AARHUS UNIVERSITY, SCHHOL OF BUSINESS & SOCIAL SCIENCES
MASTER THESIS
How to design effective earnouts in M&A deals?
Empirical evidence on an option-pricing model on earnout design
November 1, 2015
Author
Johann Arne Spanuth
201302396
MSc in Finance & International Business
Academic Supervisor
Stefan Hirth
Associate Professor
Department of Economics and Business Economics
Number of characters (excl. spaces): 149,379
Abstract
Literature on Mergers & Acquisitions (M&As) shows growing interest in earnouts as a
payment form that mitigates valuation risks from information asymmetry. Theory argues
that earnouts present a solution to problems of adverse selection by serving as a signalling
tool for high quality targets and to agency problems by serving as an incentive tool
towards the target’s management post-closing. (E.g. Kohers & Ang, 2000; Ragozzino &
Reuer, 2009) Indeed, research finds strong evidence that earnouts are more likely used in
deals facing high information asymmetry and report positive market reactions for the
acquiring firm. (E.g. Datar, Frankel & Wolfson, 2001; Cain, Denis & Denis, 2011) To
the contrary, little is known about how to design earnouts to ensure their effectiveness.
The thesis contributes to the limited research on earnout design by investigating both
theoretically and empirically what factors determine the design of earnouts in M&A deals.
First, the thesis develops a game-theoretic option pricing model on earnout design. The
effectiveness of earnouts is expected to depend on the likelihood that the target will
receive an earnout premium. A higher likelihood is associated with less efforts by the
target and with even low-quality firms tending to accept the earnout. The model suggests
two ways for the acquirer to react to an increased likelihood. On the one hand by
increasing the contingent part of the acquisition price, i.e. the earnout ratio, in order to
stronger incentivize the target for post-closing cooperation and to motivate only high
quality targets to accept the deal and on the other hand by shaping the earnout parameters
such that the likelihood is controlled for. Based on the option-like characteristics of
earnouts, the model argues that a higher uncertainty about the target’s future performance,
longer earnout periods and a lower performance goal all increase the likelihood of an
earnout pay-out in the end. Given a degree of uncertainty, the acquirer is therefore
expected to adjust the remaining earnout parameters accordingly.
The empirical analysis by means of regression models based on a UK sample of 377
earnout deals finds only weak evidence for the hypothesized dynamics. However,
previous research reports empirical results in favour of the theoretical model. These
studies document higher earnout ratios and shorter earnout periods in case of high
uncertainty. Consequently, there is strong evidence that the earnout ratio is used as the
primary “control lever” if the acquirer faces the need to design strong signalling and
incentive tools and also some evidence that the earnout period is decreased to control for
the likelihood of an earnout payment. Due to data limitations the determinants of the
performance goal could not be empirically examined.
After all, the thesis to a large extent fails to find evidence for the theoretical model on
earnout design, but the available literature shows some, yet limited, support. Further
research on the option-based approach that overcomes the thesis’s data limitations is
strongly encouraged to pave the way towards the definition of optimal earnout design.
Keywords: Earnout Design; Determinants; Uncertainty; Information Asymmetry; Option
Pricing Methodology
i
Table of contents
List of tables iii
List of figures iii
1 Introduction 1
1.1 Problem statement 2
1.2 Approach and structure 3
1.3 Scope and delimitation 3
1.4 Evaluation of sources 4
2 Literature review on earnouts in Mergers & Acquisitions 5
2.1 Motives to use earnouts 5
2.1.1 Theoretical hypotheses 5
2.1.1.1 The information asymmetry hypothesis 5
2.1.1.2 The uncertainty hypothesis 8
2.1.2 Insights from empirical studies 9
2.1.2.1 Evidence for the information asymmetry hypothesis 9
2.1.2.2 Evidence for the uncertainty hypothesis 15
2.1.3 Summary 17
2.2 Design of earnouts 18
2.2.1 The earnout parameters 18
2.2.2 Current state of research 19
3 Theoretical model on the design of earnouts in Mergers & Acquisitions 22
3.1 The option-like characteristics of earnouts 23
3.1.1 Earnout premium payment profiles 25
3.1.2 Mapping the earnout parameters onto a financial call option 29
3.2 A game-theoretic option pricing model on design of earnouts 31
3.2.1 Original model by Lukas, Reuer & Welling (2012) 31
3.2.2 Own advancements to the model 34
4 Hypotheses 37
5 Data sample creation 38
5.1 Deal search 38
5.2 Collecting data on earnout parameters 39
5.3 Generating the explanatory variable “uncertainty” 40
5.4 Final data sample 42
ii
6 Methodology 43
6.1 Control variables for information asymmetry 43
6.2 Regression models on earnout parameters 44
6.2.1 Tobit regression model of the earnout ratio 45
6.2.2 OLS regression model of the earnout period 46
6.2.3 Binary choice model of the performance measure 46
7 Empirical results 48
7.1 Descriptive statistics 48
7.2 Results from the regression models 51
7.2.1 Determinants of the earnout ratio 52
7.2.2 Determinants of the earnout period 55
7.2.3 Determinants of the performance measure 57
8 Evaluation and avenues for further research 62
8.1 Discussion of the empirical results 62
8.2 Limitations to the study 64
8.3 Further research 66
9 Conclusion 67
References 70
Appendices 74
Appendix 1: Earnout premium payment profiles 74
Appendix 2: Data selection process 75
Appendix 3: Methodology 80
iii
List of tables
Table 1 Main findings of empirical studies on the choice to use earnouts 11
Table 2 Main findings of empirical studies on wealth effects of earnouts 15
Table 3 Definitions of control variables for information asymmetry 44
Table 4 Descriptive statistics on data sample 49
Table 5 Results from Tobit model on determinants of earnout ratio 53
Table 6 Results from OLS model on determinants of earnout period 56
Table 7 Results from logit model on determinants of sales measure 58
Table 8 Results from logit model on determinants of income measure 59
Table 9 Results from logit model on determinants of non-financial measure 61
List of figures
Figure 1 Payment profile and implications of earnout like a call option 26
Figure 2 Payment profile and implications of earnout like a binary option 27
Figure 3 Payment profile and implications of earnout like a call option with cap 28
Figure 4 Mapping an earnout onto a financial call option 30
Figure 5 Dynamics of model on earnout design by Lukas, Reuer & Welling
(2012)
33
Figure 6 Dynamics of advanced model on earnout design 36
1
1 Introduction
“In an uncertain economic climate, there may be more willingness to take the wait-and-
see option offered by an earnout-structured acquisition.”
Craig & Smith (2003, p. 46)
There is an extensive body of literature examining the question of whether Mergers &
Acquisitions (M&As) are value creating or value destroying events for the acquirer’s and
the target’s shareholders. By now, research agrees that on average M&A deals
significantly increase target’s shareholder value while the acquirer’s shareholders
experience a value decreasing effect. (For a comprehensive review of studies see Eckbo,
2009) Around these basic results, literature especially focuses on the firm and deal
specific characteristics that drive the negative effect for the bidders. Studies report that
next to the size of the bidder (e.g. Asquith, Bruner & Mullins, 1983; Moeller,
Schlingemann & Stulz, 2004), and the target’s status as being public or private (e.g.
