What’s Really in a Deal? Evidence from Textual Analysis Wenyao Hu Lally School of Management Rensselaer Polytechnic Institute [email protected]Thomas Shohfi* Lally School of Management Rensselaer Polytechnic Institute [email protected]Runzu Wang Michael F. Price College of Business The University of Oklahoma [email protected]January 2019 * Contact author. We thank the Donald Shohfi Financial Research Fund for support.
41
Embed
What’s Really in a Deal? Evidence from Textual Analysis · We perform textual analysis on event transcripts to determine whether sentiment inside M&A conference calls influences
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Over the past decades, voluntary disclosure has become an important topic of academic
research in both finance and accounting. Conference calls are a common medium for evaluating
voluntary disclosure and have been examined in various aspects: the factors influencing a firm's
decision to host a conference call (Tasker, 1998), the impact of conference calls on trading
(Frankel et al., 1999), and the effect of conference calls on analysts' forecasts (Bowen et al.,
2002). However, most of these previous studies focus on earnings conference calls which are
usually held at the end of each quarter. In this study, we focus on a relatively new type of
conference calls—mergers and acquisition (M&A) conference calls that focus on specific
transactions. Since much of the variation in acquirer returns remains explained (Golubov et al.,
2015), we seek to answer the question: what information is inside an M&A call and how do
investors react to this information?
M&A conference calls are usually held in conjunction with or after the M&A announcement
as a means of voluntary disclosure. To the best of our knowledge, there are only three published
papers1 that examine M&A conference calls (Kimbrough and Louis, 2011; Siougle, Spyrou and
Tsekrekos, 2014; and Fraunhoffer, Kim and Schiereck, 2018). Using a sample of 1,228 M&A
deals in the US from 2002 to 2006, Kimbrough and Louis (2011) document that conference calls
are held when transactions are large and financed by stock. They also find the average market
reaction to a merger announcement is significantly more favorable when the deal is accompanied
by an M&A conference call. Using a sample of UK firms, Siougle, Spyrou, and Tsekerekos
(2014) find the existence of equity options decreases the willingness of managers to hold M&A
1 There is one more working paper by Rohrer (2017) that examine the participation of institutional investor in M&A
calls.
2
calls as equity option markets already convey part of the information released during the call.
The most recent paper by Fraunhoffer, Kim and Schiereck (2018) investigates M&A calls in
Europe and shows that the premium return for M&A calls is only present in the UK and France
and in industries with a focus on research and development. However, none of these prior studies
examines the content (i.e. participants, tone, etc.) of M&A conference calls.
We collect M&A conference call data from 2011 to 2017 through Capital IQ and analyze
transcripts to extract information such as executive participants, sentiment, and quantitative
information of each call. After matching with the SDC Platinum M&A database, we identify 828
unique merger calls associated with 836 merger announcements. We show that percent of stock
payment increases the likelihood of both target CEO and target CFO participation in the M&A
call. This result is consistent with a tradeoff between realized and unrealized benefits for
managers of the target. We also estimate the market reaction to target management participation
and find a negative market reaction to the participation of target executives. We attribute this
negative market reaction to the retention of target executive.
We perform textual analysis on event transcripts to determine whether sentiment inside
M&A conference calls influences market reaction to a deal. We find that positive net tone has
both a statistical and economically significant negative influence on acquirer returns. We explore
two potential explanations for his negative relationship: managerial overconfidence and
information asymmetry between managers and investors. We employ three tests by dividing the
sample according to industry concentration, the listing status of the target, and information
environment, and find support for both of these explanations. Further, we show that content
within M&A calls significantly differs from both contemporaneous press releases and earnings
conference calls.
3
Compared to private targets, a public target generally has much more available information
since public traded companies have mandatory disclosures required by listing exchanges and
regulatory authorities. Ceteris paribus, investors face less information asymmetry with regards to
how a public target will influence an acquirer’s valuation and can distinguish between genuinely
positive and contrived sentiment within management discussion. We find that benefits from
private targets mitigate negative effects by around 90% which is significant at 10% level.
