": "Source repository access terms apply", "summary": "U.S. financial statements contain no standardized “AI investment” line item, although market commentary routinely compares firms’ AI spending. This paper studies the resulting measurement problem using 40 earnings events for Amazon, Alphabet, Meta, Microsoft, and Oracle from September 2024 through July 2026. Cash capital expenditure is observable in all 40 events; numeric forward capitalexpenditure guidance appears in 14, comparable same-target-year revisions in nine, and lease-liability balances in 32. Attribution is substantially less stable than the underlying numbers. The author positively attributes all 14 guidance events to AI-related investment, whereas a label-blind independent languagemodel recoding identifies positive support in nine. Exact grade agreement is 6 of 14 (42.9 percent; Cohen’s kappa = 0.026), and high-versus-not-high agreement is 7 of 14 (50.0 percent; kappa = 0.109). The disagreement is reported as a human–model classification-stability diagnostic, not conventional human inter-coder reliability. The paper introduces an AI Investment Observability Lattice that separates recognition, attribution, horizon, and consolidation boundaries. A formal non-additivity result shows that cash flows, liability stocks, forecasts, and contractual schedules cannot be combined into a presentationinvariant scalar without explicit transformations to a common valuation date, allocation rule, and perimeter. No standardized event-level AI-specific payoff is observed. The evidence is descriptive and intensive rather than representative or causal.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Artificial Intelligence", "Capital Expenditure", "Financial Reporting", "Disclosure", "Leases", "Hyperscalers" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7251624" }, { "id": "0045", "title": "Who Pays When the Data Center Never Comes? Contract Design and Stranded-Cost Risk in U.S. Large-Load Electricity Tariffs", "authors": [ "fuli-yang", "tingting-qiu", "jing-huang" ], "type": "Working paper", "series": "energy-compute-infrastructure", "doi": "", "date": "2026-08-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7250402", "access": "Source repository access terms apply", "summary": "Serving a hyperscale data center can require irreversible utility investment before the customer operates. If the project is delayed, downsized, or cancelled, who pays for unused capacity? This paper studies largeload electricity tariffs as public contracts against customer non-arrival and underutilization. The evidence comprises 27 independent U.S. utility–state–docket policy chains, 167 source-level records, and 93 firstcoder-confirmed clause observations from operative documents. Clauses are coded as present, absent only after full-document audit, or not publicly verifiable; missing evidence is not converted into weak protection. A locked model-assisted blind recoding of 40 core-case cells yields 92.5% state agreement, 85.0% exact ordinal agreement, and quadratic-weighted Cohen’s κ of 0.508; it is a reproducibility audit, not human inter-coder reliability. The paper develops a five-architecture typology and a duration-sensitive backing measure comparing enforceable customer support with the remaining life of dedicated assets. A two-period model separates pre-investment deterrence from post-signing abandonment, and a two-state benchmark shows how competition for mobile projects can weaken backing below the national zeroexposure boundary. The cases reveal strong but heterogeneous safeguards. An adversarial missing-as-zero exercise is underpowered and cannot establish rank independence, but shows why the current record cannot support a stable national cardinal ranking. Large-load tariffs determine whether speculativeinfrastructure downside remains with the large customer, moves to utility investors, or is socialized across ordinary ratepayers.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Data Centers", "Large-Load Tariffs", "Incomplete Contracts", "Stranded Costs", "Ratepayer Protection", "Irreversible Investment" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7250402" }, { "id": "0007", "title": "Real-Time Distributional Regime Monitoring: Online Wasserstein Detection of Market Stress", "authors": [ "fuli-yang", "tingting-qiu" ], "type": "Working paper", "series": "distributional-geometry", "doi": "10.2139/ssrn.7245378", "date": "2026-08-13", "manuscript_date": "2026-10-18", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7245378", "access": "Source repository access terms apply", "summary": "We develop a real-time framework for monitoring market stress through the geometry of rolling return distributions. A Wasserstein score locates each current distribution relative to low- and high-stress prototypes learned from prior observations; thresholds are calibrated before 2007 and frozen. Across four equity indices from 2000 through June 2026, raw DRP detects 21 of 28 market–event cases, compared with 15 for realized volatility, 14 for EWMA, and 13 for fixed-parameter GARCH, at similar mean offwindow alarm-day shares of 3.1%–3.5%. These coverage differences remain positive after each of the seven episodes is removed in turn. A direct distribution-sensitive comparison uses rolling Kolmogorov– Smirnov monitors with past-only reference samples. With training alarm budgets aligned, DRP and the expanding-reference KS monitor each detect 21 cases. Among their 17 jointly detected cases, DRP has a median paired lead-time advantage of seven trading days; off-window alarm-day shares are 3.37% and 2.59%, respectively. The comparison identifies the response-time and alarm-duration dimensions of prototype-based stress alignment. Controlled experiments establish the underlying information channel: under a variance-matched shape change, raw and shape-only DRP achieve AUCs of 0.842 and 0.854, versus 0.524 for realized volatility. The framework extends financial-stress surveillance from the magnitude of fluctuations to the evolving configuration of the return distribution, linking geometric state measurement to event coverage, response time, and risk review.", "keywords": [ "Wasserstein distance", "distributional monitoring", "market stress", "volatility regimes", "sequential detection" ], "jel": [ "C22", "C58", "G01", "G17" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "related_sources": [], "author_abstract": true, "abstract_source": "Author manuscript", "revision_date": "2026-08-30", "ssrn": "7245378" }, { "id": "0039", "title": "When Local Markets Close but the World Keeps Trading: Global Information Absorption at Market Reopening", "authors": [ "fuli-yang", "tingting-qiu" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-12", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7248259", "access": "Source repository access terms apply", "summary": "What happens to global information when a local equity market is closed but foreign markets continue to trade? We use exchange-calendar differences to construct UTC-stamped information windows from the last local close before a closure to the local reopening. Verified closure episodes are compared with continuously traded pseudo-closure windows of identical calendar length, matched on the magnitude and sign of the foreign shock, global risk, seasonality, and pre-event domestic volatility. In 809 matched episodes for the Nikkei 225, FTSE 100, DAX, and S&P 500, the closure-by-foreign-return interaction is 0.631 (standard error 0.064) for the reopening-day return but -0.018 (0.062) for the cumulative calendarequivalent return. Closure therefore changes when information enters local prices in every audited market, although the magnitude differs. The pooled cumulative effect is near zero, but formal heterogeneity shows attenuation outside Japan (-0.184, p=0.005) and a positive Nikkei differential (0.293, p=0.005). In the audited DAX–Nikkei opening subsample, the differential reopening loading splits between the opening gap (0.367) and the subsequent session (0.390). Absolute foreign variation also raises the absolute reopening response; the interaction is 0.260 (0.080) and survives multiplicity adjustment and conditional permutation diagnostics. The evidence supports deferred price discovery and uncertainty concentration, but not general temporal insulation. The timing result survives synchronizedclosure, weekend-structure, and within-year calendar falsifications. Because a prespecified delayedforeign-return diagnostic rejects, however, the estimates are interpreted as timing and predictive evidence rather than a causal welfare effect of closure.