Bargeron et al., 2008), also the method of payment determines the bidder’s value gains
from a transaction (e.g. Chang, 1998, Fuller, Netter & Stegemoller, 2002). Most of the
studies concerning the method of payment in M&A deals surround the value enhancing
effects of cash deals as compared to stock deals.
However, recently a new stream of research developed that examines not only the
payment currency (cash versus stock) but focuses on the form how the takeover price is
settled by the acquirer. More specifically, research shows an increasing interest in the
form of contingent payments such as the earnout. Earnouts define a part of the overall
acquisition price to be contingent on the target’s ability to meet a prespecified
performance goal within a certain time frame post-closing of the deal. (Reuer, Shenkar &
Ragozzino, 2004, p. 20) Research argues that this type of payment form secures the buyer
against adverse selection and agency problems and thereby mitigates valuation risks.
Indeed, studies find strong evidence for a positive impact from the choice to use earnouts
on acquirer’s shareholders value gains. (E.g. Kohers & Ang, 2000; Mantecon, 2009)
Knowing that bidding companies’ shareholders on average lose value from M&A events,
these findings indicate that the research on earnouts is a valuable supplement to existing
M&A literature. Moreover, as Lukas, Reuer & Welling (2012, p. 257) conclude the
research on earnouts, in contrast to other areas of the M&A literature, is still in its infancy.
2
Most of these studies have focused on the motives of using earnouts in M&A deals and
reached consensus on the deal and firm characteristics that best suits the choice to use an
earnout. (E.g. Barbopoulos & Sudarsanam, 2012) Yet only a few studies are concerned
with the design of these contracts and there is still a lack of theory and empirical insights
on the determinants of appropriate earnout design. Consequently, the question of how to
shape the parameters to make earnouts a valuable tool has yet to be answered.
Therefore, this revenue stream presents a unique and exciting opportunity for a master
thesis to contribute to a promising new research avenue. As an additional personal
motivation, the author of this thesis was concerned with the use of earnouts in an M&A
context in practical life. Throughout the process of writing this thesis, the issue of how to
specifically design an earnout according to the deal’s circumstances therefore also had
practical value for the author.
1.1 Problem statement
This thesis aims to contribute to the yet limited research on earnout design. While
earnouts are complex contractual arrangements tailored to the requirements of each
particular deal, they all comprise common parameters that together constitute the earnout
mechanism. (Cain, Denis & Denis, 2011, p. 152) The contract defines the contingent part
of the overall acquisition price (the earnout ratio), the time frame during which the target’s
performance is monitored (the earnout period), the performance measure and the exact
performance goal that the target has to reach in order to receive the earnout payment. The
thesis therefore seeks both to contribute to theory that explains the determinants of each
single parameter and also to provide empirical evidence that tests the constituted theory.
Consequently, the overall research question is:
What factors determine the design of earnouts?
The following sub research questions systemize the overall problem statement along the
common earnout parameters:
(a) What factors determine the earnout ratio?
(b) What factors determine the earnout period?
(c) What factors determine the performance measure?
(d) What factors determine the performance goal?
3
1.2 Approach and structure
The problem statement requires to set up a model on earnout design from which testable
hypotheses are derived. The thesis therefore follows a deductive approach. By applying
standard regression model methodology these hypotheses are subsequently empirically
tested. Consequently, the thesis is both a theoretical and an empirical study of the stated
problem statement with the following structure.
Chapter 2 starts with a review of available earnout literature. First, theory and empirical
findings regarding the motives to use earnouts in M&A deals are examined. Second, the
common earnout mechanism and its parameters are defined and results from studies on
earnout design are presented.
Chapter 3 develops a theoretical model on effective earnout design based on previous
work by Lukas, Reuer & Welling (2012). Referring to the similarities between earnouts
and financial options, an option pricing model on earnout design is set up. This part is the
theoretical focus of the thesis.
Chapter 4 states the testable hypotheses on effective earnout design. Chapter 5 describes
the data collection process and chapter 6 briefly outlines the empirical methodology used
to test the hypotheses.
Chapter 7 presents descriptive results on the data sample and the results of the hypotheses
testing. Chapter 8 evaluates the evidence in the context of previous studies, discusses
implications for the theoretical model and points out limitations to this thesis and further
avenues for research.
Chapter 9 finally concludes on the thesis’ results and its main theoretical and empirical
contributions to research.
1.3 Scope and delimitation
Since the body of literature on earnouts is still limited, the thesis can present an exhaustive
overview of findings from relevant literature. However, the extensive body on M&A
literature in general is not tackled as the introduction already classifies the earnout
research stream into the overall context of M&A literature.
The theory on earnouts requires insights from information asymmetry theory, game
theory and option pricing methodology. Those assumptions and rationales essential to
4
explain theory and the model on earnout design are introduced but these different
economic schools are not discusses in detail.
Theory and models in literature to explain the use of earnouts are mostly qualitative rather
than mathematical derivations. The theoretical model developed in chapter 3 is
consequently also only derived in a qualitative manner.
The collection of earnout specific data requires in depth analysis of primary sources such
as the public deal announcement of every single deal. No reliable databases are yet
available that comprehensively summarize this information. Therefore, the thesis could
only consider a limited data sample size. Thus, the data sample only includes deals from
UK-based acquirers in the period between 01.01.2006 and 30.06.2015. (See chapter 5 for
details)
The empirical analysis is limited to standard cross-sectional regression models as learned
within the master studies and as dominantly used in earnout research. Since the focus is
on identifying determinants of earnout design, the statistical details of these
methodologies are not part of the thesis.
1.4 Evaluation of sources
The literature referred to in this thesis was retrieved through a broad search in literature
databases made available by Aarhus University and through the bibliography of earnout
related studies. Since the earnout research is still new and its literature is limited, the
literature review in chapter 2 is able to present an exhaustive overview. The downside of
this fact is that many articles must be heavily referred to along the analysis. Also, no
selection criteria regarding the quality of publishing journals could be used in order to
remain a meaningful bibliography size for this thesis.
The empirical data is retrieved from different databases. Through the Bureau van Dijk
database Zephyr a sample of earnout deals was searched. The Bureau van Dijk database
Orbis is used in order to identify proxy companies for the deals’ targets. (See chapter 5
for details) Thomson Reuters Datastream is used for all required stock market data.
Finally, Investegate, a database of public deal announcements of UK acquirers, was used
as the primary source for detailed data on the earnout parameters. The search for deal
announcements was unproblematic, however the announcements itself differ in their
5
information content regarding earnout specific data what consequently reduced the
sample size.
2 Literature review on earnouts in Mergers & Acquisitions
This chapter seeks to give an exhaustive overview of the theory and empirical evidence
of earnout literature so far. In order to examine the problem statement of how to design
effective earnouts, the purposes of using earnouts in M&A deals must be cleared first.
Research related to the motives of using earnouts in fact has dominated earnout literature
so far. Section 2.1 covers the relevant theory and empirical evidence. As the next step,
section 2.2 defines the basic structure of an earnout mechanism and covers the yet limited
literature on earnout design.
2.1 Motives to use earnouts
Literature agrees that an earnout contract is a valuable instrument to mitigate the valuation
risk an acquirer faces in an M&A transaction. Research has identified two hypotheses that
explain how earnouts mitigate this problem. The first and most dominant hypothesis is
that earnouts are able to serve as a solution to problems of information asymmetry
between the acquirer and the target in a deal. The second and less studied hypothesis is
that earnouts also mitigate pre-contractual uncertainty about the target’s future
performance that is present in a deal even in the absence of information asymmetry. Both
hypotheses are described in terms of economic theory first. Subsequently, empirical
evidence for both rationales is summarized. In this way, chapter 2.1 aims at answering
the question of why earnouts should be used in M&A deals in the first place.
