Industry Factors

Industry factors capture the systematic component of stock returns driven by membership in a particular sector or industry. Firms within the same industry share exposure to common demand cycles, input costs, regulation, and competitive dynamics. In commercial risk models, industry factors typically explain more residual variance than any single style factor.

Why industry factors exist

Companies in the same industry face correlated shocks: oil price changes for energy firms, interest rate movements for banks, regulatory action for pharmaceuticals. These industry-level drivers create return comovement that style factors alone cannot absorb. Industry factors also serve a practical role: portfolio managers need to understand and control sector bets, and risk models must attribute sector-driven tracking error separately from style tilts.

Classification systems

The choice of industry taxonomy determines factor granularity:

System	Provider	Levels	Industries (finest)
GICS	MSCI / S&P	4	163 sub-industries
BICS	Bloomberg	4	214 sub-industries
ICB	FTSE Russell	4	173 subsectors
SIC	US government	4	~1,000+ codes
NAICS	US/Canada/Mexico	5	~1,000+ codes

MSCI Barra GEM3 uses 34 GICS-based industry factors. Bloomberg MAC3 uses the BICS hierarchy, choosing a granularity level that maximizes both explanatory power and statistical significance.

Construction techniques

Binary (dummy variable) approach

The traditional method, used by Barra GEM3. Each stock receives exposure 1 to its assigned industry and 0 to all others:

$r_n={\sum }_k{X}_{nk}{f}_k+{\sum }_j{D}_{nj}{h}_j+u_n$

where $D_{nj}$ is a 0/1 indicator for industry $j$ and $h_j$ is the industry factor return. Cap-weighted industry factor returns are constrained to sum to zero.

Advantages:

• Clean, deterministic exposures with no estimation error
• Easy to interpret: a stock is simply “in” one industry
• Additive decomposition of portfolio industry bets

Limitations:

• Conglomerates and diversified firms are forced into a single industry, misrepresenting their risk profile
• All stocks within an industry are treated as equally sensitive to industry-level shocks
• A biotech startup and a mature pharma giant in the same GICS code get identical industry exposure

Beta-based approach

Used by Bloomberg MAC3. Each stock’s industry exposure is estimated by regressing its excess return against the cap-weighted industry portfolio return:

$r_{n,t}={\alpha }_n+{\beta }_n^j\cdot r_{j,t}+{ϵ}_{n,t}$

Betas are estimated with exponentially weighted least squares and trimmed (roughly 0.4 to 2.0), then standardized to cap-weighted mean 1 within each industry.

Advantages:

• A conglomerate with diversified revenue streams gets a lower industry beta than a pure-play
• Empirically increases explanatory power without adding factors
• Captures differential sensitivity to industry shocks
• Reduces spurious correlations between factor and specific returns

Limitations:

• Estimated with noise; requires sufficient return history
• Less intuitive than binary: a “healthcare” stock might have an industry beta of 0.6
• Requires regular re-estimation

Multi-industry exposure approach

Introduced in MSCI Barra GEMLT (~2016). Diversified firms receive fractional exposures to multiple industries based on business segment data. A company with 60% revenue from technology and 40% from financial services would have exposures of 0.6 and 0.4 to the respective industry factors.

Advantages:

• Directly addresses the conglomerate problem
• Uses fundamental business segment data rather than statistical estimation
• Intuitive interpretation

Limitations:

• Depends on segment reporting quality and availability
• Adds complexity to the exposure matrix
• Not all firms report segment data at sufficient granularity

Country-industry interaction

Industry composition varies substantially across countries. Technology dominates the US market; financials dominate many emerging markets. This creates confounding: is a US tech stock’s outperformance a country effect or an industry effect? Global risk models address this by jointly estimating country and industry factors in the same cross-sectional regression, with both sets of factor returns constrained to have zero cap-weighted sum. This disentanglement is one of the core contributions of global factor models over single-country models.

Granularity trade-off

Finer industry classifications explain more variance but increase the number of factors, raising estimation noise and reducing the number of stocks per industry. The optimal granularity depends on the model’s estimation universe:

• A global model with 80,000+ stocks can support 40-60 industry factors
• A single-country model with 3,000 stocks might use 10-20 sectors
• Too fine-grained: sparse industries with few stocks produce noisy factor returns
• Too coarse: lumps together firms with different risk profiles (e.g., banks and insurance)

See also

• country-vs-industry-effects for empirical evidence on the relative importance of country and industry effects
• country-factors for the analogous geographic decomposition
• msci-barra-gem and mac3-global-equity-risk-model for how industry factors are used in commercial risk models

Sources

• MAC3 Global Equity Risk Model (File)
• Barra Global Equity Model (GEM3) (File, URL)