Figure 3, Panel C - We employ a nonlinear regression to model IO based on stock characteristics. Logistic S-shape functional form provides a convenient form to identify markets segmentation and allows us to explain 91% of variations in IO only via Capacity variables.

Segmentation in Institutional Investment: Implications for Asset pricing factors and Performance Evaluation

  • with Roger Edelen and Greg Kadlec, Working paper

We demonstrate that segmentation is an important feature of institutional investors' demand for stocks. A specification for the number of institutional owners that incorporates segmentation increases the Rsquare from 75\% to 90\% while reducing the standard list of 8-10 relevant stock characteristics to 1 (market cap). Smaller institutions exhibit greater segmentation than larger institutions, despite the argument that they face lower implementation costs. Our evidence suggests that segmentation is driven by agency considerations. Stocks that fall outside of institutions' investable universe have disappeared over time, suggesting that growing dominance of segmented institutional investment may help explain the listing puzzle.

  • Presented in Virginia Tech Seminar, Virginia Tech BXBR Workshop, Georgia State University

Figure 2, Panel A- This graph presents the fraction of stocks within the exogenous decile breakpoints. The fraction of stocks in each exogenous ’decile’ bin is approximately 10% pre-millennium, but it decreases sharply for small stocks post-millennium.

Segmentation in Institutional Investment: Implications for Asset pricing factors and Performance Evaluation

  • with Roger Edelen and Greg Kadlec, Working paper

We examine how IO segmentation influences pricing factors and find support for the demand-based asset pricing view (e.g., Koijen and Yogo, 2019, and Gompers and Metrick, 2001). Second, segmentation implies a constrained portfolio optimization and therefore impaired performance. Capacity frictions can lead to narrow investment opportunity sets, which limit money managers’ ability to take advantage of profitable opportunities outside their investment segment. In this respect, we construct pricing factors (i.e., IO factors) that are feasible (ex-ante) for institutions and benchmark their performance.

  • Presented in Virginia Tech BXBR Workshop, University of Illinois at Chicago, University of Technology Sydney, University of Sydney, Florida International University, Monash University, Deakin University, and Queensland University.

Figure 2 - Instrumented Principal Component Analysis (IPCA) allows factor loadings to partially depend on observable asset characteristics that serve as instrumental variables for the latent conditional loadings (KPS). For example, this graph illustrates a mapping between stock characteristics and the first factor loading.

Preferred Habitat in Institutional Ownership and Equity Costs of Capital

  • with Roger Edelen and Bradley Paye, Work-in-progress paper

Kelly et al. (2019) (KPS) use thirty-six stock characteristics to instrument five principal components of risk (latent return factors), as well as the time-varying exposure of each sample-firm to those risks. Equity discount rates derived from this model are both effective (outperform state-of-the-art factor models) and intuitive (tied to observable firm characteristics). Our study uses the KPS framework to construct instrumented principal components for institutions’ equity demands (latent IO-habitat factors). We then evaluate (1) the degree to which IO habitat subsumes the latent return factors in KPS and (2) the extent to which the orthogonal component of IO habitat explains equity returns.