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "International Price Discovery", "Market Closures", "Trading Calendars", "Information Transmission", "Market Reopening" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7248259" }, { "id": "0096", "title": "The 5:30 Filing Cliff: Administrative Deadlines, Disclosure Timing, and Event-Time Measurement", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-11", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7264259", "access": "Source repository access terms apply", "summary": "Public disclosure is not governed by a single timestamp. In the U.S. Securities and Exchange Commission's EDGAR system, the time a submission is accepted, the date legally assigned to it, and the time it becomes publicly available can diverge. For most forms, a direct transmission begun after 5:30 p.m. Eastern Time is assigned the next business day's filing date and is not disseminated until that day. I document this administrative discontinuity and study a 5 February 2024 reform that extended same-day treatment for Schedules 13D and 13G to 10:00 p.m. In a pilot covering 500 large current U.S.-listed issuers, Schedule 13G filings increased their post-5:30 share from 2.62% to 11.07%, while 8-K, 10-Q, and 10-K filings were nearly unchanged. A differencein-differences design with issuer, calendar month-day, and form fixed effects estimates a 12.71 percentage-point increase (CR1 two-way clustered standard error 6.88 points; p=0.066). The estimate is driven substantially by the annual February filing concentration: omitting February reduces it to 3.73 points (p=0.252), and none of six postaudit exploratory alternatives survives Holm adjustment. A daily-price illustration further shows that assigning shifted earnings filings to the acceptance date moves returns and volume away from the economically relevant availability date. The institutional evidence establishes a form-specific source of event-time measurement error; the behavioral estimate is economically large but statistically suggestive. The evidence does not support a maintained claim that firms use the cutoff to conceal bad news.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-12", "ssrn": "7264259" }, { "id": "0040", "title": "Artificial Intelligence Regulatory Uncertainty (AIRU): A Standard for Measuring Institutional Regulatory Uncertainty in AI Governance", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-11", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7238658", "access": "Source repository access terms apply", "summary": "This document establishes the Artificial Intelligence Regulatory Uncertainty (AIRU) Standard, an open and version-controlled measurement standard for institutional uncertainty in AI governance. AIRU treats regulatory transitions—not document counts, keyword frequency, or regulatory strictness—as the basic observation unit. Each transition is anchored to an official source, linked to a regulatory chain, coded across uncertainty dimensions, and preserved in an auditable database. The first financial validation uses a preregistered event-study design on 20 US regulatory transitions grouped into 15 calendar clusters. A high-exposure basket (SOXX and AIQ) is compared with a low-exposure basket (SPY and QQQ), with XLP as a placebo. The confirmatory endpoint is the cluster-level high-minus-low standardized abnormal absolute residual volatility (SAARV) contrast. Its mean is 0.0036, with a one-sided cluster sign-flip pvalue of 0.491 and a 95% bootstrap interval of [-0.190, 0.205]. Pre-specified robustness checks do not support a positive directional claim. The evidence establishes a transparent first validation, not a confirmed financial effect. Long-horizon GARCH-MIDAS validation is deferred under the G6 Tier C power gate.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Artificial Intelligence", "Regulatory Uncertainty", "Institutional Events", "Measurement Standards", "Financial Volatility", "Event Study", "GARCH-MIDAS" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7238658" }, { "id": "0012", "title": "Global Stress without Global Shape: Volatility Synchronization and Distributional Fragmentation across Financial Markets", "authors": [ "fuli-yang", "tingting-qiu", "shiyu-lin" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-08-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7259360", "access": "Source repository access terms apply", "summary": "Global financial stress can synchronize market movements while generating distinct loss distributions. We establish this separation using rolling empirical distributions for the S&P 500, FTSE 100, Nikkei 225 and Hang Seng from 2000 through June 2026. Global scale measures stress intensity, cross-market volatility gaps measure scale convergence, and standardized Wasserstein distance measures differences in loss geometry. At 252 observations, the GFC displays shape convergence while COVID combines narrower volatility gaps and lower raw distance with substantially higher standardized shape divergence. The common-calendar COVID distance is 0.1469 against 0.1001 outside the named episodes, even as average return correlation rises from 0.321 to 0.464. A common innovation distribution combined with heterogeneous observed volatility paths reproduces the long-window COVID coefficient, connecting the configuration to variance mixing under shared stress. Its risk consequence is measurable: target-excluded empirical stress templates retain approximately 7.5% average relative error in standardized 5% expected-shortfall magnitude during COVID, versus 19.53% for a matched common Student-t and 22.35% for a Gaussian. The findings establish distributional shape as a separate coordinate of cross-market stress and identify market-specific tail calibration as a substantive part of transferring common scenarios.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Financial contagion", "Distributional risk", "Wasserstein distance", "Volatility synchronization", "Systemic risk", "Market regimes" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7259360" }, { "id": "0006", "title": "Distributional Regime Persistence: Geometric Market Memory and Its Forecasting Limits", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "distributional-geometry", "doi": "10.2139/ssrn.7230659", "date": "2026-08-06", "manuscript_date": "2026-09-01", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7230659", "access": "Source repository access terms apply", "summary": "We define distributional regime persistence as the duration for which a rolling empirical return distribution remains geometrically aligned with a high-volatility prototype. Building on Wasserstein market-regime clustering, we estimate low- and high-volatility barycenters from rolling return distributions and locate each window between them using a normalized relative-distance score. This score yields distributional recovery times and cumulative excess memory that can be compared with filtered high-state probabilities from an ordered, variance-targeted Markov-switching GJR-GARCH model. Daily returns for the S&P 500, FTSE 100, Nikkei 225, and Hang Seng from January 2000 through June 2026 produce three main findings. First, raw distributional persistence is strongly scale-related: log realized volatility explains 59.5%–74.3% of its in-sample variation, while a conditional shape-only score retains modest residual information. Second, FTSE and Nikkei produce negative recovery gaps across alternative thresholds and geometries, but their bootstrap intervals include zero; the negative direction appears in 82%–85% of replications. S&P 500 COVID provides the strongest inferential result: its distribution-lagging gap has a 95% bootstrap interval of 4–94 trading days and is positive in 98.5% of replications. Hang Seng Markov-switching estimates are weakly identified and excluded from parametric inference. Third, observed DRP persistence exceeds an overlapping-window IID Student- t null. In a realtime forecast design with pre-2007 fixed prototypes, a multi-horizon volatility benchmark, expanding and rolling estimation, and MSE, MAE, and QLIKE losses, however, DRP produces no stable incremental forecast gain. A separate frozen-cutoff replacement race also rejects a stronger claim that DRP can substitute for the Markov-switching state: DRP loses all three predefined axes—crisis timing, controlled misspecification, and source-market transfer—and therefore does not qualify as an independent portability result. The evidence supports DRP as a measurement and model-diagnostic framework—not as a new clustering claim or a universal forecasting improvement. The negative forecast result is an informational boundary: geometric alignment describes the realized state and model disagreement, but it is not an omitted volatility factor once multi-horizon volatility history is observed.", "keywords": [ "Wasserstein distance", "regime persistence", "volatility forecasting", "Markov-switching GARCH", "distributional geometry" ], "jel": [ "C14", "C22", "C58", "G01", "G17" ], "year": 2026, "site_date": "2026-10-27", "summary_kind": "Author abstract", "rights": "All rights reserved for manuscript files; see source terms.", "related_sources": [], "author_abstract": true, "abstract_source": "Author manuscript", "revision_date": "2026-09-01", "ssrn": "7230659" }, { "id": "0041", "title": "Artificial Intelligence