Local predictive ability of analyst recommendations

• Date: 01 July 2019
• Published date: 01 July 2019
• DOI: http://doi.org/10.1002/rfe.1055
• Author: Jorida Papakroni,Serkan Karadas

Rev Financ Econ. 2019;37:351–371. wileyonlinelibrary.com/journal/rfe

|

351

© 2019 University of New Orleans

1

|

INTRODUCTION

This paper finds that aggregated analyst recommendations at Metropolitan Statistical Area (MSA-) and state-level predict

future locally aggregated stock returns. Previous studies looking at the information sets used by the sell-side analysts show

that analysts incorporate market-level, industry-level, and firm-specific information into their recommendations.1 For exam-

ple, Howe et al. (2009) find that analyst recommendations aggregated at the market-level predict future excess market returns

and that analyst recommendations aggregated at the industry-level predict future excess industry returns. Recent research also

shows that stock returns are affected by local economic fundamentals such as the unemployment rate and housing collateral

(Korniotis & Kumar, 2013). The results of our study suggest that sell-side analysts use local information when they issue rec-

ommendations.2 We use the monthly state coincident index constructed by the Federal Reserve Bank of Philadelphia as a proxy

for local economic fundamentals to run additional estimations at state-level for the geographically concentrated firms and find

a positive relationship between locally aggregated analyst recommendations and future local economic fundamentals. Overall,

these results suggest that the ability of locally aggregated analyst recommendations to predict stock returns depends on the geo-

graphic concentration of firms. Finally, our analysis also shows that the predictive power of analyst recommendations on local

returns is independent of the non-fundamental investor sentiment.

For any given month, we group firms into MSAs and states based on their headquarters’ locations, following Pirinsky and

Wang (2006). Locally aggregated returns are measured as the average returns of all stocks located in a given MSA or state. Locally

aggregated analyst recommendations are measured as the change in the average recommendations for the stocks of firms whose

headquarters are located in a particular MSA or state. Such aggregation techniques are common in the literature. For example,

Seyhun (1988) aggregates the transactions by corporate insiders and finds a positive relationship between aggregate insider trading

Received: 10 April 2018

|

Revised: 4 September 2018

|

Accepted: 23 November 2018

DOI: 10.1002/rfe.1055

ORIGINAL ARTICLE

Local predictive ability of analyst recommendations

SerkanKaradas1

|

JoridaPapakroni2

1Department of Economics,Sewanee:

The University of the South, Sewanee,

Tennessee

2Department of Business,Organizations,

and Society, Franklin & Marshall College,

Lancaster, Pennsylvania

Correspondence

Jorida Papakroni, Department of Business,

Organizations, and Society, Franklin &

Marshall College, Lancaster, PA, USA.

E-mail: jpapakro@fandm.edu

Abstract

This article shows that locally aggregated analyst recommendations at the

Metropolitan Statistical Area (MSA-) or state-level predict future locally aggregated

excess returns. The results hold even after controlling for macroeconomic variables,

industry and market returns, as well as investor sentiment. We also show that the

local predictive ability of analyst recommendations is stronger for geographically

concentrated firms. Additional analysis at the state-level for the geographically con-

centrated firms reveals that locally aggregated analyst recommendations predict fu-

ture local economic fundamentals. Overall, our findings suggest that analyst

recommendations contain information at the MSA- and state-level, and that local

information content is richer for geographically concentrated firms.

JEL CLASSIFICATION

G12, G11, G14, G24

KEYWORDS

analyst recommendations, local stock returns, MSA, return predictability

352

|

KARADAS AnD PAPAKROnI

and future stock market returns. Howe et al. (2009) apply this technique in order to aggregate analyst recommendations at na-

tion- and industry-level. The main advantage of this technique, as pointed out by Howe et al. (2009, p.800), is that it “cancels out

the idiosyncratic components of analyst recommendations and isolates their common response to systematic factors.” Similarly,

this aggregation technique applied for analyst recommendations at the local level (MSA or state) theoretically should capture the

common response of analysts to information regarding local economic factors and local stock returns, regardless of the location of

analysts. We measure locally aggregated analyst recommendations based on the location of the firms followed by the analysts. To

the extent that local analysts have superior local economic information compared to non-local analysts, we note that the information

content of analyst recommendations is diluted because we cannot distinguish between the recommendations of these two groups.3

However, even with a potentially diluted measure, we still find a significant relationship between locally aggregated recommenda-

tions and locally aggregated returns, even after controlling for important confounding economic factors.

