Fred Houze Li

Research

Flagship

Index Rebalancing, Market Friction & Implementation Shortfall

interactive study + white paper · Dec 2025 – 2026

A survivorship-free event study of 1,219 1,219 Constituent-change events in the study panel. Authors' calculations (paper, Section 2) STOXX Europe 600 rebalancing events. The market front-runs the rule-based reconstitution list: +410 +410 95% CI [+290, +552] Pre-announcement run-up (selection-list). Placebo q=1.00; bootstrap CI excludes 0. Authors' calculations (paper, Sections 3–4) bps of median run-up accrues before the addition is announced, while the residual left after it is statistical noise. For the passive funds obliged to replicate the list, that friction costs a tracker 2.8 2.8 bps/yr No-reversal bound (s=0.70). Authors' calculations (paper, Section 6) –5.1 5.1 bps/yr Full-reversal (s-free) bound. Authors' calculations (paper, Section 6) bps a year.

+410 +410 95% CI [+290, +552] Pre-announcement run-up (selection-list). Placebo q=1.00; bootstrap CI excludes 0. Authors' calculations (paper, Sections 3–4) bps pre-announcement 1,219 1,219 Constituent-change events in the study panel. Authors' calculations (paper, Section 2) events 2.8 2.8 bps/yr No-reversal bound (s=0.70). Authors' calculations (paper, Section 6) –5.1 5.1 bps/yr Full-reversal (s-free) bound. Authors' calculations (paper, Section 6) bps/yr tracker cost
Read the interactive study White paper (PDF)

The Second Moment of the Index Effect

interactive study + working white paper · Dec 2025 – 2026 · companion to “The Index Rebalancing”

A two-sided, correction-battery study of comovement around STOXX Europe 600 reviews. Added stocks' daily beta rises. It is real at daily frequency, gone under every synchronicity correction, and traced to a +26.2% +26.2% exp(mean Δlog turnover)−1, printed as a MEAN shift; genuine (price-free) turnover rise on additions; n=278 Authors' calculations (see paper) jump in genuine turnover rather than any change in fundamentals. The one battery-robust window, 2018–21, sits on flat passive AUM and is gone during the fastest passive growth on record. Surviving demotions lose neither comovement nor trading.

289 289 usable scheduled additions (cohort 421), G=46 Authors' calculations (see paper) additions 285 285 always quoted as '285 usable (365 survivors / 406 scheduled)' — never '285 survivors' Authors' calculations (see paper) usable demotions (365 365 survivors among 406 scheduled deletions Authors' calculations (see paper) survivors / 406 406 scheduled deletion cohort Authors' calculations (see paper) scheduled) 46 46 review cycles clustering the headline battery (G=46); the AUM series spans 49 cycles 2014Q1–2026Q1 Authors' calculations (see paper) review cycles
Read the interactive study Working paper (PDF)

The Decrement Index That De-levers Itself

Working paper · index engineering & design · companion to the index-effect studies

A decrement index subtracts a synthetic dividend from an underlying each day. Fixed in index points, that deduction behaves like a leveraged short: as the level falls the same points cost a larger share, so the effective fee climbs into a crash — from 5.00% 5.00% launch design rate Authors' calculations (see paper) at launch to 15.32% 15.32% 2020 crash trough (2020-04-21) Authors' calculations (see paper) at the 2020 trough on one family. Six 6 certified indices rebuilt to the published cent Authors' calculations (see paper) published families are rebuilt here to the published cent; a hybrid deduction, min(d, ρ·IV), de-levers the fee by construction.

6 6 certified indices rebuilt to the published cent Authors' calculations (see paper) families rebuilt 1,858 1,858 per-step own-input tick-window reproduction, zero exclusions Authors' calculations (see paper) index-days to the cent 153 153 active single-stock decrement lines on the public VCS sheet (survivorship caveat: active-lines only) Authors' calculations (see paper) single-stock lines
Read the interactive study Working paper — on request

Hull-White Model Calibration for ATM Caplets and Caps

Graduate fixed-income coursework (2024) · recalibrated 2026

dr(t)=(θ(t)−a⋅r(t)) dt+σ⋅dW(t)θ(t)=∂f(0,t)∂t+a⋅f(0,t)+σ22a(1−e−2at)\begin{aligned} &dr(t) = \big(\theta(t) - a \cdot r(t)\big)\, dt + \sigma \cdot dW(t) \\[4pt] &\theta(t) = \frac{\partial f(0,t)}{\partial t} + a \cdot f(0,t) + \frac{\sigma^2}{2 a}\left(1 - e^{-2 a t}\right) \end{aligned}

A one-factor Hull-White model calibrated to a USD ATM caplet strip. Correctly targeted, it prices the book to 1.4 vol pts 1.4 vol pts mean absolute model-vs-market implied-vol gap at the recalibrated optimum export_web_json.py stage 4 of implied volatility. But the accuracy is partly bought by a free r(0) tilting the discount curve, and pinned to the curve the model cannot produce the rising normal-volatility term structure the market implies. The miss is structural. It prices σ(t) or a second factor.

1.4 vol pts 1.4 vol pts mean absolute model-vs-market implied-vol gap at the recalibrated optimum export_web_json.py stage 4 implied-vol error, r(0) free 11.6 vol pts 11.6 vol pts IV MAE under the pinned-r0 restriction export_web_json.py stage 4 the miss, r(0) pinned 118 118 caplets entering the corrected calibration (row 0 has no preceding discount, matching the archived Black loop) export_web_json.py stage 3 caplets priced
Explore the calibration GitHub

Earlier research

Research Assistant · WashU Olin & Cornell · 2023 — 2024

As a paid Research Assistant, I built the empirical plumbing for three political-economy working papers: BERTopic modelling over ~50 years of U.S. congressional-hearing transcripts (embedding pretraining, UMAP, HDBSCAN, c-TF-IDF), sentiment-trend anomaly detection, and a UK budget-shock event study addressing omitted-variable bias. The methods and pipeline are mine to discuss. The findings belong to the PI's forthcoming papers.