Using equity analyst coverage to determine stock similarity

TY - GEN

T1 - Using equity analyst coverage to determine stock similarity

AU - Yaros, John Robert

AU - Imielinski, Tomasz

N1 - Publisher Copyright: © 2014 IEEE.

PY - 2014/10/14

Y1 - 2014/10/14

N2 - With the observation that equity analysts tend to cover similar stocks, we propose a simple, intuitive method to convert their coverage sets into pairwise similarity values among stocks. These values are shown to have a strong positive relationship with future stock-return correlation. Further, these values are easily combined with historical correlation. Together, they produce more accurate predictions of future correlation than either does separately. Using an agglomerative clusterer and a genetic algorithm in a pipeline approach, we use the pairwise values to form clusters of similar stocks. We compare these clusters against a leading industry classification system, GICS, finding that the clusters from the combined analyst and correlation pairwise values tend to perform at least as well as GICS and often better. In an application of our pairwise values, we consider a hypothetical scenario where an investor wishes to hedge a long position in a single stock. Our results indicate that using the analyst similarity values to select a hedge portfolio leads to greater risk reduction than using GICS or hedging with a broad-market index. Using a combination of historical correlation with the analyst values leads to even greater improvements.

AB - With the observation that equity analysts tend to cover similar stocks, we propose a simple, intuitive method to convert their coverage sets into pairwise similarity values among stocks. These values are shown to have a strong positive relationship with future stock-return correlation. Further, these values are easily combined with historical correlation. Together, they produce more accurate predictions of future correlation than either does separately. Using an agglomerative clusterer and a genetic algorithm in a pipeline approach, we use the pairwise values to form clusters of similar stocks. We compare these clusters against a leading industry classification system, GICS, finding that the clusters from the combined analyst and correlation pairwise values tend to perform at least as well as GICS and often better. In an application of our pairwise values, we consider a hypothetical scenario where an investor wishes to hedge a long position in a single stock. Our results indicate that using the analyst similarity values to select a hedge portfolio leads to greater risk reduction than using GICS or hedging with a broad-market index. Using a combination of historical correlation with the analyst values leads to even greater improvements.

UR - http://www.scopus.com/inward/record.url?scp=84908123161&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84908123161&partnerID=8YFLogxK

U2 - https://doi.org/10.1109/CIFEr.2014.6924101

DO - https://doi.org/10.1109/CIFEr.2014.6924101

M3 - Conference contribution

T3 - IEEE/IAFE Conference on Computational Intelligence for Financial Engineering, Proceedings (CIFEr)

SP - 399

EP - 406

BT - 2014 IEEE Conference on Computational Intelligence for Financial Engineering and Economics, CIFEr Proceedings

A2 - Almeida, Rui Jorge

A2 - Maringer, Dietmar

A2 - Palade, Vasile

A2 - Serguieva, Antoaneta

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2014 IEEE Conference on Computational Intelligence for Financial Engineering and Economics, CIFEr 2014

Y2 - 27 March 2014 through 28 March 2014

ER -