Abnormal Returns from Takeover Prediction Modelling: Challenges and Suggested Investment Strategies

• Author: Antonios Siganos,Jo Danbolt,Abongeh Tunyi
• DOI: http://doi.org/10.1111/jbfa.12179
• Date: 01 January 2016
• Published date: 01 January 2016

Journal of Business Finance & Accounting

Journal of Business Finance & Accounting, 43(1) & (2), 66–97, January/February 2016, 0306-686X

doi: 10.1111/jbfa.12179

Abnormal Returns from Takeover

Prediction Modelling: Challenges and

Suggested Investment Strategies

JODANBOLT,ANTONIOS SIGANOS AND ABONGEH TUNYI∗

Abstract: While takeover targets earn significant abnormal returns, studies tend to find no

abnormal returns from investing in predicted takeover targets. In this study, we show that the

difficulty of correctly identifying targets ex ante does not fully explain the below-expected returns

to target portfolios. Target prediction models’ inability to optimally time impending takeovers,

by taking account of pre-bid target underperformance and the anticipation of potential targets

by other market participants, diminishes but does not eliminate the potential profitability of

investing in predicted targets. Importantly, we find that target portfolios are predisposed to

underperform, as targets and distressed firms share common firm characteristics, resulting in

the misclassification of a disproportionately high number of distressed firms as potential targets.

We show that this problem can be mitigated, and significant risk-adjusted returns can be earned,

by screening firms in target portfolios for size, leverage and liquidity.

Keywords: takeover prediction, abnormal returns, portfolio strategies, investment timing, firm

size, rumours

1. INTRODUCTION

A large number of studies show that takeover targets experience significant stock price

increases around merger announcements. Jensen and Ruback (1983), one of the first

systematic reviews of the mergers & acquisitions (M&A) literature, reports weighted

average abnormal returns of 29.1% for US targets in the month or two surrounding

an offer. A comparative study by Franks and Harris (1989) based on UK data reports

abnormal returns of a similar magnitude. More recent studies employing US, UK and

∗The first author is from the University of Edinburgh. The second author is from the University of Glasgow.

The third author is from Liverpool Hope University. The authors are grateful to an anonymous referee,

Ronan Powell (associate editor), Bill Rees, Jos´

eYag

¨

ue and participants at the BAFA (Manchester) 2015

conference, the BAFA Scottish (Edinburgh) 2015 conference and the XI Workshop on Empirical Research

in Financial Accounting (Cordoba) 2015 for helpful comments on previous versions of this paper. Jo

Danbolt holds the Baillie Gifford Chair in Financial Markets, and his research is partially funded by a Baillie

Gifford endowment held by the University of Edinburgh Business School. Baillie Gifford has no role in or

influence over the research conducted. (Paper received December 2014, revised version accepted January

2016)

Address for correspondence: Jo Danbolt, The University of Edinburgh Business School, 29 Buccleuch Place,

Edinburgh EH8 9JS, UK.

e-mail: Jo.Danbolt@ed.ac.uk

C

2016 John Wiley & Sons Ltd 66

ABNORMAL RETURNS FROM TAKEOVER PREDICTION MODELLING 67

EU samples (e.g., Goergen and Renneboog, 2004; Danbolt, 2004; and Gregory and

O’Donohoe, 2014) show comparable abnormal returns to targets, ranging from 19.5

to 31% in the days and months surrounding a bid.

Unsurprisingly,a number of studies (e.g., Palepu, 1986; Powell, 2001; and Brar et al.,

2009) explore whether a successful investment strategy can be developed by predicting

potential targets. However, these studies report limited success despite finding that

prediction models are fairly successful in identifying future targets. Powell (2001), for

example, finds that the market- and size-adjusted returns to his UK target portfolio

between 1 January 1996 and 31 December 1996 were –11.0% and –4.0%, respectively.

Mirroring the conclusions of earlier studies, such as Palepu (1986) and Barnes (1999),

Powell (2001, p. 1008) contends that ‘developing statistical models to predict takeover

targets is unlikely to result in a profitable investment strategy’. However, a small

number of recent studies have reported that the generation of abnormal stock returns

from takeover prediction is feasible. Using a small sample of Australian listed firms,

Rodrigues and Stevenson (2013), for example, find that their model performs in line

with the market in 2009, but outperforms the market in 2010 and 2011. Their results

are, however, only adjusted for risk using a simple market-adjusted returns model. To

our knowledge, the most optimistic results have been reported by Brar et al. (2009).

Their study, based on an EU sample, shows that a ‘takeover timing portfolio’ generates

a modest alpha of 0.58% per month between 1995 and 2003 (Brar et al., 2009, p. 448).

As discussed below, one potential reason for the gains reported by Brar et al. (2009)

is the size restriction they impose by focusing on deals involving targets with market

capitalisation in excess of $100 million.