and Financial Volatility: From Model Augmentation to Market Disruption — A Systematic Critical Review", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-08-05", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7196800", "access": "Source repository access terms apply", "summary": "The rapid integration of artificial intelligence across financial markets has generated a heterogeneous literature whose findings are frequently compressed into conflicting partial verdicts about whether AI stabilizes or destabilizes markets. This review organizes that literature around a Technology–Shock–Agent (TSA) framework, distinguishing the epistemic channel through which AI improves forecasting from the channels through which AI events and AI trading agents affect realized market outcomes. Following a PRISMA-guided systematic review, we identify 72 Core studies (2015–2026) drawn from 140 unique sources. The Technology arm documents forecasting-accuracy improvements but offers only epistemic, not market-outcome, evidence. The Shock arm supplies direct market-outcome evidence, including seven causal or quasi-experimental entries. Twenty-one direct market-outcome sources are directionally consistent with volatility compression in normal conditions and amplification under stress — consistent with, but not establishing, a causal regularity. Critically, zero qualifying studies isolate adaptive AI trading agents during real stress events using causal identification. We conclude that precautionary regulatory action is warranted while full empirical confirmation awaits future research.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Artificial Intelligence", "Financial Volatility", "Systemic Risk", "Systematic review", "GARCH models", "Algorithmic trading", "Financial stability" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-06", "ssrn": "7196800" }, { "id": "0042", "title": "Leveraged ETF Reform and Intra-Sector Correlation Dynamics: Evidence from Korea’s KOSPI", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-04", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7158999", "access": "Source repository access terms apply", "summary": "This paper documents a sharp escalation in intra-sector conditional correlation among large-cap KOSPI equities coinciding with South Korea's introduction of single-stock 2× leveraged exchange-traded funds (ETFs) in 2026. Using a DCC-GJR-GARCH(1,1) model estimated on daily returns for a 24-stock panel of KOSPI constituents from July 1, 2024 to July 21, 2026 (N=467), we find that average pairwise dynamic conditional correlation rose from 0.2912 in the pre-reform baseline to 0.3675 during the ETF-live period (May 27–July 15, 2026), and further to 0.4509 in the three trading days following the Financial Services Commission's (FSC) emergency regulatory tightening of July 16, 2026. All three regime-to-regime differences are statistically significant under Welch two-sample t-tests (pre-reform to ETF-live: Δ=+0.0763, t=3.286, p=0.0011; ETF-live to post-FSC: Δ=+0.0834, t=2.719, p=0.0102; pre-reform to post-FSC: Δ=+0.1597, t=2.087, p=0.0375). An Andrews (1993) SupF structural break test identifies an overwhelming break (SupF=278.597, far exceeding the 1% critical value of 12.16) on March 2, 2026 — 86 days before the ETFs began trading but only 32 days after the FSC's January 29, 2026 policy announcement — suggesting that anticipatory repositioning may be central to the correlation shift. A complementary event-window analysis around the July 16 tightening shows a further increase (baseline ρ=0.399 to post-event ρ=0.506, t=2.752, p=0.009). Pair-level analysis shows that the largest correlation increases concentrate not in the ETF underlyings but in non-underlying pairs spanning batteries, healthcare, financials, and technology — for example Samsung SDI × Rino Industrial (Δ=+0.507) and Samsung Biologics × Woori Financial (Δ=+0.481). Causal attribution to any single mechanism is not possible given the short post-FSC window and absence of ETF flow data; results are offered as a quasi-natural-experiment description of an ongoing episode.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Leveraged ETFs", "Dynamic Conditional Correlation", "KOSPI", "Structural Break", "Financial Contagion", "Market Microstructure", "Korean Capital Markets", "Regulatory Event Study" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7158999" }, { "id": "0114", "title": "When Markets Move as One: Institutional Synchronization after Trading Halts -Evidence from Korea's 2026 Halt Cluster", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-03", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7148038", "access": "Source repository access terms apply", "summary": "Do market-wide trading halts stabilize financial markets, or do they amplify market-wide stress once trading resumes? This paper addresses this question using an unprecedented sequence of nine institutional trading halts implemented by the Korea Exchange (KRX) between March and July 2026, including seven market-wide circuit breakers and two index-futures sidecars. Unlike prior studies that examine isolated halt events, the clustered nature of the 2026 Korean episode provides a rare opportunity to study repeated institutional interventions within a common regulatory regime while market conditions evolve rapidly. Using high-frequency intraday data covering more than 2,400 listed firms, together with index, ETF, options-implied volatility, and cross-market information, we examine how trading halts affect realized volatility, cross-market transmission, dynamic dependence, and cross-sectional synchronization. Our identification strategy combines event-time comparisons, high-frequency volatility measures, dynamic correlation models, placebo tests, structural-break analysis, and international falsification exercises. We document four main findings. First, trading halts are followed by significantly higher realized volatility for Korea’s dominant semiconductor firms rather than lower post-halt volatility, indicating that reopening does not immediately stabilize trading conditions. Second, directional volatility transmission shifts sharply during halt episodes, with shocks originating from the broader KOSDAQ market propagating toward large-cap semiconductor firms, while the reverse channel remains weak. Third, cross-sectional dispersion across KOSPI 200 constituents contracts substantially despite dramatically higher aggregate volatility, suggesting that institutional trading concentrates market-wide risk into a common systematic component. Finally, every halt occurs when options markets have already priced extreme uncertainty, with event-day VKOSPI reaching the upper tail of its historical distribution. Taken together, the evidence is consistent with the Institutional Synchronization Framework (ISF) developed in this paper. We formalize the ISF as a two-period rational-expectations model in which institutional investors are subject to value-at-risk and margin constraints while retail investors trade on heterogeneous private information. Building on the funding-liquidity framework of Brunnermeier and Pedersen (2009), the model embeds an explicit halt-and-reopen mechanism and yields closed-form comparative statics for each empirical proposition. Under the ISF, trading halts are implemented after market stress has already become systemic and reopening mechanisms fail to dissipate aggregate uncertainty. We interpret the joint behavior of realized volatility, correlations, spillovers, dispersion, and options-implied volatility as complementary manifestations of institutional synchronization: common risk-management constraints translate heterogeneous security-specific information into homogeneous institutional trading behavior, temporarily reorganizing the architecture of market risk. Rather than viewing circuit breakers solely as volatility-control devices, our findings suggest that they may function as institutional coordination events during episodes of severe market stress. These findings contribute to the literature on market microstructure, volatility transmission, and regulatory intervention by proposing the ISF and providing new evidence from the first clustered sequence of market-wide trading halts observed in a major developed Asian equity market.