Author: Vikalp Thukral Program: UCLA Anderson MFE III — Quantitative Asset Management Course Instructor: Prof. Bernard Herskovic Term: Spring 2025
This repository consolidates five major projects completed as part of the Quantitative Asset Management course (MGMTMFE 431). Each problem set replicates a canonical empirical asset-pricing factor or model — from the market and risk-free components through size, value, momentum, and finally betting against correlation (BAC).
The sequence reflects both theoretical depth and technical rigor: beginning with data handling and portfolio formation using WRDS/CRSP/Compustat data, progressing to multi-factor replication, and culminating in an original factor-engineering project.
| Folder | Factor / Theme | Core Objective |
|---|---|---|
| PS1/ | Market and Risk-Free Factor Construction | Construct market (MKT-RF) and risk-free (RF) series from CRSP monthly returns; replicate CAPM components used in Fama–French factor models. |
| PS2/ | Profitability and Investment Factors (RMW, CMA) | Recreate the Fama–French 5-Factor extensions (2015). Compute RMW (robust-minus-weak) and CMA (conservative-minus-aggressive) factors using Compustat accounting data and CRSP linkages. |
| PS3/ | Momentum Factor Replication | Reproduce the momentum anomaly as in Daniel & Moskowitz (2016) and Kenneth French. Define ranking returns, form decile portfolios, and validate the WML (winners-minus-losers) factor versus published benchmarks. |
| PS4/ | Size and Value Factor Replication (SMB & HML) | Integrate CRSP and Compustat to replicate SMB (small-minus-big) and HML (high-minus-low) factors from Fama & French (1992, 1993). Build decile portfolios on market equity and book-to-market, test against Ken French library data. |
| FinalProject/ | Betting Against Correlation (BAC) | Design and backtest the BAC factor, going long low-correlation stocks and short high-correlation stocks. Implement rolling correlation estimation, beta-neutral portfolio formation, and performance attribution vs. standard risk factors. |
- Derived the market excess return (MKT-RF) by aggregating CRSP equity data.
- Constructed a consistent risk-free rate series aligned with Fama–French inputs.
- Verified that the computed market portfolio tracks the official Fama–French series closely.
- Replicated RMW (robust minus weak) and CMA (conservative minus aggressive) factors using firm-level operating profitability and total asset growth.
- Created cross-sectional deciles on profitability and investment, value-weighted monthly.
- Compared replication statistics (mean, volatility, Sharpe, correlation) to the French 5-factor library.
- Constructed rolling ranking returns based on 12-month past performance, skipping the most recent month.
- Formed decile portfolios both under Daniel–Moskowitz and Kenneth French definitions.
- Calculated long–short WML returns and validated performance vs. benchmark data (correlations >0.99 across deciles).
- Discussed momentum crashes and implementation limits in recent years.
- Merged CRSP and Compustat via the CCM link table to compute market equity (ME) and book equity (BE).
- Formed decile portfolios by NYSE breakpoints for ME and BE/ME ratios.
- Constructed SMB and HML factors and evaluated replication accuracy vs. Ken French portfolios (correlations ≈0.99).
- Analyzed factor consistency over time — confirming persistence of size and value premia but attenuation in the post-2010 period.
- Implemented a novel factor isolating the correlation component of beta, as in Frazzini–Pedersen-style decompositions.
- BAC goes long low-correlation, short high-correlation stocks; scaled to maintain dollar- and beta-neutrality.
- Computed rolling market correlations, formed monthly BAC portfolios, and backtested performance.
- Demonstrated that BAC captures the correlation-risk premium independent of volatility, with low market beta and moderate positive alpha over 2010–2024.
- Benchmarked BAC against standard Fama–French factors and observed low correlation, indicating orthogonal return drivers.
- Data Sources: WRDS (CRSP, Compustat North America), Kenneth French Data Library.
- Portfolio Construction: Decile-based, value-weighted, NYSE breakpoints, July–June rebalancing.
- Evaluation Metrics: Annualized mean, volatility, Sharpe ratio, skewness, and correlations with official Fama–French series.
- Factor Validation: Replication correlations >0.95 across all factors (MKT, SMB, HML, RMW, CMA, WML).
- Codebase: Modular Python notebooks for data ingestion, cleaning, portfolio formation, and result visualization.
- Fama, Eugene F., and Kenneth R. French (1992, 1993, 2015) — Common Risk Factors in the Returns on Stocks and Bonds.
- Daniel, Kent, and Tobias Moskowitz (2016) — Momentum Crashes, Journal of Financial Economics.
- Kenneth R. French Data Library (U.S. Factor Portfolios).
- WRDS (CRSP, Compustat, CCM Link Tables).
- Frazzini, A., and Pedersen, L. H. (2014) — Betting Against Beta and correlation-based decompositions.