Our paper is motivated by the finding that target prediction models have some

predictive ability (at least, better than random classification); targets gain substantially

from takeover activity, yet portfolios of predicted targets fail to outperform the market.

We contribute to the literature in two broad areas. First, we explore the underlying

factors that influence the stock returns of predicted target portfolios. Amongst these

are issues of poor timing, prediction errors and the tendency for distressed firms

to be identified as potential targets. Second, we investigate potential strategies for

mitigating the effects of these factors. In particular, we show that by screening firms

in target portfolios for size, leverage and liquidity, the effect of distressed firms can

be reduced and significant abnormal returns can be generated from the investment

strategy.

With respect to the factors influencing target portfolio stock returns, we highlight

three potential reasons why portfolios generated from takeover prediction models fail

to generate abnormal returns, and we investigate these by developing nine testable

hypotheses. First, we examine the predictive ability of current prediction models and

whether the prediction models’ inability to correctly predict a substantial number of

future targets explains the low returns to target portfolios (as suggested by Palepu,

1986; Barnes, 1999; Powell, 2001, 2004; and Cahan et al., 2011). Prediction models

underperform if the predicted targets (or firms with high takeover likelihood) do

not eventually receive takeover bids. Prior studies suggest that these firms (described

as type II errors) are strategically better off if acquired. Given that takeover targets

generally underperform prior to takeovers (Palepu, 1986; and Powell, 2001), the

expectation is that such firms are likely to continue to underperform unless they

become a takeover target. The presence of type II errors in the portfolio of predicted

targets will, perhaps, explain a substantial portion of the low returns to these portfolios.

C

2016 John Wiley & Sons Ltd

68 DANBOLT, SIGANOS AND TUNYI

We hypothesise that (i) type II errors underperform the typical non-target (hypothesis

1); and (ii) target portfolios stripped of type II errors earn significant abnormal returns

(hypothesis 2). If the presence of type II errors explains the low returns to target

portfolios, then the literature will benefit from the development of better prediction

models.

Second, we explore the challenge of optimal timing in target prediction, and

whether poor timing in prediction potentially explains the low returns to target

portfolios. The challenge arises from the finding that targets tend to perform poorly

over several months prior to the bid announcement, but start to generate significant

returns as the announcement date draws closer (we confirm this later in our study).

This suggests that holding targets later rather than sooner might be of some benefit.

Nonetheless, other market participants are likely to anticipate potential takeover bids

as the bid date draws near, thus reducing any abnormal returns to be generated from

the strategy. To our knowledge, our study is the first to consider the significance of

timing when predicting takeover targets.

To elaborate, optimal timing is hampered, as target prediction models employ

firm fundamental values; therefore, annual portfolio rebalancing is used in takeover

prediction modelling (see, e.g., Palepu, 1986; Barnes, 1999; Powell, 2001; Powell and

Yawson, 2007; Brar et al., 2009; and Cremers et al., 2009). However, the predictions

may be made so early that the stock returns of targets experience a significant decrease

before the pre-bid upward movement in stock price commences. We might, therefore,

find very low returns to correctly predicted targets. Further, takeover prediction

models exist in the public domain, and other market participants are likely to employ

similar strategies. We should, therefore, expect to find evidence of market anticipation

of predicted targets. We hypothesise that the returns to actual targets in target

portfolios are low, as several potential targets are included in the portfolio either too

early or too late (hypothesis 3). To directly test the effect of anticipation by other

market participants using similar models, we also hypothesise that takeover targets

which are correctly predicted by our model are less of a surprise to the market, and,

hence, earn significantly lower abnormal returns than targets which we are unable to

predict (hypothesis 4).

Third, we explore whether target characteristics that match those of firms that

are predisposed to experience financial distress explain the difficulty of generating

abnormal stock returns from target prediction models. Prior research (e.g., Pastena

and Ruland, 1986; Clark and Ofek, 1994; and Powell and Yawson, 2007) suggests that

targets and bankrupt firms share similar characteristics, and, therefore, mergers and

bankruptcy are alternative forms of reorganisation facing firms in distress. Targets

and distressed (bankrupt) firms tend to be small firms with poor stock performance

(Powell and Yawson, 2007). Takeoverprediction models are therefore likely to select as

potential targets firms that are simultaneously classified as candidates for bankruptcy.

Prior takeover prediction studies (e.g., Palepu, 1986; Powell, 2001; and Brar et al.,

2009) ignore the impact of this tendency and hence overstate portfolio returns by

not recognising the significant loss occurring when predicted targets exit the portfolio

through bankruptcy.

To our knowledge, our study is the first to consider the impact of bankruptcies

on the profitability of investing in predicted takeover targets and to propose a

strategy for mitigating its effect. We can directly address the issue of investing in

firms that eventually go bankrupt by explicitly trying to identify these firms ex ante

C

2016 John Wiley & Sons Ltd