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "trading halts", "circuit breakers", "institutional synchronization", "funding liquidity", "margin spiral", "common-factor risk", "cross-sectional dispersion", "volatility spillovers", "dynamic conditional correlation", "options-implied volatility", "KOSPI", "KOSDAQ" ], "jel": [ "G12", "G14", "G15", "G18", "C58" ], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7148038" }, { "id": "0043", "title": "Institutional Synchronization Theory: A Framework for Market Dynamics under Common Institutional Constraints", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-03", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7148838", "access": "Source repository access terms apply", "summary": "We develop Institutional Synchronization Theory (IST), an equilibrium framework in which market-wide synchronization arises endogenously as the consequence of institutional optimization under a common constraint. Heterogeneous institutions solve their own mean-variance problems but face the same regulatory environment. A change in the common constraint enters every institution’s optimization problem through a shared gradient term, aligning individually optimal portfolio adjustments in direction and concentrating aggregate institutional demand. Market clearing translates this demand concentration into amplified price responses, elevated return comovement, and stronger cross-asset volatility transmission. We term the resulting equilibrium a synchronization equilibrium. The theory yields four chained predictions — institutional demand concentration, dynamic return correlation, volatility spillovers, and temporary liquidity deterioration — that we evaluate using the 2026 Korea Exchange trading halt cluster as a natural experiment. IST is the first component of a broader research program on Institutional Market Dynamics (IMD) that reinterprets comovement, spillovers, liquidity fragility, and price impact amplification as manifestations of a single underlying mechanism.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Institutional Synchronization", "Common Constraints", "Market Equilibrium", "Regulatory Intervention", "Return Comovement", "Volatility Spillovers", "Trading Halts", "Korea Exchange", "DCC-GARCH" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7148838" }, { "id": "0044", "title": "Institutional Shocks and GARCH Mis-specification: Evidence from China’s 2016 Circuit Breaker Episode and AH Dual-Listed Pairs", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-08-01", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7137098", "access": "Source repository access terms apply", "summary": "This paper examines whether standard GJR-GARCH volatility specifications remain structurally stable when subjected to abrupt institutional shocks, using China’s January 2016 market-wide circuit breaker episode as a natural experiment. The circuit breaker mechanism was triggered twice within four trading days of its introduction and was abolished on January 8, 2016, producing a sharply defined exogenous policy shock within a market already characterized by ±10% daily price limits and T+1 settlement. Using a three-layer empirical design combining Bai-Perron structural break tests, a difference-in-differences comparison against Hong Kong equity benchmarks, and a novel GARCH Portability Index applied to 18 AH dual-listed pairs, the paper documents systematic mis-specification in baseline GJR-GARCH(1,1)-t models. Structural breaks cluster around the circuit breaker dates, leverage asymmetries diverge significantly (mean Portability Index 0.553, below the 0.80 threshold), and baseline models fail Kupiec coverage tests in all 16 sectors. Augmenting with institutional friction terms improves VaR pass rates from 0% to 44%.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Circuit Breaker", "GJR-GARCH", "Structural Break", "Price Limits", "T+1 Settlement", "AH Dual-Listed Pairs", "GARCH Portability", "VaR" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7137098" }, { "id": "0097", "title": "Central Bank Language and Long-Run Volatility: A Harmonized Four-Bank Test of Measurement and Regime Dependence", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "central-bank-communication", "doi": "", "date": "2026-07-30", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7205359", "access": "Source repository access terms apply", "summary": "Does the volatility relevance of central-bank language travel across institutions, measures, and policy regimes? Using a harmonized text-scoring and GARCH-MIDAS pipeline, we study 362 scheduled statements from the Federal Reserve, ECB, Bank of England, and Bank of Japan during 2015–2026. Only ECB uncertainty remains significant after Holm correction, and that loading disappears after the ECB’s 2021 strategy redesign. Formal interactions also identify loading shifts around the 2024 Bernanke Review and the Bank of Japan’s exit from negative rates and yield-curve control. A Federal Reserve reconciliation shows strong sensitivity to lexicon choice. Textual volatility channels are measurementand regime-contingent.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-07", "ssrn": "7205359" }, { "id": "0098", "title": "A Constraint Geometry State-Space Model of Dynamic Correlation With an Empirical Demonstration on ENSO-Driven Commodity Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-07-27", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7188139", "access": "Source repository access terms apply", "summary": "We propose a Constraint Geometry State-Space Model (CGSSM) for dynamic correlation. The CGSSM replaces the correlation matrix Rt with an interpretable two-part latent state: a scalar market-wide synchronization level λt and a low-dimensional geometry factor ft that encodes the direction of dependence relative to a fixed reference correlation structure. Within the tangent space at that reference point, the decomposition is exact and orthogonal, and correlation change decomposes additively into level movement, geometry movement, and their interaction. We give a constructive identification result for the CGSSM under a fixed blockindicator loading design and a companion reconstruction result for the identifiable complement when the block design is rank-deficient. Inference proceeds by iterated-Laplace state-space filtering whose approximation quality we audit in a controlled simulation environment. As an empirical demonstration, we apply the model to a panel of eleven commodity futures with the Oceanic Niño Index as the exogenous climate channel. Under a deliberately parsimonious restricted specification, the CGSSM estimates cleanly, reconstructs symmetric positivedefinite correlation paths, and delivers stable smoothed states for the synchronization and geometry components. The demonstration establishes that the CGSSM is operationally estimable on real-world panels; formal inference on climate transmission and richer empirical applications are deferred to companion work.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-06", "ssrn": "7188139" }, { "id": "0099", "title": "Constraint Geometry of Institutional Synchronization", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-07-25", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7178198", "access": "Source repository access terms apply", "summary": "Market synchronization can emerge independent of contemporaneous information arrival, driven instead by the common component of institutional optimization under shared constraints. This paper develops the constraint geometry of that mechanism, introducing three theoretical objects. First, the ISTadmissible constraint class — institutional constraint functions that admit a Common-Idiosyncratic Decomposition, delineating the scope of Institutional Synchronization Theory (IST) as a definitional rather than assumptional restriction. Second, Synchronization Pressure — a vector-valued theoretical primitive derived from constraint shadow prices and common gradients, invariant to how the econometrician projects returns. Third, the Synchronization Law — a monotonic relationship between the equilibrium pressure-aligned covariance shift and the capacity-adjusted Synchronization Pressure normalized by market capacity, derived from a minimal Kyle-style linear price-impact assumption. A single Dominance Principle unifies the four established transmission channels of institutional finance — benchmarktracking, VaR, ETF creation/redemption, and short-sale constraints — as regime-specific corollaries. The Law's linearization recovers the empirical Scaling Law documented in Yang (2026b), building on the theoretical framework of Yang (2026a); preliminary cross-market calibration is consistent with universality.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-05", "ssrn": "7178198" }, { "id": "0100", "title": "Market Architecture and Exogenous State Transitions: Evidence from Climate Regimes in Commodity Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-07-24", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7170658", "access": "Source repository access terms apply", "summary": "Do exogenous climate regimes reorganize the dependence structure of commodity markets, or do they merely shift the general level of co-movement? We introduce the concept of Market Architecture — defined as the persistent topological organization of long-run dependence, invariant to common shifts in the level of correlation — and develop an estimable Block Factor DCC-MIDAS model whose long-run correlation target is analytic and positive definite. Using twenty commodity futures spanning five sectors from November 2, 2000 to June 29, 2026 (T = 6,469), and ENSO states as an exogenous indexing device, we find that El Nino and La Nina operate through two distinct channels. La Nina raises system-wide average long-run correlation by 9.5 percentage points, constituting a system-wide integration shock. After removing this level effect, El Nino is associated with differentiated architecture: Grains and Metals tighten relative to the system mean while Energy and Livestock remain flat, reorganizing the relative structure of dependence across sectors. The evidence supports the theoretical distinction between level and topology proposed in the paper’s foundational propositions. A three-factor specification is selected by BIC (ΔBIC = −833.3) and confirmed by nine independent starts. The results document a predictive association between climate regimes and market architecture, with implications for regime-conditional risk management.