HOA-ASSETS Pricing

Assets pricing questions, machine learning applications

Who critiques the explosion of asset pricing factors and proposes higher standards for factor discovery?

Harvey, Liu, and Zhu (2016)
Theme: False discoveries in the “factor zoo” due to multiple testing and data mining.argue for more stringent criteria in empirical finance to mitigate false discoveries.

What is the main problem with traditional t-statistics according to Harvey et al.?

Standard t > 2.0 thresholds are too lenient in a multi-factor setting.
Proposal: Use t > 3.0 to reduce false positives from data mining.

What testing methods do Harvey et al. recommend to improve reliability?

• FWER: Very strict, controls even one false positive

• FDR: Allows some false positives for higher discovery power

• Bayesian: Uses prior beliefs to evaluate factor credibility

What three corrections did Hou et al.(2020) make to address bias from microcaps?

1. NYSE breakpoints to reduce microcap overload in extreme deciles

2. Value-weighted returns to reflect actual capital allocation

3. FM-WLS regression to reduce outlier influence from small firms

Hou et al (2020) Tested the replicability of anomaly-based factors using robust methods.Which anomalies passed replication and what does that imply?

Only size and momentum survived.
Conclusion: Most anomalies are not robust; markets are more efficient than previously believed.

What is IPCA and who introduced it for bond return modeling?

Kelly et al. (2023)
IPCA = Instrumented Principal Components Analysis
Goal: Capture latent risk factors in corporate bond markets

What cross-market insight does IPCA provide?

Reveals strong integration of equity and bond market risks
Systematic return components are closely aligned across markets

Who explores intraday market predictability using machine learning?

Aleti et al(2025)
Goal: Forecast 15-minute returns on market portfolios using cross-sectional factor data

Why do Aleti et al(2025). filter out return jumps in their analysis?

Jumps are unpredictable under continuous-time finance theory
Filtering helps models focus on smoother, more predictable patterns

according to Aleti et al(2025), When is intraday market return predictability strongest?

During high uncertainty and volatility
Liquidity and tail-risk factors drive short-term return patterns

Bali, et al (2023) used ML to predict delta-hedged equity option returns

They focused on capturing nonlinearities and interactions between option and stock characteristics

According to Bali, et al (2023), why is machine learning well-suited for predicting option returns?

• Captures nonlinear pricing behavior

• Handles complex interaction effects (e.g., implied volatility × moneyness)

• Mitigates overfitting in high-dimensional settings

Bali, et al (2023) – Main Findings
What predictors are most important for equity option return forecasts?

Option-based characteristics dominate
Stock-based features add value when used together with option variables

Bagnara (2024) – Survey Focused on for return prediction and factor modeling to address overfitting, dimensionality, and theory alignment. Which ML techniques address asset pricing econometric issues?

• Regularization (e.g., Lasso) reduces overfitting

• Dimension reduction (e.g., PCA, IPCA) simplifies factor structures

• Adaptive models capture nonlinearities and interactions