Zong-Wei (Eric) Yeh

Publications

Empirical Research on the Taiwan Stock Market: Review and Outlook

with Hsin Yu Chiu, Kendro Vincent, and Wei-Che Tsai

Conditionally Accepted at NTU Management Review (TSSCI)

Abstract:
Taiwan’s stock market has gained global prominence, underpinned by its leadership in the semiconductor and AI sectors. Distinguished by a retail-dominated microstructure, the market offers a unique setting for empirical finance research. This paper systematically reviews four major themes: asset pricing, anomalies, mutual funds, and investor heterogeneity. The literature reveals a definitive shift from the CAPM to advanced multi-factor frameworks, with the Fama-French six-factor and q-factor models demonstrating superior explanatory power. While traditional anomalies often appear fragile, robust abnormal returns emerge when conditioned on earnings quality, trading frictions, or machine learning signals. Mutual fund research highlights distinct stock-selection skills but persistent deficits in market timing. Furthermore, investor heterogeneity plays a central role in price formation: foreign institutions lead price discovery, whereas retail investors provide liquidity while driving sentiment-induced mispricing. Future research should leverage Taiwan’s specific institutional environment to explore non-linear factor models, AI applications, and the impact of regulatory shocks.

What Drives Jumps in the Secured Overnight Financing Rate? Evidence from the Arbitrage-Free Nelson-Siegel Model with Jump Diffusion

with Dong-Jie Fang, Jie-Cao He, and Shih-Kuei Lin

Pacific-Basin Finance Journal, 2024 (NSTC ATier 2)

Abstract:
In this paper, the arbitrage-free Nelson–Siegel (NS) model with jump diffusion (AFNSJ) is proposed to describe the Secured Overnight Financing Rate (SOFR). The parameters of this model are estimated through particle filtering conducted with a weighted maximum likelihood estimation approach. The empirical results of this study indicate that the AFNSJ outperforms the arbitrage-free NS model in fitting market data. SOFR jumps are highly related to Federal Open Market Committee meetings. Moreover, even under different interest rate changes, these jumps are mainly driven by a short-term factor. The risk adjustment term can suitably capture changes in the US Federal Reserve rate caused by the jump risk component.

Delta Hedging in the USD/JPY Options Market: Insights from Implied Stochastic Volatility

with Shih-Kuei Lin, Kendro Vincent, and Chung-Jen Lin

Management Review, 2024 (TSSCI)

Corresponding Author

Abstract:
Purpose–The purpose of this paper is to explore the performance of the implied stochastic volatility (ISV) approach in fitting the implied volatility surface (IVS) of USD/JPY and to analyze the hedging performance under standard and minimum variance (MV) delta. Design/methodology/approach–In this study, the ISV approach proposed by Aït-Sahalia et al.(2021) is mainly used to construct the IVS and parameter estimation using the Heston model as the model base. Findings–Firstly, based on the Heston model, the ISV approach has better in-sample and out-of-sample fitting ability compared to the conventional approach. On the other hand, we demonstrate that the ISV approach is effective in improving the accuracy of the hedging, especially in terms of the MV delta, using a delta-neutral hedging strategy.

Working Papers

Carbon Policy Paradox: The Divergent Impacts of Short-term versus Long-term Policies

with I-Hsuan Ethan Chiang and Shih-Kuei Lin

Presentations: CICF, ESWC, FMA, EFMA, AsianFA, MSFC, EcoSta, TFA, FeAT, TRIA, STSC

Job Market Paper Corresponding Author

Abstract:
We investigate how carbon policies with different time horizons affect fossil-fuel use. Our model shows a simple yet important mechanism: near-term tightening policies raise today’s carbon cost and slow down extraction, while long-term tightening policies push costs further into the future and encourage oil producers to extract more sooner. We use a joint affine term-structure model to uncover level and slope factors from the EUA futures curve, which capture short- and long-horizon policy signals. We find a one-standard-deviation short-horizon tightening reduces European oil supply and demand by 0.3–0.4\% within three months. In contrast, an equally sized long-horizon tightening increases both series by about 0.9\% around the six-month horizon. Furthermore, short-horizon policies shift energy use toward natural gas, while long-horizon policies crowd it out, amplifying the run-up in oil use.

