Roll Measure

Quick Reference

Property	Value
Dimension	signal
Category	liquidity
Version	v0.9.0 (Beta)
Output Column	roll_measure

Roll measure proxy: negative autocorrelation of bar-to-bar price changes

Formula

sqrt(clip(neg_autocorr(diff(close, 1), window), 0.0))

CDM Inputs

Column	CDM Table	Description
close	cdm_*	CDM source table

Parameters

Parameter	Type	Default	Description
window	integer [5, 200]	20	Rolling window for autocorrelation

Output

Column: roll_measure

Roll effective spread proxy from price change autocorrelation

Market Intuition & Trading Rationale

Roll measure estimates the effective bid-ask spread from price change autocorrelation: sqrt(-autocorr(Δprice, window)). Roll (1984) showed that bid-ask bounce creates negative serial correlation in price changes — a trade at the ask pushes price up, then a trade at the bid pushes it down, creating an oscillating pattern. The magnitude of this oscillation reveals the effective spread, even without quote data.

Usage Cases

• Spread estimation without quote data: Estimate transaction costs for markets or historical periods where quote data is unavailable. The Roll measure is the standard approach in academic research.
• Liquidity comparison across markets: Compare Roll measures across instruments or time periods. Higher Roll measure = wider effective spread = higher transaction costs.
• Transaction cost estimation: When working with trade-only data (no quotes), use Roll measure to estimate realistic transaction costs rather than assuming fixed costs.

YAML Definition

name: roll_measure
description: 'Roll measure proxy: negative autocorrelation of bar-to-bar price changes'
category: liquidity
version: v0.9.0 (Beta)
dimension: signal
status: Pre-release
required_inputs:
  - close
output_column: roll_measure
output_description: Roll effective spread proxy from price change autocorrelation
parameters:
  window:
    type: integer
    description: Rolling window for autocorrelation
    required: false
    default: 20
    constraints:
      min: 5
      max: 200
formula: sqrt(clip(neg_autocorr(diff(close, 1), window), 0.0))