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-04", "ssrn": "7170658" }, { "id": "0050", "title": "Policy Uncertainty and Asset Volatility: A Reproducible Review and Controlled GARCH-Family Demonstration", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-07-24", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7099179", "access": "Source repository access terms apply", "summary": "Where an uncertainty variable enters a volatility model determines the estimand, yet this distinction is often blurred. We combine a reproducible review of 36 GARCH-family applications with a controlled expanding-window demonstration. All primary S&P 500 comparisons use the same 1,193 forecast dates. Rankings vary across QLIKE, MSE, one- and five-day targets, transformations, and benchmarks; realized-volatility models lead point rankings, but 10% model-confidence sets retain all ten candidates. The observed one-day QLIKE gain is far below the design’s 80%-power minimum detectable effect. Official geopolitical-risk revisions are directionally positive in 52 of 58 months, showing that first releases systematically understate the later vintage, yet final-versus-real-time losses remain statistically indistinguishable after realized volatility is included. A 1% loss-trimming audit changes the QLIKE leader, revealing stress-day dependence. Applying the same model ladder to Japan, EUR/USD, and BRL/USD produces different winners across markets and losses. Thus uncertainty measures can be conditional signals, but neither coefficients nor rankings are portable across architecture, information set, target, benchmark, and market; the designs do not identify causal effects.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Policy uncertainty", "Geopolitical risk", "Volatility forecasting", "GARCH-MIDAS", "Real-time data", "Model confidence set" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-26", "ssrn": "7099179" }, { "id": "0101", "title": "Central Bank Uncertainty and Long-Run Financial Volatility: Evidence from FOMC Communication and a GARCH-MIDAS Framework", "authors": [ "fuli-yang", "tingting-qiu", "jing-huang" ], "type": "Working paper", "series": "central-bank-communication", "doi": "", "date": "2026-07-22", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7161818", "access": "Source repository access terms apply", "summary": "This paper examines whether Federal Open Market Committee (FOMC) statement language is associated with the long-run component of equity market volatility. We construct a dataset of 89 scheduled post-meeting statements issued between January 2015 and June 2026 and score each statement along three textual dimensions: Loughran-McDonald (LM) net sentiment, a phrase-based hawkish-dovish (HD) index, and an uncertainty ratio based on hedging language. GARCH-MIDAS estimates link each dimension to the long-run variance component of daily S&P 500 returns. In the full sample, the Uncertainty Ratio is positively associated with long-run variance (theta = 0.197, SE = 0.061, p = 0.001), implying that a one-standard-deviation increase in linguistic uncertainty is associated with a 21.7% increase in the long-run variance component. The HD loading is insignificant, and the LM loading becomes insignificant when jointly modeled with uncertainty; the uncertainty loading remains significant in that joint specification (p = 0.013). To our knowledge, this is the first study to use textual measures extracted from FOMC post-meeting statements as meeting-frequency drivers of the long-run equity-variance component in a GARCH-MIDAS model. The estimand concerns the level of uncertainty language in the ratified statement; sequence, deletion, and market-based uncertainty measures define extensions rather than substitutes for this core design.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-05", "ssrn": "7161818" }, { "id": "0036", "title": "Artificial Intelligence, Algorithmic Trading, and Financial Market Microstructure: A Systematic Literature Review", "authors": [ "fuli-yang", "shiyu-lin", "jing-huang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-07-20", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7123419", "access": "Source repository access terms apply", "summary": "The past two decades have witnessed a structural transformation in financial market microstructure: human traders have been displaced by algorithmic agents operating at microsecond timescales. This paper provides a systematic review of the academic literature on how algorithmic trading (AT), high-frequency trading (HFT), and artificial intelligence—including large language models (LLMs) and machine learning classifiers—have reshaped three foundational dimensions of market quality: liquidity provision, price discovery, and market stability. Following a PRISMAcompliant protocol, we screened 847 initial records across Web of Science, Scopus, SSRN, and Google Scholar, ultimately synthesizing 60 studies meeting pre-specified inclusion criteria. We organize the literature around three central debates. The first concerns whether algorithmic market-making constitutes genuine liquidity supply or phantom liquidity that evaporates under stress, amplifying flash crashes across equity, Treasury, and foreign exchange markets. The second examines the tension between AIenhanced informational efficiency and systemic fragility induced by algorithmic herding, correlated machine-learning strategies, and retail-driven social media trading. The third surveys the regulatory challenge of detecting AI-facilitated market manipulation—including spoofing and quote stuffing—through natural language processing and high-frequency surveillance. Drawing on foundational contributions from Kyle (1985), Glosten and Milgrom (1985), and Budish, Cramton, and Shim (2015), we characterize the current paradigm as \"algorithmic ecology\"—a coevolutionary ecosystem in which AI agents, human traders, exchanges, and regulators continuously adapt to one another's strategies. We identify a fundamental paradigm shift: market microstructure research must transition from models of rational human AI, Algorithmic Trading & Market Microstructure agents toward frameworks accommodating the strategic behavior, correlated failures, and emergent dynamics of algorithmic agents.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Algorithmic Trading", "High-Frequency Trading", "Market Microstructure", "Liquidity", "Price Discovery", "Flash Crash", "Artificial Intelligence", "Large Language Models", "Market Manipulation", "Financial Regulation" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7123419" }, { "id": "0102", "title": "AI Computing Demand and Energy Market Co-movement: Evidence from EGARCH-X and DCC-GARCH Models", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "energy-compute-infrastructure", "doi": "", "date": "2026-07-16", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7128543", "access": "Source repository access terms apply", "summary": "This paper examines whether the expansion of AI-related computing demand has created a new financial linkage between major technology equities and energy infrastructure assets. Using 817 daily observations from March 2023 to July 2026, we estimate an EGARCH(1,1)-X model augmented with the MOVE Index as an exogenous variance regressor (Stage 1), and a bivariate DCC-GARCH model (Engle, 2002) on GARCH-filtered standardized residuals (Stage 2) for seven asset pairs spanning AI equities (NVDA, MSFT, SOXX) and energy infrastructure (URA, XLU). The MOVE Index coefficient δ is statistically significant for NVDA (δ = 0.023, p < 0.01), SOXX (δ = 0.030, p < 0.001), and XLU (δ = 0.018, p < 0.001), confirming macrofinancial stress transmission to AI-equity volatility. DCC-GARCH estimates reveal a significant increase in NVDA–URA co-movement (Δρ = +0.175, p < 0.001) and SOXX–URA co-movement (Δρ = +0.108, p < 0.01) following July 2024, while the control pair URA–XLU shows no structural break (Andrews sup-Wald p = 0.597). These findings suggest an AI-specific, rather than broad-market, origin for the documented co-movement increase. Policy implications for energy infrastructure investment and systemic risk monitoring are discussed.