The time‐varying performance of UK analyst recommendation revisions: Do market conditions matter?

• Author: Robert S. Hudson,Chen Su,Hanxiong Zhang
• Published date: 01 May 2020
• Date: 01 May 2020
• DOI: http://doi.org/10.1111/fmii.12126

DOI: 10.1111/fmii.12126

ORIGINAL ARTICLE

The time-varying performance of UK analyst

recommendation revisions: Do market conditions

matter?

Chen Su1Hanxiong Zhang2Robert S. Hudson3

1Newcastle University Business School

2Sheffield University Management School

3Hull University Business School

Correspondence

ChenSu, Newcastle University Business School, 5

BarrackRoad, Newcastle upon Tyne NE1 4SE, UK.

Email:chen.su@ncl.ac.uk

Fundinginformation

NewcastleUniversity Business School; Fac-

ultyof Humanities and Social Sciences, New-

castleUniversity, Grant/AwardNumber:

OSR/0372/FR19/0001

Abstract

This study examines the time-varying performance of investment

strategies following analyst recommendation revisions in the UK

stock market, with specific emphasis on the impact of changing mar-

ket conditions. We find a negative relationship between the rec-

ommendation performance and market conditions as measured in

terms of past market return and market volatility.In particular, the

upgrade (downgrade) portfolio generates significantly positive (neg-

ative) net abnormal returns in bad market conditions (e.g., the dot-

com bubble burst in 2000 and the credit crisis in 2007), but not in

other periods of time. Moreover,our non-temporal threshold regres-

sion analysis shows that the reported negative relationship disap-

pears when market conditions become better, i.e., when the past

marketreturn (market volatility) is higher (lower) than a certain level,

indicating the importance of taking non-linearity into account in the

long sample period as examined in this study.Our time-series boot-

strap simulations further confirm that the superior recommendation

performance in bad market conditions is not due to random chance;

analysts have certain skills in making valuable up/downward revi-

sions in bad markets.

KEYWORDS

analyst recommendation revisions, bootstrap simulations, market

conditions, non-temporal threshold regression model

JEL CLASSIFICATION

G11, G12, G14, G24

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and repro-

duction in anymedium, provided the original work is properly cited.

c

2020 The Authors. Financial Markets, Institutions & Instruments published by New York University Salomon Center and Wiley

Periodicals LLC

Financial Markets,Inst. & Inst. 2020;29:65–89. wileyonlinelibrary.com/journal/fmii 65

66 SU ET AL.

1INTRODUCTION

There are numerous studies on the information role of financial analysts, while the existing literature mostly ignores

thequestion of whether the performance of their stock recommendations is related to the state of the economy (Chang

& Choi, 2017; Loh & Stulz, 2018). Loh and Stulz (2018) provide empirical evidence showing that analysts’ advice (e.g.,

stock recommendations and earnings forecasts) is more valuable in bad market conditions. They further argue that, in

bad marketconditions, analysts work harder and market investors rely more on analysts’ advice than on other informa-

tion sources. This argument, however,does not seem to align with the common sense view that bad market conditions,

such as financial crises and recessions, usually give rise to increased uncertainty, making it much harder for analysts

to make accurate stock recommendations (see, Amiram, Landsman, Owens, & Stubben, 2018; Bloom, 2009; Chopra,

1998). For example, Barber,Lehavy, McNichols, and Trueman (2003) report that in the years of 2000 and 2001 after

the dot-com bubble burst, the most (least) favorableanalyst recommendations lead to an average annualized abnormal

return of –7.06% (13.44%).

In contributing to this debate, we shed fresh light on the time-varying performance of UK analyst recommendation

revisions, with specific emphasis on the impact of market conditions. Our particular attention to the UK stock mar-

ket is motivated by the following considerations. Specifically, despite the existenceof extensive analyst research in

the US market, empirical evidence on the performance of analyst recommendations remains mixed and far from con-

clusive (see relevant literature in Section 2). Jegadeesh and Kim (2006, p. 275) call for an in-depth investigation into

other developed marketsto provide “a more comprehensive picture of the extent to which the unique skills of analysts

are useful for investors”.The UK stock market, as a highly developed and sophisticated market, offers an appropriate

setting to examine the recommendation performance and its implication on marketefficiency. In particular, its institu-

tional settings and tradingpractices are partially different from and independent of those in the US market;1as a result,

the existing US evidence maynot justify the UK investment practices. However, there is surprisingly little related ana-

lyst research in the UK stock market, exceptfor two published studies that report conflicting results, using very small

samples over relatively short time periods. For example,Dimson and Fraletti (1986) examine an unpublished sample

of 1,649 telephone recommendations made by one UK brokerin 1983 and find no significant abnormal returns for the

recommended stocks. Ryan and Taffler(2006, p. 372) argue that the sample of analyst recommendations employed in

Dimson and Fraletti (1986) is made by “a single UK brokeragehouse only and is biased towards large capitalization

stocks”.Ryan and Taffler (2006) investigate 2,506 analyst recommendation revisions made by six London-based bro-

kers from December 1993 to June 1995, showing that stock prices are significantly affected by the changes in analyst

recommendations.

This study examines a comprehensivesample of 70,220 UK analyst recommendation revisions over the period Jan-

uary 1995 to June 2013. As such, our sample is much larger than has been employed in the two prior UK studies; the

long sample period also helps identify the evolution of the recommendation performance in different market condi-

tions. Specifically, we construct an upgrade portfolio, including all upgrades to buy-related stock recommendations

from previous sell/hold-related stock recommendations, as well as a downgrade portfolio, including all downgrades to

sell/hold-related stock recommendations from previous buy-related stock recommendations. The up/downgradeport-

folio is updated daily; for each up/downward revision, the recommended stock enters the up/downgrade portfolio at

the close of trading on the day the revision is announced and then remains in the portfolio for up to five trading days

(oneweek). Like Barber, Lehavy,McNichols, and Trueman (2001), we take an investor-oriented,calendar-time perspec-

tive to track the evolution of the portfolio performance on a rolling window basis.2This allows us to directly measure

the time-varying performance of the up/downgrade portfolios using the intercepts derived from various asset pricing

models and to estimate portfolio turnover.Consequently, we are able to determinate whether the up/downgrade port-

folio can generate statistically significant net abnormal returns over time, after taking transaction costs into account

(see, also, Barber et al., 2001). Accordingly, our empirical investigations proceed in three major parts using various

alternative regression models.