Increased market power is another important motive for acquisitions (Borenstein, 1990).
Overly optimistic manager sentiment about an acquisition may be offset when investors
anticipate gains from increased pricing power of the firm. Ali, Klasa, & Yeung (2014) shows that
firms in more concentrated industries tend to have a more opaque information environment. We
use top quartile Herfindahl-Hirschman Index (HHI) to measure industry concentration in a given
year. Consistent with these explanations, we find that a one standard deviation increase in net
positive tone from concentrated (diffuse) industries is associated with a 15% (-42%) relative
increase in CAR from the mean.
Moreover, we also use information intensity (Zhao, 2017) to measure the information
environment for a firm. We find support for the information asymmetry explanation as a one
standard deviation increase in net positive tone for a low information intensity firm will increase
market reaction by 5% relative to the mean compared to a 62% decrease for high information
intensity firm.
We also extract other pertinent information including the level of quantitative information
within each M&A call. Prior research demonstrates that the impact of quantitative versus
qualitative information on the market is different (Hutton, Miller, and Skinner, 2003). In recent
work, Zhou (2017) proposes that quantitative information is more precise and has a more
4
positive market reaction than qualitative information around quarterly earnings conference calls.
We predict and find that acquiring firm shareholders react more positively to M&A conference
calls with higher levels of quantitative information. Specifically, we find that a one standard
deviation increase in the percent of numbers spoken increase the market reaction by nearly 1.1%
(significant at 5% level) representing a $121 million increase relative to mean firm market
capitalization in our sample.
Previous research (Rodrigues and Stegemoller 2014; Gorbenko and Malenko, 2014)
employing novel data of M&A characteristics is often restricted by the limited sample size of
hand collected data. By using textual analysis, we build dictionaries to represent the
characteristics of each deal. Specifically, we identify relevant keywords in M&A calls and
broadly separate them into two groups: financial and strategic words. We find that the market
reacts more positively to more frequent occurrence of financial words and negatively to strategic
words. A one standard deviation increase in financial (strategic) words increases the market
value for mean size company in our sample by approximately $64 million (-$58 million). This
result is consistent with previous qualitative research identifying the difference between strategic
and financial mergers.
Our study makes several contributions to the literature. First, our findings provide insight
into variation within acquirer returns. Many prior studies examine M&A announcement returns
(Jensen and Ruback, 1983; Bradley et al.,1988; Moeller et al., 2005; Masulis et al., 2007; Phan,
2014), however, there is still a large proportion of unexplained variation in acquirer
announcement stock returns. Our composition analysis of M&A conference calls provides a
better understanding the gains/losses to acquirers.
5
Second, our study contributes to the existing literature on voluntary disclosure by analyzing
the contents of management communication regarding a specific, often unexpected, corporate
activity. Other than previous research which mainly focuses on earnings conference calls, our
research highlights the usefulness of voluntary disclosure for an unscheduled corporate event
particularly when management overconfidence and/or information asymmetry about the event is
high (i.e. acquisitions of private targets).
Third, our research advances existing research on M&A calls by adding detailed content
analysis. Previous research (Kimbrough and Louis, 2011; Siougle, Spyrou, and Tsekerekos,
2014; Rohrer, 2017; Fraunhoffer, Kim and Schiereck, 2018) on this topic mainly focuses on the
determinants or outcomes of hosting a call. Our paper is the first, to the best of our knowledge, to
analyze the content of M&A calls and examine its impact on acquirer returns.
This paper is organized as follows. In section 2, we develop our hypotheses related to
information inside M&A calls. Section 3 describes the procedure of obtaining our M&A sample.
Empirical results are described in section 4 while section 5 concludes.
2 Hypothesis Development
2.1 Attendance of Top Executives
The participation of top executives in conference calls is probably one of the most
fundamental pieces of information contained in conference calls. In this paper, we try to discover
the driving factors of top acquirer and target executives attending a conference call about a
proposed transaction. Previous literature shows that in a stock-based merger, the target
executives face a tradeoff between the tax benefit of stock sales and the risk minimizing benefit
of cash consideration (Faccio and Masulis, 2005). When they choose to accept a stock offer,
6
target managers have greater incentives to attend an M&A since their future value is linked to the
ongoing performance of the acquiring company. Target managers are also more likely to retain
positions within the acquiring firm when the deal has a higher stock component (Ghosh and
Ruland, 1998), and are thus more likely to participate in the conference call. We summarize our
first hypothesis accordingly.
