Numinor
Use Cases
Industry Analysis
3 min readJanuary 22, 2026

Proportional Sector Exposure: Precision Risk Management with Revenue Data

Modern companies don't fit neatly into single industry boxes—diversified businesses span multiple sectors. Learn how proportional industry exposures based on actual segment revenues enable more precise risk models and cleaner sector rotations.

Datasets Used
SAM Value Chain

The Question

Barra risk models and sector-neutral portfolio construction rely on a simplifying assumption: every company belongs to exactly one industry. But conglomerates, diversified manufacturers, and multi-segment businesses don't fit this mold.

A company deriving 40% revenue from industrials, 35% from materials, and 25% from technology gets forced into one bucket—usually the plurality segment. But that "industrial" label masks two-thirds of its actual business risk.

When you neutralize industry exposure in portfolio construction, are you actually neutral? Or are you inadvertently taking concentrated bets on the hidden segments of misclassified companies?

The Approach

We modify traditional Barra-style factor models by replacing single-industry dummy variables with proportional multi-industry exposures derived from SAM segment revenue data.

Instead of assigning a stock binary industry membership (1 for "Technology," 0 for everything else), we assign fractional exposures summing to 1:

  • 40% Industrials
  • 35% Materials
  • 25% Technology

For factor return estimation, we run weighted least squares regressions with market-cap weights and constraints that industry exposures sum to 1 for each stock. This produces cleaner industry factor returns that reflect actual revenue exposure, not just plurality assignment.

The test: Does this approach improve risk model explanatory power and produce more intuitive industry factor returns?

The Finding

Proportional industry classification produced more balanced and stable industry factor returns. Standard deviation of returns across industries decreased—reducing extreme outliers caused by misclassification of diversified firms. Inter-industry correlation also dropped, suggesting the approach better isolates true sector effects from company-specific noise.

Risk model performance improved for diversified companies. Conglomerates with significant multi-sector revenue saw better fit between predicted and realized risk when using proportional exposures. This has direct implications for portfolio construction: if your risk model understates exposure to hidden segments, your "sector-neutral" portfolio isn't actually neutral.

Style factor returns remained stable across both approaches, indicating that the proportional method doesn't distort non-industry factors—it purely refines sector attribution.

From a practical standpoint, the approach more accurately describes economic reality. When a "technology" company derives 50% of revenue from manufacturing services, treating it as pure tech for factor analysis is simply wrong. The proportional method corrects this at scale.

Try It Yourself

Implementing proportional sector exposures requires restructuring your factor model pipelines—but the infrastructure investment pays dividends across multiple use cases:

  • Sector-neutral portfolio construction: Build truly neutral portfolios that account for diversified business models, not just plurality industry codes
  • Sector rotation strategies: Execute cleaner rotations by precisely controlling exposure to each segment, avoiding unintended bets from misclassified stocks
  • Risk decomposition: Attribute portfolio risk to the actual industry drivers, not oversimplified single-sector labels
  • Event studies: When analyzing sector shocks (commodity price moves, regulatory changes), correctly identify which companies are truly exposed based on revenue mix

This approach works best for institutional investors with in-house risk models (Barra, Axioma, custom implementations) who can modify factor construction logic. For those using vendor models, SAM data can still improve stock selection by screening for segment-level characteristics.

Ready to refine your risk model with revenue-based industry exposures? Book a call to discuss data integration and factor model architecture.

Want to explore this with your own data?

We'll walk you through the methodology, provide sample code, and help you adapt this approach to your specific research questions.

Book a Call

Related Use Cases