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-27", "ssrn": "7128543" }, { "id": "0103", "title": "Central Bank Communication as Literature: A Narrative Economics Survey on Fed Speak and Market Expectations", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "central-bank-communication", "doi": "", "date": "2026-07-14", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7119058", "access": "Source repository access terms apply", "summary": "Central bank communication has evolved from a peripheral policy instrument into its primary transmission mechanism. This survey treats \"Fed speak\" as a narrative-economic phenomenon, applying Shiller's (2017) framework to bridge qualitative rhetoric analysis and quantitative volatility modeling. We organize forty years of measurement methodology into four phases: manual coding, dictionary-based analysis, machine learning and NLP, and large language models. For each phase we assess strengths and limitations specific to central bank text's constructive ambiguity and committee-authored register. We then map four transmission channels — signaling, portfolio rebalancing, coordination, and cross-border spillover — through which textual narratives generate measurable market volatility, and connect this literature to GARCH-family models. We propose hybrid GARCH-MIDAS architectures incorporating high-frequency textual sentiment as a research frontier.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-27", "ssrn": "7119058" }, { "id": "0106", "title": "Supply Chain Stress and Industrial Metal Volatility: A GJR-GARCH-MIDAS Approach with Geopolitical Risk Conditioning", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-07-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7109258", "access": "Source repository access terms apply", "summary": "This paper investigates whether global supply chain pressure exerts a measurable influence on the long-run volatility of industrial metal futures, and whether this influence is conditional on the geopolitical environment. Using a GJR-GARCH-MIDAS framework applied to copper and nickel LME front-month futures over January 2005–June 2026, I decompose conditional variance into a short-run asymmetric GARCH component and a long-run component driven by the New York Fed’s Global Supply Chain Pressure Index (GSCPI). The MIDAS slope coefficient is positive and significant for both metals (copper θ = 0.149, p ≈ 0.06; nickel θ = 0.109, p = 0.013), indicating that supply chain stress raises the long-run volatility baseline. Splitting the sample by geopolitical risk (GPR) terciles reveals that copper’s GSCPI sensitivity is concentrated almost entirely in high-GPR regimes (Δθ = +0.249), while nickel retains a smaller but significant effect even in calm periods (Δθ = +0.116). Nickel further displays an anomalous positive leverage coefficient, consistent with the 2022 LME short-squeeze episode. This is the first study to use the GSCPI as a MIDAS long-run driver for industrial metals, and the first to condition that relationship on geopolitical risk.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-24", "ssrn": "7109258" }, { "id": "0105", "title": "ENSO-Driven Correlation Dynamics in Agricultural and Energy Markets: A DCC-MIDAS Approach", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-07-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7110198", "access": "Source repository access terms apply", "summary": "This paper embeds the Oceanic Niño Index (ONI) into a two-stage DCC-MIDAS model to examine how ENSO drives long-run correlation between soybean futures and WTI crude oil over September 2000 to June 2026 (N = 6,465 daily observations). The estimated long-run ENSO slope is negative and significant (θρ = −0.145, p<0.05), indicating that El Niño episodes reduce soybean-oil correlation while La Niña episodes raise it. An asymmetric decomposition confirms this pattern is driven entirely by La Niña (θ− = +0.284, p<0.05), with El Niño having no significant effect (θ+ = −0.031); a Wald test rejects symmetry at the 1% level (χ²(1) = 7.459, p = 0.006). The mechanism operates through opposing variance channels: ENSO lowers soybean long-run variance (θsoybean = −0.541) but raises oil long-run variance (θoil = +0.421), mechanically decoupling the two markets under El Niño. Estimated MIDAS effective lags reveal that correlation integrates climate information (2.8 months) far more rapidly than individual market volatilities (11.2 months for soybean, 15.2 months for oil). These findings have direct implications for commodity portfolio diversification and dynamic hedging strategies conditioned on ENSO forecasts.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-29", "ssrn": "7110198" }, { "id": "0104", "title": "Volatility Spillovers and Macro Fragility in the AI Super-Cycle: A Two-Stage GARCH Transmission Chain from Tech Giants to Energy Infrastructure", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-07-13", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7113358", "access": "Source repository access terms apply", "summary": "This paper documents a two-stage volatility transmission chain running from sovereign bond market stress, through the core equity assets of the artificial intelligence (AI) super-cycle, into physical energy infrastructure markets. Stage 1 estimates EGARCH(1,1)-t models augmented with the first difference of the MOVE Index (ΔMOVE) for NVIDIA (NVDA) and Microsoft (MSFT) over a sample of 597 daily observations (March 29, 2023 to July 2, 2026). Both AI core assets exhibit significant ARCH effects (NVDA α = 0.142**), significant leverage effects (NVDA γ = −0.065*; MSFT γ = −0.073**), and extremely high volatility persistence (β > 0.96), establishing macro-fragility rather than macro-immunity. Stage 2 extracts GARCHfiltered standardized residuals and computes rolling 60-day dynamic conditional correlations between AI assets and energy infrastructure proxies — the nuclear-energy ETF (URA) and the utilities ETF (XLU). The central finding is that the NVDA–URA correlation more than doubled from a pre-July-2024 mean of 0.168 to a post-July-2024 mean of 0.434 (t = −19.16, p < 0.001), coinciding with the AI-power narrative crystallized by Microsoft's September 20, 2024 Three Mile Island restart agreement. This pattern generalizes across SOXX–URA and MSFT–URA pairs, while the URA–XLU control pair remains stable. The NVDA–XLU pair undergoes a sign reversal from −0.226 to +0.042. A vol-on-vol regression shows near-zero R² (0.0001), indicating that contagion operates through the correlation structure rather than through volatility-level transmission. These findings document, for the first time, a measurable physical-financial nexus linking digital infrastructure investment to energy markets.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-29", "ssrn": "7113358" }, { "id": "0107", "title": "Pricing the Physical Gray Rhino: Extreme Climate Shocks and Long-Run Volatility Dynamics in Soft Commodities", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-07-11", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7101698", "access": "Source repository access terms apply", "summary": "Against the backdrop of intensifying global climate variability, extreme weather events are increasingly inflicting physical disruptions on agricultural supply chains; yet traditional financial asset pricing models predominantly neglect these long-term physical constraints. This paper investigates the transmission mechanism of physical climate risks into financial markets by employing a GJR-GARCH-MIDAS (Mixed Data Sampling) framework, utilising daily returns of soft commodity futures—cocoa and coffee—alongside the monthly Oceanic Niño Index (ONI) released by NOAA, over the period January 2000 to June 2026. Empirical results reveal that low-frequency ENSO climate anomalies significantly and positively elevate the long-run secular volatility of tree-based soft commodity futures (θ = 0.233, p < 0.01 for cocoa; θ = 0.105, p < 0.10 for coffee), while exerting no statistically significant effect on annual crop futures such as soybeans (θ = −0.417, p < 0.01 in the opposite direction). This striking heterogeneity is rooted in the divergent biological supply elasticities between perennial tree crops and short-cycle annual crops. Our findings offer a robust empirical foundation for correcting long-run volatility estimates in derivative pricing, and provide forward-looking risk management insights for multinational food corporations and central banks navigating agricultural inflation.