Sector Investment Growth Rates and the Cross Section of Equity Returns*

Qing Li, Maria Vassalou, Yuhang XingThe Journal of Business 2006 · DOI

TL;DR

This paper introduces a novel asset pricing model that uses sector-specific investment growth rates as factors to explain stock returns. The authors demonstrate that this "sector investment growth model" significantly outperforms traditional models like CAPM and Fama-French, particularly in explaining the returns of small growth portfolios, which have historically been difficult to price.

What Problem Does This Paper Address?

For decades, financial economists have sought to understand what drives the differences in stock returns across various companies and portfolios. The Capital Asset Pricing Model (CAPM) was an early attempt, but it proved insufficient. The Fama-French (FF) model, incorporating factors for size and book-to-market value, became the industry standard, but questions remained about its theoretical underpinnings and its ability to fully explain all market phenomena. Specifically, there's a desire to link asset returns to macroeconomic factors, moving beyond purely return-based factors. Cochrane's work on investment-based asset pricing offered a new direction, but aggregating investment across broad sectors (like residential and non-residential) still left room for improvement. This paper tackles the challenge of finding more granular and explanatory factors for equity returns, proposing that disaggregating investment growth by economic sector is key.

Methods

The authors employ two primary econometric methodologies:

  • Stochastic Discount Factor (SDF) approach: This is implemented using the Generalized Method of Moments (GMM). The core idea is to estimate a "pricing kernel" that relates asset returns to economic factors. They incorporate sector investment growth rates as these factors.
  • Fama-MacBeth (FM) regressions: This is a classic two-step approach for testing asset pricing models. First, time-series regressions are run to estimate factor betas for each portfolio. Second, cross-sectional regressions are run to estimate the risk premiums associated with these factors.

To assess model performance, they use standard statistical tests like Hansen's J-statistic (for overidentifying restrictions) and the Hansen-Jagannathan (HJ) distance measure. They also conduct robustness tests, including examining the model's performance with scaled returns and excluding influential observations.

Datasets

The primary dataset used in this study consists of:

  • Sector Investment Growth Rates: Derived from the Federal Reserve Board Statistical Releases, covering the period from 1952Q1 to 2000Q4. These are broken down into five sectors: Households and Nonprofit Organizations (HHOLDS), Non-farm Nonfinancial Corporate Business (NFINCO), Non-farm Noncorporate Business (NONCOR), Farm Business (FARM), and Financial Business (FINAN).
  • Fama-French (FF) 25 Portfolios: These are portfolios formed by sorting stocks based on size and book-to-market ratios. Data was obtained from Kenneth French's website.
  • Market Portfolio Returns: Value-weighted returns of all stocks in the CRSP database.
  • Three-month T-bill rate (RF): Obtained from CRSP.
  • Dividend Yield: Calculated based on CRSP data.
  • Default Premium: The difference between Moody's Seasoned yields on Baa and Aaa corporate bonds, sourced from the Federal Reserve Economic Data.

Key Insights and Ideas

  • Granularity Matters: Disaggregating investment growth into sector-specific components is crucial. Aggregating investment across broad sectors, as in previous models, masks important variations that drive equity returns.
  • Real Investment Sectors are Key: Investment growth in sectors involving "real" economic activity (like NFINCO, NONCOR, and HHOLDS) appears to be more important for explaining cross-sectional returns than the financial sector (FINAN). This suggests that the drivers of real economic growth are more directly linked to asset pricing.
  • Small Growth Portfolios are Priced: The proposed sector investment growth model excels at explaining the returns of small growth portfolios, a notoriously difficult task for existing models. This is a significant contribution, as these portfolios often represent companies with unique risk profiles.
  • HML and SMB Become Redundant: When sector investment growth factors are included, the traditional Fama-French factors (HML and SMB) lose their explanatory power, suggesting that the new factors capture the information that HML and SMB were previously proxying for.

Key Results

  • The proposed four-factor sector investment growth model explains 87.8% of the cross-sectional variation in the average returns of the 25 FF portfolios, significantly outperforming CAPM (6.1%), FF model (65.4%), and Cochrane's model (0%).
  • The model is not rejected by Hansen's J-test (p-value of 0.987) and the HJ-distance measure (p-value of 0.893), indicating it prices the 25 assets correctly.
  • The model's parameters are found to be stable over time.
  • The sector investment growth model also performs well when pricing alternative sets of test assets, scaled by dividend yield and default spread, though it performs better with dividend yield scaling.
  • The model's performance is robust to the exclusion of influential observations, suggesting its findings are not sample-specific.

Limitations and Future Work

  • Emerging Markets: The authors note that their findings, particularly regarding the financial sector's role, might not apply to emerging markets where financial sector developments could play a more central role in economic growth.
  • Theoretical Underpinnings: While the paper focuses on empirical performance, further theoretical work could explore how these sector-specific investment growth rates are derived from fundamental economic principles.
  • Data Limitations: The appendix mentions that data for some sectors (like nonprofit organizations) are not always available separately, requiring assumptions.

The study's success in disaggregating investment growth information indirectly supports multi-sector business cycle models, suggesting future research could explore these models further for their asset pricing implications.

Summary updated

Abstract

We examine the importance of the information contained in sector investment growth rates for explaining the cross section of equity returns. We propose an empirical specification that outperforms the capital asset pricing model and Cochrane's (1996) model and performs at least as well as the Fama-French (1993) and Lettau-Ludvigson (2001) models in explaining the 25 Fama-French size-sorted and book-to-market-sorted portfolios, as well as other sets of test assets.
BibTeX
@article{li2006sector,
  author = {Qing Li and Maria Vassalou and Yuhang Xing},
  title = {{Sector Investment Growth Rates and the Cross Section of Equity Returns*}},
  journal = {The Journal of Business},
  year = {2006},
  doi = {10.1086/500687},
}
Figure