2.1.1 Theoretical hypotheses
2.1.1.1 The information asymmetry hypothesis
One of the pivotal parts of negotiating an M&A deal is the agreement upon the purchase
price. While the acquirer faces the risk of overpayment, the seller at the same time might
be concerned with underpayment. This type of risk is especially distinctive in situations
of information asymmetries. (E.g. Travlos, 1987; Chang, 1998) If the bidder of a deal is
less informed about the true value of the target firm than the seller, the two parties might
face a disagreement about the fair valuation of the target resulting in a valuation gap. The
acquirer’s lack of information ultimately leads to problems of adverse selection. (Lukas
6
& Heimann, 2014, p. 482) The rationale how earnout agreements mitigate these problems
is therefore referred to as the adverse selection hypothesis. Moreover, the acquirer might
overpay for a target if key human capital with informational advantages leave the new
combined firm post-closing. In this case the acquirer faces an agency problem. (Beard,
2004) The rationale of how earnouts can mitigate these risks from post-contractual
information asymmetry is referred to as the agency problem hypothesis. Together, the
rationales of how earnouts mitigate risks from information asymmetries present the
dominant hypothesis in research. (See Kohers & Ang 2000; Ragozzino & Reuer 2009;
Cain, Denis & Denis 2011)
The basic assumption underlying the adverse selection hypothesis is information
asymmetry between the bidder and the target that leads to suboptimal outcomes of the
negotiations. While the seller knows about the true value of the target firm, the bidder has
limited access to this information and therefore faces an informational disadvantage.
(Myers & Majluf, 1984) Since the acquirer has to base his valuation of the target firm on
limited information, he will bear a greater risk of misvaluation and overpayment. In face
of this risks, the bidder would only be willing to offer a relatively low acquisition price.
(Beard, 2004, p.26) At the same, the seller would be unable to command an attractive
sales price. Thus, the parties have to deal with a valuation gap. This situation on the M&A
market resembles the problems of a “market for lemons” as famously introduced by
Akerlof (1970). Due to its informational disadvantage, the bidder cannot distinguish
between high-quality and low-quality targets and therefore only offers a price to which
the high-quality sellers are unwilling to close the deal. As a consequence, the high-quality
targets will leave the market and the bidder is left with low-quality targets. This dynamic
is defined as the adverse selection problem. (Akerlof, 1970, p. 493) In the extreme case,
the market might collapse. In any case, the information asymmetry causes costs since the
parties to the deal are forced to accept suboptimal outcomes in order to agree on a
purchase price at all. (Akerlof, 1970, p. 495)
As Krishnamurti & Vishwanath (2008, p. 134) point out, the motive of bridging a
valuation gap between bidders and targets is most commonly discussed in earnout
literature. To overcome this problem caused by information asymmetry game theory
suggests that the seller needs a tool to credible signal the quality of the target firm to the
bidder. (See e.g. Spence, 1973) In fact, the literature on earnouts agrees that an earnout
mechanism can serve as such a signal: “By deferring the valuation decision until post-
7
closing, sellers can create credible information signals and thereby resolve potential
adverse selection problems.” (Quinn, 2012, p. 138-139) By accepting that a part of the
acquisition price is contingent on future performance of the target, the target reveals his
confidence of meeting the prespecified performance benchmark. For low quality firms,
this signal would be too costly to replicate since they are not confident to reach the
demanded performance goals of the earnout. (Beard, 2004, p. 27)
By allowing high-quality sellers to convey hidden information about their unobservable
quality to the bidder, the earnout helps separating the high quality sellers from the low
quality sellers. In this way, a contingent payment such as the earnout functions as a so
called separating equilibrium. (See Spence, 1973, p. 358) Ragozzino & Reuer (2009)
suggest that earnouts mitigate the adverse selection problem by transferring the risk of
overpayment from the acquirer to the bidder. The bidder adheres its initial valuation and
pays a corresponding price upfront while any earnout premium paid at a later point in
time would only reflect that the target outperformed the bidder’s expectations. (Beard,
2004, p.26) At the same time, the seller is compensated for this performance and in total
receives a purchase price closer to his original valuation. In this way, both bidder and
seller potentially benefit from the features of a contingent earnout. All in all, earnouts
helps to manage risks associated with the adverse selection problem, and thereby
ultimately might bridge the valuation gap.
The agency problem hypothesis assumes that the value of a target firm strongly depends
on key human capital that has an informational advantage as compared to the acquirer.
Kohers & Ang (2000) argue that a loss of key management post-closing would
consequently diminish the target’s value. These agency costs, if significantly high,
potentially jeopardize the entire future success of the business and ultimately result in
overpayment by the acquirer. (Lukas & Heimann, 2014, p. 482) Research agrees that
earnouts help to mitigate these agency problems due to its incentivizing structure. The
earnout agreement defines a part of the overall acquisition price to be contingent on future
performance. Consequently, in order to cash-in the maximal purchase price, the key
management, respectively the target’s shareholders, is incentivized to remain with the
target post-transaction and to reach the agreed performance goal of the earnout
mechanism. (Datar, Frankel & Wolfson, 2001) Thus, the earnout can benefit the M&A
deal by retaining the key management. In a more general sense, an earnout thereby helps
to align the interest of the key agents with the interest of the shareholders of the acquiring
8
firm. (Beard, 2004, p. 27) Even in the case that the target’s value is not to a large part
dependent on key human capital and the management is replaced after closing of the deal,
the earnout still functions as an incentive tool to the agent in place.
To sum it up, in theory earnouts carry the value to mitigate problems of information
asymmetry in M&A. Potentially, these contingent payment agreements ultimately
diminish the risk of overpayment for the acquirer by bridging valuation gaps and retaining
key management that would have otherwise jeopardized or even prevented the deal from
closing. Kohers & Ang (2000, p. 445) conveniently wrap up the two potential benefits of
earnouts as “agreeing to disagree” (bridging the valuation gap) and “agreeing to stay”
(management retention).
2.1.1.2 The uncertainty hypothesis
The uncertainty hypothesis offers an opposing explanation for the value of utilizing
earnouts in M&A. It argues that earnouts still help to overcome diverging expectations
and valuations even in the absence of information asymmetries. While acknowledged by
several scholars, the explicit formulation of this hypothesis is only to be found in a recent
empirical study by Quinn (2012). The author assumes acquirer and target to have
symmetric information pre-closing of the deal. However, in order to agree on a fair value
estimation of the target, the bidder and seller are required to have joint expectations about
the future. Their expectations in turn are intimately tied to their risk preferences. Less
risk-averse parties would be more optimistic about the future, while more risk-averse
parties would be more pessimistic. If buyers and sellers do not share the same preference
for risk, the expectations differ and would consequently result in a valuation gap. In this
case, the valuation gap emerges from uncertainty about future states rather than from
information asymmetry. The phrase “uncertainty” in this thesis therefore is always related
to pre-contractual uncertainty about future states and does not refer to the problems
arising from information asymmetry.