Hypothesis 1: The attendance of top executives of the target is positively affected by the
percentage of stock in the M&A offer.
Moreover, Fich et al. (2016) point out that acquirers retaining the target CEO in M&A
transactions receive a significantly lower return in both short- and long-time periods. The
unexpected conclusion of Bargeron et al. (2017) partly supports this result by providing evidence
that the target shareholders gain higher acquisition premiums if the target CEO is retained after
the transaction. Therefore, if the main driving mechanisms we describe above are true, we would
also expect the market reacts negatively to target executives’ attendances.
Hypothesis 2: The attendance of top executives of the target has a negative impact on the
CAR of acquirer.
2.2 Net Tone and Market Reaction
The net tone of language used in information disclosures, i.e., the net usage of positive
words versus negative words, is probably the most common metrics used in previous literature
employing textual analysis. Early studies have examined the impact of net tone in different kinds
of information disclosures, including press releases (Davis et al. 2012), earnings conference calls
(Price et al. 2012), and MD&A disclosures (Davis and Tama-Sweet 2012). In this paper, we
investigate on how net tone of M&A conference calls affects their abnormal returns.
7
Constant with previous literature, one would expect the marketplace to react positively to
positive tone of information disclosure events. However, in comparison to other scheduled
disclosure events like earnings conference calls, the level of information asymmetry is much
higher between the management and investor/analyst participants of an M&A conference call.
Moreover, managers of acquiring firms express overconfidence regarding to successful outcomes
of merger and acquisition transactions (Malmendier and Tate, 2007). As a result, words spoken
by an acquiring management team are expected to be overly positive which often means that the
sentiment they use to describe a deal is "too good to be true." Therefore, a potential question is
how much the market believes such information and whether the market can distinguish whether
managers are over-optimistic or not. While an earnings conference call is preceded an earnings
surprise that drives sentiment, there is no analogous metric for M&A events. Therefore, we
propose two opposite hypotheses, and let the data lead us to the correct one.
Hypothesis 3a: The net tone of an M&A conference call has a positive impact on the
market reaction.
Hypothesis 3b: The net tone of an M&A conference call has a negative impact on the
market reaction.
2.3 Information in Numbers
Early literature shows that disclosure of both qualitative and quantitative information is
valuable to investors. However, more recent studies point out that the impact of these two types
of information on the market is different. For example, Hutton, Miller, and Skinner (2003) find
that when there are no numbers in the headline of a press disclosure, the market tends to
underreact to the disclosure. Based on this fact, the authors argue that numbers in disclosure
serve as salient information.
8
Since quantitative information has a stronger impact on the market, Chuprinin, Gaspar, and
Massa (2017) and Zhou (2017) point out that the proportion of quantitative information and
qualitative information also carries information. A higher percentage of quantitative information,
defined as tangible information in Chuprinin et al. (2017), is related to more precise and easier
interpretation.
It is therefore possible that information tangibility with M&A conference calls plays an
important role in explaining acquirer equity market reaction. In this paper, we follow the
approach used by Zhou (2017), which uses the percentage of numbers in a conference call as a
measure of information tangibility, and check its impact on market reaction to M&A conference
calls. Since higher tangibility of information implies higher quality disclosure, we predict that
the market reacts more positively to M&A conference calls with a higher percentage of numbers.
Hypothesis 4: The market reacts more positively to M&A conference calls with a higher
percentage of numbers.
2.4 Recognition of M&A Transaction Type via Conference Calls
Market reactions to M&A transactions have been studied extensively in the literature
(Travlos (1987), Moeller, Schlingemann and Stulz (2004), Masulis, Wang and Xie (2007),
Rodrigues and Stegemoller (2014)). Many of these prior works focus on how different firma nd
deal characteristics affect the market reaction. These results are often clear and strong. However,
in many cases, the size of the sample of deals non-standard characteristics is small due to the cost
associated with hand collected data. However, with M&A conference call transcripts and textual
analysis techniques, we might able to identify such characteristics and use them to better explain
the market reactions to acquisitions.