Based on 7,851 MSA-month and 4,680 state-month observations, we find that one standard deviation increase in locally

aggregated recommendations is associated with a 2.91% and 2.15% increase in one-quarter-ahead MSA-level and state-level

excess returns, respectively. The results are robust to controlling for macroeconomic variables, market returns, and industry

returns. MSA-level regression results are stronger relative to state-level ones. MSA-level estimates are statistically significant at

the 1% level while state-level estimates are statistically significant at the 10% level. The overall evidence suggests that analysts

use local information at the MSA- and state-level in their recommendations.

We further examine whether the geographic concentration of firms affects the ability of analyst recommendations to predict

future local excess returns. García and Norli (2012) identify the list of states where a given firm has presence by counting the

number of times a state is mentioned in that firm’s filings. Using their data, we separate our sample into two groups: firms that

operate in five or fewer states and firms that operate in more than five states. We find a strong positive relationship between

locally aggregated analyst recommendations and locally aggregated returns for firms that operate in five or fewer states (i.e.,

geographically concentrated firms). The relationship between the predictive ability of locally aggregated analyst recommen-

dations and the geographic concentration of firms suggests that local economic fundamentals play a relatively larger role for

geographically concentrated firms relative to geographically dispersed firms (firms that operate in more than five states).

We further support this conjecture for geographically concentrated firms by documenting two sets of results. First, when we

control for the future changes in local economic fundamentals, the statistical and economic significance of locally aggregated

analyst recommendations in our estimations is reduced. Second, locally aggregated analyst recommendations predict the future

changes in local economic fundamentals.

Our study makes several contributions to the literature. First, it expands the body of knowledge on analyst recommendations

by showing that locally aggregated analyst recommendations contain local information at both the MSA- and state-level in

addition to market-, industry-, and firm-level information. Second, our study contributes to the literature on how geographic

location of firms affects decisions made by analysts by showing that local economic factors play a larger role in analyst rec-

ommendations for geographically concentrated firms. Finally, under the assumption that analysts are market participants, an-

alyzing the relation between stock returns and analyst recommendations adds to the literature on the predictive content of the

aggregated decisions by individual market participants.

The rest of the paper proceeds as follows. We summarize the related studies and develop the hypotheses in Section 2. We

describe the data and the construction of variables in Sections 3 and 4, respectively. Section 5 sets up the empirical framework

and discusses the regression results. Section 6 presents robustness checks, and Section 7 explores the relationship between lo-

cally aggregated analyst recommendations and local economic fundamentals. Section 8 discusses the overall implications and

the limitations of the paper, and Section 9 concludes it.

2

|

LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT

Research shows that analyst recommendations predict abnormal returns and prices.4 Elton, Gruber and Grossman (1986) and

Womack (1996) find significant positive abnormal returns associated with “upgraded” recommendations, which persist up to

6 months. Barber et al. (2001) extend the investigation of abnormal returns to consensus recommendations. They show that

a trading strategy of buying the most highly recommended stocks and short selling the least favorably recommended stocks

generates positive abnormal returns, even after controlling for market risk, size, value, and momentum effects. However, they

also find that these abnormal returns dissipate after controlling for transaction costs. Contrary to the previous studies, Jegadeesh

et al. (2004) find that the predictive ability of analyst consensus recommendations is not robust to the inclusion of other predic-

tive signals such as momentum or contrarian signals. Instead, they find that the change in analyst recommendations over the

prior quarter predicts future returns, even after controlling for the aforementioned predictive signals.