Following Quinn (2012), an earnout contract would still be able to overcome this
problem. If the parties can agree on a transaction structure that distributes the probability
of an adverse event to the party with the larger risk-preference, then the parties are finally
able to create uniform assumptions and to generate an efficient price for the seller.
Earnouts shift uncertainty to the seller, which is commonly expected to be more risk-
seeking, not because he is better informed about future states but because the seller agrees
9
to bears the costs of being wrong due to its higher risk preference. Thereby, earnouts
allow the buyers in turn to reduce their exposure to uncertainty. In this way, the bidder
reduces its risk of overpaying if the target turns out to be less successful as ex ante
predicted by the seller. Earnouts therefore undertake an important function of distributing
pre-contractual uncertainty. (Lukas & Heimann, 2014, p. 484)
So, even in the absence of information asymmetry, an earnout contract serves as an
effective risk reducing tool by hedging against the risk of misvaluing targets and thereby
ultimately closing the valuation gap. (Kohers & Ang, 2000)
2.1.2 Insights from empirical studies
2.1.2.1 Evidence for the information asymmetry hypothesis
Empirical research reports strong evidence for the adverse selection hypothesis. The
studies follow two approaches to prove the potential benefits of earnouts. On the one
hand, most of the studies examine if earnouts are used more likely in situations that
indicate information asymmetry. (Kohers & Ang, 2000; Datar, Frankel & Wolfson, 2001;
Ragozzino & Reuer, 2009; Cain, Denis & Denis, 2011) Their findings are presented first.
On the other hand, some studies test if the hypothesized benefits of using earnouts is
reflected in wealth effects for the acquirer’s shareholders. (Lukas & Heimann, 2014;
Barbopoulos & Sudarsanam, 2012; Kohers & Ang, 2000) These studies are briefly
summarized as the second step. At the end of each section, the main empirical findings
are summarized in a table.
The choice to use earnouts
The ground-breaking empirical study on earnout use was conducted by Kohers & Ang
(2000). Their results report strong evidence that these contingent payments are more
likely to be used for deals that face a high degree of information asymmetry as indicated
by certain deal and firm characteristics. For a sample of 938 earnout deals in the US
between 1984 and 1996, the authors try to predict the use of earnouts by means of a
logistic regression on indicators of information asymmetry and the importance of human
capital. High-tech firms are considered to imply information asymmetry since their value
primarily depends on difficultly verifiable growth opportunities. Firms from the service
industry imply information asymmetries as they usually carry low tangible, easily
valuable assets and heavily rely on intellectual property and key human capital. Privately
10
held firms are considered problematic since they lack previously disclosed information.
Also, acquisitions in unfamiliar territories such as cross-industry deals are considered to
indicate information asymmetries due to a gap of market-specific knowledge between
acquirer and target. Finally, bidders that can least afford to absorb the risk of
overpayment, i.e. small bidders, should tend to use earnouts more likely.
The reported results strongly support the expectations and suggest that the use of an
earnout is significantly more likely for acquisitions that carry these deal and firm
characteristics. Further US-based research carried out by Data, Frankel & Wolfson
(2001), Ragozzino & Reuer (2009) and Cain, Denis & Denis (2011) consistently confirms
these basic findings for more recent deal samples. Barbopoulos & Sudarsanam (2012)
prove these results to be robust also for a sample of UK-based acquirers. Consequently,
there is wide consensus among scholars that earnouts are chosen as a payment form in
situations where information asymmetry problems are expected.
Apart from evidence regarding the choice to use earnouts, Kohers & Ang (2000) further
examine the retention rate of target management in the post-acquisition phase by a long-
term analysis. Statistics suggest that the majority of the tracked managers remain with the
firm after the earnout period. The authors therefore conclude that earnouts serve the two
main functions hypothesized from theory: As a mechanism to shift mispricing risk from
the acquirer to the seller, and as a retention incentive for valuable target human capital.
(Kohers & Ang, 2000, p. 475) Similar conclusions are drawn by Ragozzino & Reuer
(2009) who find that earnouts as contractual agreements substitutes for other ways by
which a bidding firm might otherwise deal with information asymmetries such as equity
partnerships. This result implies that using earnouts is not only a choice of the payment
method, but also a strategic decision.
Another focus of research surrounds cross-country deals in particular. Kohers & Ang
(2000) expect a high degree of information asymmetry to be present in these deals,
however their empirical evidence shows that earnouts are not significantly more likely
opted for in these circumstances. Datar, Frankel & Wolfson (2001) and Barbopoulos &
Sudarsanam (2012) underline this exception by reporting evidence that for cross-country
deals per se earnouts are even less likely to be used. The studies explain this result by the
fact that the difficulties of enforcing an earnout contract in a foreign country under a
different legal system may offset the benefits. In fact, a more differentiated analysis
documents that earnouts are more likely used if acquirer and target of a cross-country deal
11
operate under similar legal system and less likely used if these legal system are not
considered congruent. (Kohers & Ang, 2000) These insights shed light on the operational
requirements for an earnout to serve as an effective tool.
Finally, empirical research also investigate certain acquirer and industry characteristics
that make the use of earnouts less valuable and therefore less likely. Datar, Frankel &
Wolfson (2001) empirically prove that earnouts are less likely used in industries with an
active M&A market. The authors argue that in an active market, the acquirer is provided
with more reference points to base its valuation decision on and consequently face less
need to secure against valuation risk. Barbopoulos & Sudarsanam (2012) argue that older
acquirers cope better with valuation risks due to more experience and that larger acquirers
are less exposed to overpayment risk and therefore have less need for an earnout contract.
Indeed, their empirical analysis reports that earnouts are less likely chosen by older and
larger acquirers. Furthermore, there is evidence that larger deals favour the use of
earnouts. The authors conclude that an increasing deal size increases the absolute
valuation risk and accordingly increases the value to use an earnout.
All in all, there is strong empirical support in literature for the information asymmetry
hypothesis. For deal characteristics that imply adverse selection and agency problems to
be severe, earnouts are more likely used as the payment form. These results strongly
indicate that acquirers rely on earnouts to serve as effective signalling and incentive
instruments. Table 1 summarizes the evidence from research and shows the deal and firm
characteristics that are considered as indicators for information asymmetry, their effect
on the valuation risk and their impact on the probability than an earnout is used.
Table 1: Main findings of empirical studies on the choice to use earnouts
Deal characteristic Effect on
valuation risk
Probability to
use earnouts Study
High-tech target + + Kohers & Ang (2000)
Datar, Frankel & Wolfson (2001)
Barbopoulos & Sudarsanam (2012)
Service-industry target + + Kohers & Ang (2000)
Datar, Frankel & Wolfson (2001)
Barbopoulos & Sudarsanam (2012)
Private target + + Kohers & Ang (2000)
Datar, Frankel & Wolfson (2001)
Cain, Denis & Denis (2011)
Barbopoulos & Sudarsanam (2012)
Small target + + Datar, Frankel & Wolfson (2001)
Young target + + Ragozzino & Reuer (2009)
12
Cross-industry deal +
+
Kohers & Ang (2000)
Datar, Frankel & Wolfson (2001)
Ragozzino & Reuer (2009)
Cain, Denis & Denis (2011)
Barbopoulos & Sudarsanam (2012)
Cross-country deal + ~
-
Kohers & Ang (2000)
Datar, Frankel & Wolfson (2001)
Barbopoulos & Sudarsanam (2012)
if different legal systems - Kohers & Ang (2000)
if similar legal system + Kohers & Ang (2000)
Active M&A market - - Datar, Frankel & Wolfson (2001)
Acquirer’s size - - Kohers & Ang (2000)
Barbopoulos & Sudarsanam (2012)
Acquirer’s age - - Barbopoulos & Sudarsanam (2012)
Source: Author’s table
The wealth effect of earnouts
The second stream of research presented here follows the logic that the hypothesized
benefits of earnouts should ultimately be reflected in terms of value gains to the
shareholders. As commonly defined in M&A literature, these wealth effects are measured
in terms of stock market reactions around the deal announcement date.