9
It is commonly accepted that bidders in M&A transactions can be divided into two groups:
financial bidders and strategic bidders. Financial bidders seek undervalued firms and treat the
target firm as a part of their portfolio, while strategic bidders focus on long term operational
synergies and try to integrate the target into their own business (Gorbenko and Malenko, 2014).
Using a small sample of special purpose acquisition corporations, Rodrigues and Stegemoller
(2014) conclude that, compared to strategic M&A transactions, financial M&A transactions have
significantly higher announcement abnormal returns. We create two custom M&A motive
dictionaries using keywords (provided in Appendix III) of previous studies in financial and
strategic M&A transactions and use the proportion of words in these two dictionaries as
measures of how likely a deal is to being financial or strategically motivated. According to prior
literature, we expect that the market reacts more positively to higher percentages of financial
words and lower to percentages of strategic words.
Hypothesis 5: The market reacts more positively to more frequent use of financial words
and less frequent use of strategic words during M&A calls.
3 Data
3.1 M&A Conference Calls
We start the data collection process by extracting a list of all announced mergers and
acquisitions between January 2011 and December 2017 from SDC Platinum Database. 2011 is
the year in which merger call transcripts are widely available through Capital IQ.2 In screening
the deal in SDC database, we set several restrictions. We require that the deal value is at least 1%
of the acquirer's market value, that the acquirer is public, the acquirer can be linked to CRSP and
2 The first year in which M&A calls appears is 2009, but the coverage of M&A call transcripts in Capital IQ is low
for the first two years.
10
COMPUSTAT, and that the deal status is completed. Applying these screening filters, we
identify 2,538 unique merger announcements.
For each of the 2,538 deal announcements, we manually collect M&A conference call
transcripts from Capital IQ. When matching the conference call with deal information in the
SDC database, we require that the date of the M&A conference call be within 7 days following
deal announcement to increase the accuracy of the matching process. After careful matching, we
identify 828 unique merger calls associated with 836 merger announcements.3 Among the
identified calls, we obtain 757 unique M&A conference call transcripts in Capital IQ.4 Merger-
related conference calls occur for about 34% of all the deals in our sample. Compared to
Kimbrough and Louis's (2011) sample, our coverage rate is higher since we use a 1% deal ratio
which is smaller than theirs (10% deal ratio). We report descriptive statistics for the full sample
in Table 1.
[Insert Table 1]
From Table 1, we find distinguishing differences between deals with and without M&A
conference calls. Consistent with prior research, we show that acquirers who choose to hold
conference calls are much larger, have a lower book to market ratio, and have higher analyst
coverage than firms that do not hold M&A conference calls. For deal level characteristics, we
find that firms with conference calls have higher payment in stock, a larger deal ratio, and the
3 One M&A call can involve discussion about several different deals. 4 Some of the transcripts are not available in Capital IQ since the company either did not open the call to the general
public or did not provide an event transcript to Capital IQ.
11
target is less likely to be private. At the industry level, bidders are more likely to hold a call
when the target is in the same industry or the acquirer is in the finance industry.5
3.2 Information in M&A Calls
After obtaining individual merger call transcripts from Capital IQ, we parse and structure
the obtained HTML file for each conference call. We identify each participant's name and the
paragraph associated with that person in each part (presentation or Q&A) of the transcript using
Python’s Beautifulsoup package. We also identify the following information for each merger
call: the affiliation of each person (e.g. acquirer/target executive or analyst), tone, and numerical
content.
3.2.1 Participation of Executives
For identification of executive participant affiliation, we first use the summary page in each
M&A call to broadly separate participants into two groups: executives or analysts. For
executives, we first identify position by extracting keywords such as CEO (Chief Executive
Officer) and CFO (Chief Financial Officer) from each participant’s title in the executive
subgroup. If there is no information in the call summary page, we manually check each person’s
position through the content in the M&A call or through internet searches. We then use
Execucomp to check whether the executive is from the acquirer or target.7 In Table 2, we list the
top executive participant types in M&A calls.