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-24", "ssrn": "7101698" }, { "id": "0053", "title": "Generative AI Mania and Macro Vulnerability: Asymmetric Volatility Dynamics of Tech Giants via EGARCH-X (2023-2026)", "authors": [ "fuli-yang", "shiyu-lin" ], "type": "Working paper", "series": "ai-financial-systems", "doi": "", "date": "2026-07-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7071718", "access": "Source repository access terms apply", "summary": "The period from January 2023 to July 2026 witnessed an unprecedented surge in investor enthusiasm surrounding generative artificial intelligence (AI), propelling technology-sector equities—most notably NVIDIA Corporation (NVDA)—to extraordinary cumulative excess returns. This paper examines whether the AI innovation narrative confers immunity to macroeconomic systemic risk transmission. Using 866 daily observations, we estimate Exponential GARCH models augmented with exogenous macro-stress variables (EGARCH-X), incorporating the ICE BofA MOVE Index as a proxy for sovereign bond market dislocation. Our cross-sectional analysis covers five assets: NVDA, AMD, MSFT, GOOGL, and the S&P; 500 (SPY). Three findings emerge. First, the leverage effect is monotonically weaker as assets move closer to the AI narrative core: SPY exhibits the strongest leverage (γ = −0.198, p < 0.001), while AMD shows no significant asymmetry (γ = +0.028, n.s.). This cross-sectional gradient is consistent with a behavioral hypothesis: momentum-driven buy-the-dip dynamics suppress the institutional fear-leverage channel in high-narrative assets. Second, NVDA's ARCH coefficient (α = 0.142) substantially exceeds SPY's (0.088), confirming greater sensitivity to information shocks. Third, a structural break analysis centered on the January 27, 2025 DeepSeek announcement reveals a fundamental regime shift in NVDA's volatility dynamics, while SPY parameters remain stable—evidence of AI-specific structural fragility.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Generative AI", "Asymmetric Volatility", "Leverage Effect", "MOVE Index", "Structural Break", "DeepSeek", "EGARCH-X", "GARCH" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7071718" }, { "id": "0051", "title": "Attributing Forecast Gains in Hybrid GARCH–Deep Learning Models: A Critical Review of Architecture, Information, and Evaluation", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-07-10", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7090918", "access": "Source repository access terms apply", "summary": "Hybrid GARCH–deep learning models are often presented as a cumulative progression in which econometric structure and neural flexibility jointly improve volatility forecasts. This critical review asks a narrower question: what does the connection itself add? A structured search and citation-chaining exercise retained 30 qualifying studies published or circulated between January 2018 and 28 August 2026: 27 peer-reviewed articles and three working papers. Seventeen use state transfer, six use forecast or error combination, and seven use recursion embedding. The evidence supports conditional complementarity but not a monotone ranking of integration depth. Reported gains commonly bundle architecture with unequal information, search effort, targets, validation windows, and stress-period composition. The hybrid attribution framework therefore defines performance contrasts on a common expected-loss scale and translates them into information-matched comparisons, nested ablations, common tuning rules, and identical forecast origins. Four patterns emerge: GARCH-derived states are representations rather than new information; simple diversification is an underused benchmark; embedded recursions improve coherence but raise identification costs; and point-volatility gains need not improve tail calibration or economic value. The principal limitations are partial 2026 coverage, English-language discovery, single-author coding, heterogeneous outcomes, and reliance on reported rather than replicated results. The next advance depends less on another hybrid label than on identifying when, why, and for which endpoint the econometric–neural link adds value.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "GARCH", "Deep learning", "Volatility forecasting", "Hybrid models", "Forecast evaluation", "Model comparison" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-08-28", "ssrn": "7090918" }, { "id": "0054", "title": "Sentiment Shocks, Asymmetric Volatility, and Market Segmentation: Evidence from AH Dual-Listed Pairs", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-07-08", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7053938", "access": "Source repository access terms apply", "summary": "This paper examines whether investor sentiment shocks affect conditional volatility asymmetrically in China’s equity markets, using a sample of 32 AH dual-listed firm-market pairs over 2018–2024. We augment the standard GJR-GARCH(1,1) model with positive and negative sentiment indices derived from the Loughran-McDonald financial dictionary — Chinese-language news for A-shares (SCN) and English-language news for H-shares (SHK) — as exogenous variance regressors. Estimating the GJR-GARCH-X model via maximum likelihood with Student-t innovations for 15 A-share and 17 H-share series, we find strong evidence of sentiment-driven volatility. For A-shares, the negative sentiment coefficient (δ−) is statistically significant in 13 of 15 cases (87%), with a mean magnitude of 0.260, compared with only 7 of 17 H-share series (41%). The sentiment-augmented model yields ΔAIC < 0 (improved fit over the baseline |δ+| for the majority of pairs, confirming sentiment asymmetry — negative news amplifies volatility more than positive news dampens it. Sector analysis reveals that energy and banking A-shares exhibit the strongest sentiment sensitivity, while construction and telecom firms show negligible effects. H-shares display significantly weaker sentiment incorporation, consistent with the hypothesis that Hong Kong’s more informationally efficient market prices sentiment shocks faster. Mean GARCH persistence is high in both markets (A: 0.979, H: 0.980), indicating long-lived volatility clustering. Our findings have implications for volatility forecasting, risk management, and the ongoing debate on A-share market efficiency.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "GJR-GARCH-X", "Sentiment analysis", "Loughran-McDonald", "AH shares", "Volatility asymmetry", "Market segmentation", "China equity markets", "Investor sentiment" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7053938" }, { "id": "0052", "title": "The Physical Limits of Compute: Dynamic Risk Contagion Networks Between Semiconductor and Energy Infrastructure Assets Under the AI Super Cycle", "authors": [ "fuli-yang", "shiyu-lin" ], "type": "Working paper", "series": "energy-compute-infrastructure", "doi": "", "date": "2026-07-08", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7080180", "access": "Source repository access terms apply", "summary": "Artificial intelligence depends on a production chain connecting semiconductors, data centres, electrical equipment, generation, nuclear supply, cooling and gas infrastructure. This paper shows how that physical chain becomes a financial risk network. We combine daily adjusted equity prices through June 2026 with published electricity-capacity outcomes, power agreements and a regulatory access decision. A seven-layer system uses 28 continuously represented stocks and an unbalanced extension containing five additional infrastructure firms. Factor exposures are estimated using strictly lagged rolling regressions. Between the compute-expansion and infrastructure phases, return connectedness rises from 6.34% to 21.72%, and connectedness in transformed standardised downside innovations rises from 5.27% to 22.12%. Both changes survive the three-outcome multiplicity correction and remain significant across beta windows, control sets, block lengths, layer deletions, extreme-date deletions and firm membership. Boundary analysis locates the transition within 2024: the downside increase holds for every boundary from January 2024 through January 2025, and the return-network increase is strongest around mid-2024. The observed 15.38-point return-network increase exceeds 97.3% of 5,000 size-preserving random layer assignments, while the seven-layer diversification ratio falls by 17.8%. The network peaks during the 2025 infrastructure build-out and remains significantly above the compute-expansion benchmark in the final 252 sessions. Grid equipment has the largest outgoing variance attribution. Competitive generators outperform regulated utilities by 12.18 percentage points around the Crane agreement, a contrast that remains significant against its pre-event placebo distribution after Holm correction across ten source-dated events. Most measured connectedness is already present at the one-step horizon. The evidence establishes a stronger, uneven financial risk network spanning the physical inputs to AI, with rapid joint repricing as its dominant daily manifestation.