In connection with their findings that information asymmetry triggers the use of earnouts,
Kohers & Ang (2000) also conducted the first event-study of stock market reactions to
earnout transactions. The authors find that acquiring firms using earnouts per se show
positive and significantly larger abnormal returns than bidders using different method of
payments. The empirical study reports abnormal returns for bidders in earnout deals of
1.35% on announcement date as compared to 0.9% for bidders in cash or stock deals.
Beard (2004) and Mantecon (2009) confirm these results for US-based acquirers,
Barbopoulos & Sudarsanam (2012) report the same evidence for UK-based acquirers and
Lukas & Heimann (2014) are finally able to document the same positive effect for a
sample of German acquirers.
Moreover, these studies reveal that market reactions are larger the higher the degree of
information asymmetry present in a deal. Consistently, studies show that for deals facing
high information asymmetry bidders that utilize earnouts enjoy significantly larger
returns than bidders that opt for other methods of payment. (Kohers & Ang, 2000; Beard,
2004; Barbopoulos & Sudarsanam, 2012) The same deal and firm characteristics that
trigger the use of earnouts also positively impact the market reaction, i.e. for private
13
targets, targets from high-tech or service industries and in the case of cross-industry deals.
These results imply that especially in cases where earnouts are the appropriate choice, the
bidding firm is better off using an earnout as compared to other payment agreements.
Also, the takeover premium paid to target shareholders is significantly higher for earnout
deals than for comparable deals using simple stock or cash offers indicating that both
parties benefit from the earnout’s features. Consequently, this evidence suggests that
contingent payments are perceived by the market to mitigate adverse selection and agency
problems. (Barbopoulos & Sudarsanam, 2012, p. 693)
Again, scholars explicitly focus on market reactions to cross-country earnout deals.
Mantecon (2009) finds no evidence, that buyers benefit from using earnouts in cross-
country deals. To the contrary, buyers gained from earnouts in domestic transactions. In
a more differentiated approach, Lukas & Heimann (2014) show that earnouts in cross-
country deals does not add value to the acquirer per se, while a positive effect is examined
if the legal systems are considered similar. In reference to Kohers & Ang (2000) these
findings are in line with the argument that the costs of enforcing and monitoring earnout
contracts in countries with differing legal systems and accounting standards outweigh the
benefits.
Furthermore, research on the market reactions identifies several acquirer characteristics
that tend to positively impact the wealth effect for the shareholders. Lukas & Heimann
(2014) argue that earnouts are more likely used in transactions involving Research &
Development (R&D)-intensive targets that studies often classify as targets from the high-
tech industries. However, the researchers point out that earnouts are not by definition an
effective tool to mitigate R&D-induced information asymmetry but that the effectiveness
depends on the “absorption capacity” of the acquirer. This capacity is defined as “the
ability to identify, process and interpret encoded knowledge” (Lukas & Heimann, 2014,
p. 486). For acquirers that carry a large stock of R&D related knowledge themselves this
capacity should be more established. The conducted event study confirms this hypothesis
and reports significant positive abnormal returns for acquirers with large absorption
capacity. The study also shows that larger acquirers tend to profit more from the use of
earnouts, which might indicate that these firms have better access to information on the
target and therefore more appropriately use earnouts in the M&A context. (Lukas &
Heimann, 2014, p. 486).
14
Finally, Barbopoulos & Sudarsanam (2012) accumulate the results of empirical research
on earnouts in their model of optimal earnout choice. The main innovative finding of the
paper is, that not only the simple use of earnouts is value enhancing or the use in case of
appropriate deal and firm characteristics, but that the wealth effect is also significantly
higher when it is optimal to use earnouts on the industry level. The authors identify those
target industries in which the information asymmetries between acquirer and target are
particularly high and show that the market reaction is significantly more positive if
acquirers use earnouts in these circumstances. The industries deemed optimal for earnout
use are characterized to carry large intangible assets such as intellectual property and to
be R&D-intensive. Clearly, the earnout’s benefits of retaining human capital and shifting
risks due to uncertain project outcomes to the target are more valuable for these target.
The only contradictory results are reported in a long-term study on the performance of
earnouts from an accounting perspective by Quinn (2012). To answer the question if
earnouts successfully address the problem of adverse selection, the author analyses the
actual fair value estimates of the earnouts in the acquirers’ balance sheets. Though the
author finds that earnouts are more prevalent in circumstances where one might expect
adverse selection to be a potential problem, there is little evidence to suggest that earnouts
actually function as a solution to the adverse selection problem. The researcher argues
that in order to function as a credible signalling tool for high quality targets, then post-
closing fair value estimates would have to increase as the acquiring firms confirm the
target’s quality. However, the comparison of fair value estimates of earnouts at
acquisition date and during the earnout period reported no significant differences. From
these results, the author concludes that earnouts fail to sort high quality from low quality
targets and therefore fail as a valid signalling tool. A limiting factor to this study is a small
sample of 140 transactions only. Furthermore, the author himself points towards the issue
that accounting data might be biased by management decisions. Especially, firms have an
incentive to overestimate the likelihood of an earnout payment and thereby its fair value
at acquisition date due to conservative accounting practices. Consequently, most of the
earnout fair values would decline over time as the acquirer adjusts its estimation of the
likelihood to pay an earnout to more realistic levels. Therefore, on average no upwards
adjustment of the fair value of earnouts would be observable over time.
After all, these results are not confirmed by further research studies and should not change
the overall acceptance of the information asymmetry hypothesis. The main findings from
15
the presented studies are summarized in table 2, reporting the focus of the studies and the
observed value effect to the acquirer’s shareholders around the deal announcement date.