[Insert Table 2]
5 Probit regression results for deal announcements that are followed by M&A conference calls are provided in
Appendix II. 7 For private targets, we assume all remaining unmatched executives are from the target firm.
12
Table 2 tells us that the acquirer CEO is the most frequent participant n M&A calls. Since
mergers and acquisitions usually lead to a major change in company's future direction, acquirer
CEOs need to participate and explain to analysts and investors how the deal will positively
impact the future of the firm. Following the acquirer CEO, the acquirer CFO attends near 85% of
calls in our sample as the acquisition usually contains a great amount of payment and financial
restructuring details. On the target side participation is much lower: the target CEO (CFO)
participates in nearly 30% (5%) of M&A calls.
3.2.2 Content Inside M&A calls
After identifying the participants in M&A calls, we use Loughran and McDonald’s
dictionaries (2011) to separately measure positive and negative tone as the number of positive
(negative) words divided by the total number of words in merger calls. We use net positive tone
which is positive tone minus negative tone for our regression analysis. For quantitative
information inside the merger calls, we follow the same procedures as described by Zhou (2017).
We look for any number (excluding numbers more likely to be years) that begins with a space or
dollar sign ($) and followed by numbers (0-9), a comma (,), or a period(.). We calculate the
percentage of numbers as the total count of numbers divided by the total number of words in
each section of the transcript. Table 3 reports summary statistics for the information inside M&A
conference calls.
[Insert Table 3]
From Table 3, we find that the average number of words in an M&A transcript is around
5500. There are approximately 38% (62%) of words in the presentation (Q&A) section of a
typical merger call. Panel B shows that net positive tone is around 1.14% which means the
13
average of the tone for a transcript is positive and there are around seven questions in a merger
call.
To compare the tone in M&A calls, we construct two related measurements from the
merger-related press release and earnings conference calls. We define the merger-related press
release as the nearest 8-K filling around the M&A announcement date with merger relevance
keywords inside the document. Moreover, we follow the same data collection procedure as Call,
Sharp, and Shohfi (2017) to obtiain the corresponding earnings calls for the firms in M&A
sample during the sample period.
[Insert Table 4]
Table 4 gives us a clear view that sentiment inside M&A calls is much higher than the other
two types of disclosure. Compared to an M&A press release, an M&A call expresses more
positive tone and less negative tone. The results is consistent with a press release in an 8-K being
a mandatory disclosure which contains higher legal risk and is therefore more conservative. We
also collect all related earnings conference calls from acquiring firms during the sample period.
As we can see in the third column of Table 4, the net positive sentiment is still significantly
higher for M&A calls. This is initial support for our argument that managers express far more
optimism during M&A calls compared to other forms of corporate disclosure.
3.3 Endogenous Choice of Hosting an M&A Call
There is a potential endogeneity concern regarding holding an M&A conference call in that
the decision to do so is made by the acquiring firm. This problem has been raised by previous
research (Kimbrough and Louis, 2011; Rohrer 2017). In our empirical approach, we address this
issue using the approach of Kimbrough and Louis (2011). We run a probit regression model to
14
determine the probability of hosting an M&A conference call in our sample. We then calculate
the inverse mills ratio from the selection model. The result is shown in Appendix II. In
subsequent empirical analyses, we do not include industry membership in our regressions as it is
highly associated with the conference call decision but has no theorized effect on announcement
returns (Kimbrough and Louis, 2011).
4 Empirical Results
4.1 Determinants of Executive Appearance
We hypothesize that the percent of stock payment will positively affect the appearance of
target executives. Target executives are sellers who are faced with a tradeoff between the tax
benefits of stock and risk minimizing benefits of cash payment (Faccio and Masulis, 2005). It
follows that target executives will have greater incentive to attend the M&A conference call with
higher levels of stock payment.
To test our hypothesis related to the appearance of top executives, we use the following