And Pythia said: ``Buy not sell''; An analysis of analysts' recommendations betting on sparsity

38 Pages Posted: 6 Apr 2022

See all articles by Veni Arakelian

Veni Arakelian

Piraeus Bank; UCL Centre for Blockchain Technologies

Mike Tsionas

Lancaster University

Date Written: March 18, 2022

Abstract

The goal of this paper is to identify influential analysts who generate abnormal returns when issuing a new recommendation by building a model that bets on the sparsity of typical analysts' recommendation data. Based on Bayesian techniques, we estimate a regression model for the abnormal returns in conjunction with a time-varying Markov switching model for the analysts' recommendations, using the $\alpha$--stable distribution as prior, and we find that the influential analysts are very few. Additionally, we identify publicly available information that contributes to the abnormal returns besides the analysts' recommendations. Moreover, we study the analysts' herding behavior as an application to exemplify the merits of our method. Our findings show that analysts' herding behavior is not pervasive when the model accounts for the deviation of the analysts' recommendations from the prevailing consensus. Finally, we show that our model performs better than LASSO, elastic net, and the horseshoe prior.

Keywords: Bayesian, Herding, Horseshoe prior, One-sided stable distributions, Shrinkage methods

JEL Classification: c11, c55, g14,g24

Suggested Citation

Arakelian, Veni and Tsionas, Efthymios G., And Pythia said: ``Buy not sell''; An analysis of analysts' recommendations betting on sparsity (March 18, 2022). Available at SSRN: https://ssrn.com/abstract=4061058 or http://dx.doi.org/10.2139/ssrn.4061058

Veni Arakelian (Contact Author)

Piraeus Bank ( email )

Vasilissis Sofias 94 & Kerasountos 1
Athens, Attica 11528
Greece
11528 (Fax)

UCL Centre for Blockchain Technologies ( email )

Malet Place
London, London WC1E 6BT
United Kingdom

Efthymios G. Tsionas

Lancaster University ( email )

Lancaster LA1 4YX
United Kingdom

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
74
Abstract Views
366
Rank
576,227
PlumX Metrics