Table 2: Main findings of empirical studies on wealth effects of earnouts
Focus of study Value effect
to acquirer Study
Earnout use in general + Kohers & Ang (2000)
Beard (2004)
Mantecon (2009)
Barbopoulos & Sudarsanam (2012)
Lukas & Heimann (2014)
For US acquirers Kohers & Ang (2000)
Beard (2004)
Mantecon (2009)
For UK acquirers + Barbopoulos & Sudarsanam (2012)
For German acquirers + Lukas & Heimann (2014)
Earnout use with high information asymm. + Kohers & Ang (2000)
Beard (2004)
Barbopoulos & Sudarsanam (2012)
Private target + Kohers & Ang (2000)
Beard (2004)
Barbopoulos & Sudarsanam (2012)
Lukas & Heimann (2014)
High-tech target + Kohers & Ang (2000)
Beard (2004)
Barbopoulos & Sudarsanam (2012)
Service-industry target + Kohers & Ang (2000)
Beard (2004)
Barbopoulos & Sudarsanam (2012)
Cross-industry deal + Kohers & Ang (2000)
Barbopoulos & Sudarsanam (2012)
Cross-country deal - Mantecon (2009)
Barbopoulos & Sudarsanam (2012)
If similar legal system + Lukas & Heimann (2014)
If different legal system - Lukas & Heimann (2014)
Earnout use of large acquirers + Lukas & Heimann (2014)
Earnout use of R&D-intensive acquirers + Lukas & Heimann (2014)
Earnout use of experienced acquirers + Beard (2004)
Deal size + Barbopoulos & Sudarsanam (2012)
Source: Author’s table
2.1.2.2 Evidence for the uncertainty hypothesis
The alternative uncertainty hypothesis is far less studied in academia than the information
asymmetry hypothesis presented before. In fact, Quinn (2012, p. 163) points out that the
information asymmetry hypothesis is dominant among scholars while the author claims
16
that the uncertainty hypothesis is common amongst practitioners. Consequently, the
review of literature could only identify two studies that explicitly examine the role of pre-
contractual uncertainty on earnouts. (Reuer, Shenkar & Ragozzino, 2004; Lukas &
Heimann, 2014)
Both studies use the volatility or standard deviation in the target’s industry as a measure
of uncertainty. Indeed, this type of uncertainty is not related to situations of information
asymmetry, rather it presents an indicator for uncertainty about the future performance of
a target that both the acquirer and the target face symmetrically. To this extent,
uncertainty creates valuation risk that cannot be reduced by the buyer by due diligence,
screening or other selection mechanisms. According to the uncertainty hypothesis, this
type of uncertainty evokes valuation gaps as well, simply because the buyer and the seller
might have different risk preferences and therefore estimate different future scenarios.
Reuer, Shenkar & Ragozzino (2004) test if earnouts are used for risk-sharing purposes
only rather than to mitigate information asymmetry problems. The uncertainty measure
is specified as the volatility of net sales in a particular industry. However, results from
their regression models indicate that uncertainty is not significant in explaining the use of
earnouts. Consequently, their test rejects the uncertainty hypothesis.
Lukas & Heimann (2014) follow the alternative approach and test if market reactions to
the use of earnouts are related to the target’s uncertainty. The authors argue that in the
case of higher volatility in the target’s cash flows, the valuation risk is more severe since
the prediction of future cash flows is problematic. Consequently, the study states the
hypothesis that with increasing uncertainty in the target’s performance, the use of
earnouts becomes more favourable. The authors define the target’s uncertainty as the
standard deviation of daily returns from comparable target companies. Indeed, their
empirical results report a positive and significant impact of the uncertainty measure on
the bidder’s abnormal return. (Lukas & Heimann, 2014, p. 491) Thus, this study finds
evidence that earnouts are viewed by the market as an appropriate instrument to manage
risks from pre-contractual uncertainty in a deal.
To sum it up, research offers limited and contradictory evidence for the uncertainty
hypothesis.
17
2.1.3 Summary
The motives of using earnouts in M&A deals is the focus of research so far. Theory
suggests that such contingent payments serve as a tool to mitigate valuation risk for the
acquirer and offers two competing hypotheses as explanations.
The information asymmetry hypothesis states that the risk of misvaluing a target firm
stems from problems of adverse selection and agency problems. This research stream
presents earnouts as a solution to this two problems. By shifting a part of the valuation
risk away from the acquirer towards the seller, the earnout contract serves as a signalling
tool for high-quality targets and thereby reduces problems of adverse selection.
Furthermore, the contingent payment functions as an incentive tool for strong post-
closing performance and helps to retain target’s key management. Empirical studies
report strong evidence for this rationale from two perspectives. First, they show that
earnouts are more likely to be used in situations where information asymmetries and
consequently adverse selection and agency problems can be expected. Second, they report
positive market reactions to the use of earnouts, especially in situations of high
information asymmetries. These results are consistent for samples of different time
frames, ranging from 1984 to 2008, and for samples of acquirers from the US, the UK
and Germany.
The uncertainty hypothesis explains that earnouts are valuable to hedge against valuation
risk even in the absence of information asymmetry. Theory argues that diverging target
valuations may simply emerge from the different risk preferences of the acquirer and the
target. By declaring a part of the overall acquisition price contingent on future
performance, an earnout therefore can close this gap by shifting the valuation risk towards
the party with the higher preference, i.e. usually the target. However, literature so far
offers only very limited and inconsistent empirical evidence for this alternative
hypothesis.
We can conclude that the concept of information asymmetry has high power to explain
why earnouts are used in M&A deals in the first place and when it is appropriate.
Logically, now the research question of this thesis arises of how to design earnouts to
make it an effective and valuable tool which is tackled in the next chapter.
18
2.2 Design of earnouts
After having explained the choice to use earnouts in M&A deals. This section now turns
towards the research question of how earnouts have to be designed in order to serve the
purpose of mitigating valuation risk. The concrete form of an earnout contract can vary
widely. Usually, the structure of an earnout is tailored to the specific characteristics of a
deal and is subject to detailed negotiations of the acquirer and the target. However, there
are some variables that are common to all earnout agreements. These earnout parameters
are introduced and defined in 2.2.1. Subsequently, chapter 2.2.2 summarizes theory and
evidence on the determinants shaping the parameters.
2.2.1 The earnout parameters
Reuer, Shenkar & Ragozzino (2004, p. 20) define earnouts as “deferred variable payments
tied to the target’s ability to meet prespecified performance goals within a certain time
frame after the deal has been consummated”. From this definition and in reference to
further attempts (e.g. Lukas, Reuer &Welling, 2012), we can conclude on the common
parameters of earnouts that together constitute the earnout mechanism:
Earnout premium:
The amount paid to the seller that is dependent on
the target’s performance. This parameter is also
referred to as the contingent payment, the deferred
variable payment or simply the earnout payment.
Earnout ratio: The earnout premium in proportion to the overall
maximal acquisition price (fixed upfront payment at
acquisition plus earnout premium).
Earnout period: The prespecified time frame over which the target’s
performance is measured.
Performance measure: The prespecified measure of the target’s
performance.
Performance goal: The prespecified goal in terms of the performance
measure that the target has to reach in order to
receive the earnout premium.
19
Required performance increase: The increase in the performance measure compared
to pre-closing performance that is required in order
to reach the performance goal.
Despite considerable heterogeneity in the specific details, the basic functionality of an
earnout mechanism is common to each contract. (Krishnamurti & Vishwanath, 2008, p.
138) At the time of the acquisition, the acquirer pays a fixed upfront payment to the seller.
During and until the end of the earnout period, the performance of the target is monitored
according to the prespecified performance measure. At the end of the earnout period, it is
determined if the target reached the prespecified performance goal. If so, the acquirer
pays the earnout premium to the seller. In case of a performance below the performance
goal, the acquirer does not pay the earnout premium.
In practice, earnouts are more complex and unique than the simplified mechanism
described above. Empirical studies show that all earnout parameters exhibit substantial
heterogeneity within a chosen sample of earnout deals. (E.g. Kohers & Ang, 2000; Beard,
2004; Cain, Denis & Denis, 2011 as referred to in the descriptive statistics in chapter 7)
The reported heterogeneity in earnout parameters indicates that earnout mechanisms are
actively designed instruments rather than standardized “off-the-shelf” agreements. This
conclusion in turn implies that there might be some identifiable characteristics that
determine the shape of each earnout parameter. The current state of research on the
determinants of earnout design is therefore examined in the next section.
