Macro Factor Model for US Sector Rotation

This repository contains a complete, end-to-end research pipeline for a macro-driven equity allocation model. It uses:

• US macroeconomic data from FRED
• Monthly returns of SPY and US sector ETFs
• Rolling OLS factor models estimated on macro factors
• Out-of-sample backtests for:
  • A SPY timing strategy (risk-on vs cash)
  • A macro-guided long/short sector rotation strategy (V2)

The project is framed as a research notebook, not as production trading code. The emphasis is on transparent design choices, avoiding look-ahead bias, and understanding why the model behaves the way it does.

---

1. Project Overview

The core question:

Can a small set of macroeconomic indicators systematically improve equity allocation decisions?

To answer this, the project builds a macro factor model that links US macro data to equity returns, then uses that model to construct and backtest simple systematic strategies.

Key elements:

• Universe:

  • SPY (broad US equity market)
  • 10 US sector ETFs: XLB, XLE, XLF, XLI, XLK, XLP, XLRE, XLU, XLV, XLY

• Macro inputs (from FRED, resampled to month-end):

  • Growth: Industrial Production (INDPRO)
  • Labour: Unemployment Rate (UNRATE)
  • Inflation: CPI (CPIAUCSL)
  • Policy: Fed Funds (FEDFUNDS)
  • Yield curve: 10Y–2Y spread (T10Y2Y)
  • Risk sentiment: VIX (VIXCLS), Baa yield (BAA)
  • Liquidity / USD: M2 (M2SL), Fed balance sheet (WALCL), USD broad index (DTWEXBGS)

• Factors:

  • YoY growth and inflation (IP_YOY, CPI_YOY)
  • 6-month changes in those YoY rates (acceleration signals)
  • Unemployment gap vs 5-year average
  • 10-year z-scores of yield curve and credit spread
  • 2-year z-score of VIX
  • Real policy rate (REAL_FF = FEDFUNDS – CPI_YOY)
  • Liquidity and USD YoY changes

• Models:

  • Rolling 84-month (7-year) OLS regressions on lagged macro factors
  • Excess returns (over Fed Funds) as the base return measure
  • SPY model: absolute excess return vs macro
  • Sector model: relative excess return vs SPY vs macro

• Strategies:

  1. SPY timing – long SPY when forecast excess return > 0, otherwise cash
  2. Sector L/S (V2) – macro-driven cross-sectional sector rotation:
     • Forecast sector relative returns vs SPY
     • Long top 3 sectors, short bottom 3 (conditional on trend)
     • Trend filter: only short sectors with P_t < MA_10(P)
     • Inverse-volatility weighting for both long and short baskets
     • Dollar-neutral portfolio