Altruism in P2P Lending: Evidence from Lending Club

with Dong-Jie Fang, Chien-Hsiu Lin, and Shih-Kuei Lin

Revise and Resubmit to Review of Quantitative Finance and Accounting (NSTC ATier 2)

Presentations: EFMA, EasternFA, TRIA, TFA

Abstract:
This study examines whether altruistic language in loan descriptions affects loan funding and default in peer‐to‐peer (P2P) lending. Using 119,379 Lending Club loans, we employ text mining to construct altruistic variables, including prosocial and proenvironmental intents, and linguistic metrics, including readability, tone, and deception cues. We find that loans framed with altruistic appeals are less likely to be fully funded, yet exhibiting lower default risk. In particular, prosocial content drives the bulk of this result, whereas explicitly pro‐environmental language has a muted impact on default but appears to induce greater investor scrutiny at funding. Linguistic factors further moderate these effects: more readable and positive descriptions attract more funding and lower defaults, while more deception cues increase default risk. Overall, our findings suggest a tradeoff between altruistic signaling and credit risk. Lenders discount altruistic appeals even though such borrowers hold a lower default probability.

Implied Volatility Surface Predictability: Evidence from TAIFEX Options

with Yen-Hsun Juan and Shih-Kuei Lin

Submitted to Pacific-Basin Finance Journal (NSTC ATier 2)

Presentations: FeAT

First Author

Abstract:
We ask which dimension of the implied volatility surface (IVS)-level, slope, or curvature-is predictable out of sample, and what information predicts it, using TAIEX index options over 213 out-of-sample months (2007--2024). We extract interpretable IVS factors via the quadratic-regression framework of Collin-Dufresne et al. (2024) and forecast their levels from a broad, economically organized set of 101 predictors using a balanced machine-learning spectrum-shrinkage, sparse, dimension-reduction, and tree-ensemble methods-each tuned by time-series cross-validation. Against a random-walk benchmark, predictability is concentrated in the slope(skew) factor, where the best models attain an out-of-sample $R^2_{RW}$ of +16.2% (p<0.01); the level and curvature are not robustly predictable. Dimension-reduction and tree-ensemble methods succeed while sparse linear models fail, and the predictive power is distributed across credit, monetary, options-microstructure, and sentiment channels. The signal is economically valuable: a conviction-weighted, delta-hedged risk reversal earns a net annualized Sharpe ratio of 1.08. The equilibrium volatility skew thus appears to be captured by methods conditioning on a distributed set of macro-financial and market state variables.

The Boundaries of Index Option Return Predictability: Evidence from Taiwan

with Pin-Chi Chen, Xian-Ji Kuang, and Tzu-Chieh Lin

Submitted to Pacific-Basin Finance Journal (NSTC A Tier 2)

First Author

Intermediate-Term Reversal

with Guan-Ying Huang and Kendro Vincent

Submitted to International Review of Financial Analysis (NSTC A-)

Abstract:
Long-term reversal is widely regarded as a fragile anomaly: the premium is concentrated in small and distressed firms and does not survive the standard risk adjustment. We ask whether this verdict reflects how the overreaction is measured rather than whether it occurs. Because the overreaction is an episode in the price path rather than a condition on its endpoints, we rank stocks on the maximum drawdown - the deepest peak-to-trough decline over a window that skips the most recent year - in place of the cumulative return. The sorts uncover an intermediate-term reversal that sits between the momentum continuation and the classic long-horizon effect: the high-minus-low decile portfolio earns 1.16% per month, with an eight-factor alpha of 0.87% per month. Unlike the conventional strategy, the premium survives value weighting, persists among the stocks above the NYSE median capitalization, strengthens after 2000, and remains significant once the inference is corrected for the search across seventeen measurement horizons. It is not subsumed by momentum, the firm fundamentals, or a broad set of return predictors, and it concentrates in the stocks that are costly to arbitrage.