2.2.2 Current state of research
Theory on the appropriate design of earnouts is limited. Some studies analyse single
earnout parameters but offer little theoretical rationale for their findings. Only Cain, Denis
& Denis (2011) examine the determinants of all relevant earnout parameters
simultaneously. Nevertheless, all these study consistently seek to explain the parameter
design by the same indicators of information asymmetry that determined the use of
earnouts in the first place. The findings are summarized for each parameter separately in
the following paragraphs.
Determinants of the earnout ratio
The earnout ratio attracted the most academic interest. Kohers & Ang (2000), Beard
(2004) and Cain, Denis & Denis (2011) find that the same characteristics that affect the
20
likelihood that earnouts are used also determine the size of the earnout ratio. For private
targets, cross-industry deals and in case of high-tech targets and service-industry targets
the earnout ratio tends to increase. Kohers & Ang (2000) argue that if problems from
information asymmetry are more severe, it is favourable to shift a larger part of the
valuation risk towards the seller. Beard (2004) further concludes that a higher earnout
ratio allows the target to even stronger signal its quality to the acquirer. Cain, Denis &
Denis (2011) add the insight that the earnout ratio is also driven by the importance of
target’s human capital. Thus, the contingent payment is raised the more the acquirer faces
the need to incentivize key management to stay. Furthermore, the earnout ratio is also
reported to be higher for targets from more volatile industries, suggesting that a higher
degree of uncertainty shifted to the target is preferable in these environments. (Cain,
Denis & Denis, 2011) Event-study analyses show that the market reacts more favourably
to larger earnout ratios. (Beard, 2004; Barbopoulos & Sudarsanam, 2012) Again, this
value enhancing effect for the acquirer might stem from the more favourable risk
distribution in case of larger earnout ratios.
All in all, there is evidence for the earnout ratio to be driven by indicators for adverse
selection and agency problems. The fact that the value of an earnout increases with the
degree of information asymmetry and uncertainty is therefore mirrored in the same
positive relationship for the earnout ratio. This result is intuitive, since increased valuation
risks not only favours the simple use of earnouts but also favours to shift more of the risk
to the target by larger contingent payments.
Determinants of the earnout period
Barbopoulos & Sudarsanam (2012, p. 693) further examine the effect of different lengths
of the earnout period on the acquirer’s abnormal returns. Their analysis reveals that the
earnout period, in contrast to the earnout ratio, has insignificant power to explain the gains
to the acquirer’s shareholders. A contradictory result is reported by Lukas & Heimann
(2014) who show that buyers benefit from shorter earnout periods. They refer to contract
theory and argue that longer earnout periods give rise to an increased probability that
results of the performance measure are manipulated or negotiated performance goals are
questioned. Ultimately, this might put the successful post-closing cooperation between
acquirer and target at risk or triggers lawsuits that create additional costs for the acquirer.
Cain, Denis & Denis (2011) find earnout periods to be longer when the information
asymmetry is likely to be resolved over time like for targets that heavily rely on the
21
successful outcome of R&D projects. Interestingly, this positive impact is offset in
situations of high uncertainty in terms of industry volatility which tends to decrease the
length of the earnout period. (Cain, Denis & Denis, 2011, p.161) The authors argue that
stronger fluctuations in the performance measure over a longer period of time would
complicate to measure the performance that is actually controllable by the target. Shorter
earnout periods are therefore favourable in case of increased uncertainty in an industry.
All in all, the empirical results are inconsistent. Even more importantly, the economic
theories referred to as explanations of the design of earnout periods are inconsistent as
well. While Barbopoulos & Sudarsanam (2012) lack any discussion of their results, Cain,
Denis & Denis (2011) acknowledge reverse effects of information asymmetry and
uncertainty. Consequently, research misses a comprehensive approach to explain the
determinants of the earnout period.
Determinants of the performance measure
Cain, Denis & Denis (2011) finally report the choice of a performance measure to be
related to proxies indicating the amount of information revealed by that measure and to
the verifiability of that measure. More specifically, if there is high information asymmetry
between the buyer and the target or if the target’s industry is highly uncertain the measure
is more likely to be sales rather than profits. This might be due to the fact that sales are
more easily verifiable. (Cain, Denis & Denis, 2011, p.162) Consequently, for less
uncertain industries and if information asymmetry is not severe, income is more likely
used as performance measure. Moreover, the authors argue and find empirical evidence
that for targets whose value is primarily derived from future growth-opportunities, i.e.
typically younger firms, the performance measure is more likely to be a non-financial
measures. Since this type of measures is often related to milestones in growth projects,
more relevant information is conveyed than it would be the case with sales or profit
measures. In contrast to the earnout period, information asymmetry indicators and a
measure of uncertainty show complementary effects on the performance measure.
Summary
To sum it up, the information asymmetry theory is referenced as the dominant theory
explaining the shape of the earnout parameters. However, except for the earnout ratio
studies so far reported inconsistent results and offer no model that integrates all of the
parameters that constitutes an earnout mechanism. Especially, no study so far comprised
22
the performance goal and consequently the required performance increase in its analysis.
Research therefore still lacks a comprehensive and complete approach to answer the
question of how earnouts should be designed.
However, only recently Lukas, Reuer & Welling (2012) formulated an alternative model
of earnout design. Its strength is that it directly follows from the well-studied and widely
accepted information asymmetry hypothesis and that it comprises the parameters of
earnout ratio, earnout period and required performance increase simultaneously. Its
novelty stems from the fact that the model dynamics are based on the option-like
characteristics of earnout mechanisms. Although it is a recurring concept in earnout
research to describe earnouts as options on a target’s fair value (e.g. Bruner & Stiegler,
2001) this approach is not deeply examined and also the model by Lukas, Reuer &
Welling (2012) lacks an empirical testing. Therefore, their model offers a promising yet
unexplored new stream in earnout literature to overcome the inconsistencies and the lack
of a comprehensive theory on earnout design so far. The model and the option-like
characteristics are analysed in detail in the next chapter.
3 Theoretical model on the design of earnouts in Mergers &
Acquisitions
A systematic approach to model the appropriate design of an earnout should inevitably
be linked to the motives of using earnouts in the first place. In order to receive the full
potential benefits of an earnout, the structure of the earnout mechanism must ensure to
serve the purposes of mitigating adverse selection and agency problems. Indeed, literature
refers to information asymmetry concepts to explain earnout design. However, the
presented studies so far miss to identify one single factor that determines the earnouts
effectiveness in solving these problems and which at the same time is directly affected by
the earnout parameters.
Lukas, Reuer & Welling (2012) identify this missing factor to be the likelihood that an
earnout payment is paid at the end of the earnout period. Clearly, in order to serve as a
powerful incentive tool, the acquirer should control for this likelihood. The more likely
the target will finally receive the earnout premium, the less efforts it will invest in the
post-closing phase. Moreover, if the likelihood for an earnout payment is high, even
relatively low quality targets are willing to accept this contract and the separating
23
equilibrium is violated. As a consequence, the earnout’s power to mitigate agency and
adverse selection problems is strongly dependent on its likelihood to result in an earnout
payment.
Therefore, the likelihood of the earnout premium to be paid has to be an important
reference point for the effective design of earnouts. Lukas, Reuer & Welling (2012)
develop a model on earnout design that is based on this very rationale. The authors state
that the likelihood is determined by the earnout parameters through dynamics that are
derived from option pricing methodology. They justify the use of option pricing
techniques with similarities between an earnout mechanism and a financial option. These
option-like dynamics in combination with the information asymmetry assumptions
outlined in the previous paragraph constitute the authors’ game-theoretic option pricing
model on earnout design.