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "Dynamic correlation", "Semiconductors", "Nuclear energy", "AI infrastructure", "Risk contagion", "Structural break", "Energy transition", "DCC-GARCH" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "7080180" }, { "id": "0108", "title": "Cointegration, Spread Dynamics, and GARCH-Type Volatility Modeling: An Empirical Econometrics Framework", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-06-30", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7025142", "access": "Source repository access terms apply", "summary": "This paper develops a unified methodological framework for studying cointegration, timevarying hedge ratios, and GARCH-type volatility dynamics in mean-reverting financial spreads. When asset prices are integrated of order one, the identification of cointegrated relationships and the estimation of spread dynamics are sensitive to the choice of testing procedure, volatility proxy, and evaluation criterion. Using Monte Carlo simulations together with an illustrative empirical design based on foreign exchange and equity markets, the paper compares EngleGranger and Johansen procedures, static and Kalman-filter-based hedge-ratio estimation, and alternative specifications for modeling conditional heteroskedasticity in the spread residual. The results indicate that model rankings are criterion-dependent: statistical loss functions, economic performance measures, and tail-risk criteria can yield materially different preferred specifications. Model confidence set evidence further suggests that, in finite samples, multiple plausible models may remain empirically indistinguishable. The paper contributes to the econometrics of cointegrated financial systems by integrating spread-volatility modeling, criterion sensitivity, and model uncertainty within a common evaluation framework.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-08", "ssrn": "7025142" }, { "id": "0109", "title": "Price Limits, T+1 Settlement, and Volatility Clustering: A Direct Portability Test of Standard GJR-GARCH Specifications to China's A-Share Market", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "institutions-market-structure", "doi": "", "date": "2026-06-28", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7012118", "access": "Source repository access terms apply", "summary": "This paper tests whether standard GJR-GARCH and DCC specifications calibrated on developed-market equity markets can be directly transplanted to China's A-share market, where two structural constraints — a ±10%/±5% daily price limit and T+1 settlement — fundamentally alter the volatility generating process. Using 2015–2025 daily data for eight CSI 300 sector indices (N = 2,695) and a panel of 18 AH dual-listed company pairs as a natural experiment, we document three systematic distortions. First, price limits truncate the return distribution and induce artificial autocorrelation in squared residuals, causing baseline GJR-GARCH to overestimate volatility persistence (mean α+γ/2+β = 0.9857 for A-shares vs. 0.9905 for matched HK counterparts, ∆ = -0.0048). Second, T+1 settlement creates an asymmetric liquidity shock — investors cannot exit positions acquired on the same day after an adverse price move — inflating the leverage coefficient γ beyond its standard 'bad news' interpretation (mean ∆γ = -0.0577 across AH pairs; 16/18 pairs reject H0: γ_A = γ_H at p<0.05). Third, VaR backtests confirm that unmodified standard GJR-GARCH specifications fail Kupiec's unconditional coverage test in 100% of sector-level cases. The aggregate GARCH Portability Index averages 0.553 (on a 0–1 scale), well below the 0.80 threshold for model transportability. We propose two minimal corrections — a limit-hit indicator in the variance equation and a T+1 asymmetry adjustment in the leverage term — that reduce persistence bias and improve VaR coverage, providing a blueprint for institutional practitioners deploying standard risk models in Chinese equity markets.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-08", "ssrn": "7012118" }, { "id": "0110", "title": "Carbon Price Uncertainty and Stock Market Volatility Transmission: A GJR-GARCH-X and DCC Approach with Evidence from China's ETS", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "climate-finance", "doi": "", "date": "2026-06-25", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6997340", "access": "Source repository access terms apply", "summary": "This paper examines whether carbon price uncertainty transmits to stock market volatility in China's high-emission sectors following the launch of the national Emissions Trading Scheme (ETS) in July 2021. Using 1,072 daily observations on China Carbon Allowance (CEA) prices and Shenwan sector indices for electricity, steel, and cement over the period July 2021–December 2025, we document three findings. First, full-sample GJR-GARCH-X estimates yield statistically insignificant delta coefficients across all three sectors (|delta| ≤ 0.0085, p > 0.30), an informative null consistent with thin ETS trading and Phase 1 restriction to the power sector only. Second, event-window analysis around annual compliance deadlines reveals time-varying transmission: carbon uncertainty significantly amplifies electricity volatility in the 60-day window preceding the 2023 deadline (delta = 0.0448, p = 0.033) and the 60-day window following the 2024 deadline (delta = 0.0981, p = 0.008). Third, Asymmetric Slope CAViaR-X models demonstrate that carbon price uncertainty significantly shifts the extreme left tail of electricity returns (delta = 0.0473, p < 0.001 at the 1% quantile) even in the full sample where conditional mean-variance effects are absent. Our results distinguish between a dormant average variance channel and an active tail-risk channel, with implications for climate stress testing and carbon-aware portfolio risk management. —1—", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-08", "ssrn": "6997340" }, { "id": "0111", "title": "Loss Function Sensitivity, Volatility Forecasting, and Dynamic Model Choice: Evidence from Foreign Exchange and Equity Markets", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-06-18", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6962001", "access": "Source repository access terms apply", "summary": "This paper examines how volatility model rankings depend on the choice of loss function when the true conditional variance is unobservable and forecast evaluation must rely on noisy volatility proxies. Using daily data for EUR/USD and the S&P 500, it compares GARCH, GJR-GARCH, and EGARCH through rolling one-step-ahead forecasts under four evaluation criteria: MSE, MAE, QLIKE, and joint VaR–ES loss. The results show that model rankings are loss-dependent: EGARCH performs best overall under MSE, QLIKE, and joint VaR–ES loss, whereas GJR-GARCH performs relatively better under MAE. At the same time, model confidence set results indicate that the three models are not sharply distinguishable in finite samples at the 10% level, implying that a superior model set is often more defensible than a unique winner. Dynamic model averaging and dynamic model selection also fail to deliver stable improvements over the best static model in the current sample. The paper contributes by integrating loss-function sensitivity, joint tail-risk evaluation, and dynamic model choice into a unified framework, showing that volatility model comparison is fundamentally conditional on the evaluation criterion, volatility proxy, and market environment.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "revision_date": "2026-07-09", "ssrn": "6962001" }, { "id": "0055", "title": "Selected Recent Developments in GARCH Models: A Narrative Review", "authors": [ "fuli-yang" ], "type": "Working paper", "series": "volatility-econometrics", "doi": "", "date": "2026-06-17", "year": 2026, "site_date": "2026-10-27", "source": "SSRN", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6868038", "access": "Source repository access terms apply", "summary": "The Generalized Autoregressive Conditional Heteroskedasticity (GARCH) framework introduced by Bollerslev remains one of the standard tools for modeling conditional volatility in financial time series, value-at-risk measurement, and related applications. At the same time, recent literature shows that the contemporary development of GARCH-based research has been shaped by several recurring themes, including asymmetric volatility modeling, dynamic dependence structures, estimation under heavy tails, and hybrid forecasting architectures. This review provides a selective narrative survey of recent developments while restricting detailed discussion to studies whose bibliographic information can be checked or substantially corroborated. In addition to recent work, the paper includes an earlier Chinese-language contribution by Yang (2012) as a background reference for distributional assumptions in VaRoriented GARCH applications.", "summary_kind": "Author abstract", "author_abstract": true, "abstract_source": "Author manuscript", "keywords": [ "GARCH", "Volatility modeling", "VaR", "Copula-GARCH", "Heavy tails", "Narrative review" ], "jel": [], "rights": "All rights reserved for manuscript files; see source terms.", "ssrn": "6868038" } ] }