Volatility Decay and Arbitrage in Leveraged ETFs: Evidence from the US and Japan

with Cheng-To Lin, Shih-Kuei Lin, and George Y. Wang

Presentations: TFA, FeAT, TRIA, STSC, TWSIAM

Corresponding Author

Abstract:
While financial theory suggests that shorting bear leveraged ETFs (LETFs) is the optimal way to harvest volatility decay, we find the opposite is true in the US market. A beta-neutral arbitrage strategy shorting US bull LETFs yields a Sharpe ratio as high as 2.12, whereas the theoretically superior bear-side strategy is largely unprofitable. This paper resolves this puzzle by examining volatility decay strategies across US and Japanese markets. We find that the overall pairwise beta-neutral strategy is robustly profitable, generating highly positive skewness and offering strong downside protection. The striking cross-market asymmetry is driven not by volatility decay itself, but by the “non-compounding effect”—a friction tied to the markets’ different replication technologies (swaps in the US vs. futures in Japan). Our results show that the optimal decay, harvesting strategy is necessarily asymmetric: shorting bull LETFs in the US but bear LETFs in Japan. Finally, we develop a jump-diffusion model to provide a theoretical basis for the strategy’s exceptionally large profits during market crises.

Risk Converge and Procyclicality in Futures Margin System: Evidence from Taiwan Stock Exchange Index Futures

with Ting-Da Yan, Ting-Fu Chen, and Chien-Hsiu Lin

First Author

Abstract:
Central counterparty (CCP) margin models face a fundamental trade-off between risk coverage accuracy and procyclicality. This study quantifies this trade-off by evaluating four distinct margin systems applied to Taiwan Stock Exchange Index (TAIEX) futures data. The models are constructed from combinations of Simple Moving Average (SMA) and Heston (1993) volatility estimators, specifications with Value-at-Risk (VaR) and Expected Shortfall (ES) risk measures. Our results show that while the Heston model achieves superior backtesting accuracy, it induces substantial procyclicality, generating more volatile and unstable margins than the SMA model. Across all specifications, ES consistently provides more robust protection against extreme tail losses. These findings highlight a critical dilemma for regulators, as the pursuit of statistical accuracy may introduce systemic instability. The incumbent SMA model, while less precise, possesses valuable stabilizing properties, suggesting that the choice of a margin model is a substantive policy decision involving competing objectives.

Reserve Adequacy: An Unspanned Risk Factor in Overnight Funding Markets

with Dong-Jie Fang

First Author

Learning to Fear: The Option-Implied Dynamics of Disaster Beliefs

with Bo-ai Tsai and Shih-Kuei Lin

Presentations: FeAT

First Author

The Price of Weather: Unspanned ENSO Risk in Commodity Futures

with Dong-Jie Fang

First Author

Option-Implied Probability Distortions and Stock Return Predictability

with Ting-Xuan Wang, Wei-Yu Kuo, and Shih-Kuei Lin

Presentations: TFA, FeAT, PBFEAM

Abstract:
As retail investors increasingly influence financial markets, understanding how cognitive biases affect asset prices is crucial. This study examines the return predictability and economic value of the curvature parameter of probability weighting function, which captures investors’ risk perception through probability distortions. Using a empirical pricing kernel approach with particle filtering, we construct a Probability Weighting Index (PW I) and analyze its predictive power. Empirical results suggest that PW I exhibits a certain degree of predictive power, provides incremental information beyond traditional predictors, and offers economic value through an investment strategy that outperforms the buy-and-hold approach. Further analysis indicates that PWI primarily predicts returns through the discount rate channel, distinguishing it from sentiment-based predictors. Our findings highlight the role of cognitive biases in asset pricing by introducing PW I as a potential behavioral-based predictor.

Do Investors Care About Carbon Higher Moment Risks

Single Author

Abstract:
TBA

Zong-Wei (Eric) Yeh

About Me

Publications

Working Papers

Get in Touch