The information asymmetry related assumptions were already introduced in detail in
chapter 2. The option-like characteristics of earnouts, however, are novel. As this
assumption is essential, chapter 3 starts with an analysis of the similarities between
earnouts and financial options. After that, the model by Lukas, Reuer & Welling (2012)
can justifiably be introduced in 3.2.
3.1 The option-like characteristics of earnouts
Several researchers point out that earnouts have option-like characteristics. (Craig &
Smith, 2003; Caselli, Gatti & Visconti, 2006; Bruner & Stiegler, 2001; Krishnamurti &
Vishwanath, 2008) More specifically, Krishnamurti & Vishwanath (2008, p. 137)
describe an earnout contract as a call option on the target’s future performance.
The upfront payment in an M&A deal would in fact only incorporate an estimation of the
target’s future performance that both acquirer and seller can agree on. The earnout
premium in turn would be based on any additional performance that the target generates
during the earnout period, at least if the performance goal is reached. Achieving the
performance goal therefore implies that the target performs stronger than initially
expected by the acquirer. In this case the target’s fair value would be adjusted upwards.
Since the acquirer then has to pay the earnout premium to the seller, both parties would
eventually benefit from this increase in value. Consequently, an earnout grants both
parties of the deal the right to benefit from this upside potential in the target’s future
performance. Caselli, Gatti & Visconti (2006) specify that the seller therefore holds a
24
long position in the call option since they potentially cash in the earnout premium at the
end of the earnout period. Since the earnout premium usually does not comprise the entire
fair value increase of the target, however, the acquirer would also benefit from a stronger
than expected performance.
This logic is straightforward and allows for the conclusion that earnouts resemble a real
option. According to Real Option Theory (ROT), a real option is the right to undertake a
particular business decision such as an investment in a company. (Berk & DeMarzo,
2014, p. 774) More specifically, earnout contracts resemble the type of a real option to
wait. The option to wait gives rise to two additional sources of value for the acquirer. (For
the general case of investment opportunities as real options see e.g. McDonald & Siegel,
1986; Trigeorgis, 1991; Luehrmann, 1998)
First, like any real option to wait an earnout allows the acquirer to defer the contingent
payment until the end of the earnout period and to enjoy the time value of money. (Del
Roccili & Fuhr, 2001) Second, the earnout includes the option to defer the final purchase
price determination until new and better information becomes available. Since earnouts
are especially used in situations of high information asymmetry, the option to wait is
valuable for the acquirer as valuation risk due to an imperfect information basis is reduced
over time. The acquirer’s informational disadvantage regarding the target’s performance
is mitigated over the length of the earnout period as the target reveals its true performance
power. Consequently, the additional information gained allows the acquirer for a more
appropriate fair value estimation.
Following Kohers & Ang (2000), we can conclude that an earnout just like a common
financial option ensures the acquirer’s ability to participate in the upside potential of the
target while at the same time it insulates him from poor performance. If the target’s fair
value goes up, the acquirer pays the earnout premium and benefits from the additional
value himself too. If the target’s fair value goes down, the acquirer does not pay the
earnout premium. From the acquirer’s perspective, this outcome is also fine. By utilizing
an earnout and waiting with additional investments until the end of the earnout period, he
avoids to overpay for the target what would have been the case if he would have followed
the more optimistic estimations of the seller upfront. (Luehrmann, 1998, p. 53)
Although the similarities between earnouts and a call option seem strong at the first
glance, there are some limitations to this approach. In general, both a financial and a real
25
option give the holder the right but not the obligation to make a certain business decision.
(Berk & DeMarzo, 2014) In case of earnouts, however, both the acquirer and the seller
enjoy the right to participate in upside potential but the acquirer also faces the obligation
to pay the additional contingent payment if the performance goal is triggered. The earnout
contract which is agreed and fixed at closing of the deal therefore also fixes all duties for
both parties. In contrast to a perfect real option situation the acquirer cannot chose the
most attractive alternative after new information on the target is available. (Berk &
DeMarzo, 2014, p. 774) To the contrary, the earnout already determines at closing date
the acquirer’s business decision at the end of the earnout period, i.e. its obligation to pay
an earnout premium. Neither can the buyer adjust his investment decision at the end of
the earnout period, nor can he choose the point in time.
Nevertheless, the earnout can be considered at least a “restricted” real option. Although
the contract restricts discretion, it offers in a simplified manner two possible outcomes.
Either the valuation of the target was understated in the beginning and both parties benefit
from the upside potential or the pessimistic view of the acquirer was appropriate and
misevaluation costs were avoided. Therefore the contingent payment still adds more
flexibility value to an investment project than any purchase agreement that only includes
a simple upfront payment. To further underline the option-like structure of earnout
mechanism, the subsequent chapter describes different earnout payment profiles with the
help of illustrative examples.
3.1.1 Earnout premium payment profiles
Describing an earnout in terms of a call option also eases the understanding of the possible
states in which the earnout might be due. Just like a financial call option, an earnout at
the end of the earnout period can expire in three states:
(1) Out of the money (OTM), if the performance goal is not reached and no premium is
paid
(2) At the money (ATM), if the target’s performance is just below the performance goal
and no premium has to be paid
(3) In the money (ITM), if the performance goal is reached or exceeded and a premium
has to be paid
26
To illustrate the option-like possible outcomes of an earnout, examples A to C present the
value of an earnout in different future scenarios. Although all earnout mechanisms
comprise the same parameters, the formulas that determine the exact earnout premium
can take many forms, varying widely from single lump sum payments to complex
formulas based on the degree to which the performance goals are exceeded. (Del Roccili
& Fuhr, 2001)
However, the following examples A to C illustrate the most common types of earnout
mechanisms. Each example is accompanied by a chart showing the earnout payment
profile with the earnout premium (y-axis) dependent on the performance achieved by the
target at the end of the earnout period (x-axis). Furthermore, each chart shows the
implications of different performance scenarios for the target’s estimated fair value (FV),
the earnout premium to be paid, the status in which the earnout expires and ultimately the
acquirer’s mispricing of the target as the difference between his valuation of the target at
closing and the true fair value of the target at the end of the earnout period. A negative
mispricing implies a loss in the target’s FV while a positive mispricing indicates a gain.
All figures are in million Euros.
Example A: Earnout premium like a call option
At closing of the deal, the target’s estimated fair value and therefore the initial payment
to the target's shareholders is EUR 10mn. Additionally, the target’s shareholders receive
50% of the amount that the Year One EBIT exceeds EUR 5mn and nothing below that
level. Figure 1 shows the payment profile and further implications.
Figure 1: Payment profile and implications of earnout like a call option
Source: Author’s figure
27
Example A describes an earnout premium that is a percentage of the amount to which the
EBIT in year one exceeds the performance goal of EUR 5mn. Clearly, the payment profile
is similar to a financial call option with the strike price EUR 5mn. Once the year one
EBIT exceeds this threshold, the target’s shareholders will cash in, at least partly, the
target’s increased fair value (FV). The accompanying table shows further implication of
this earnout mechanism. For an EBIT ≤ EUR 5mn, the call option expires OTM or at best
ATM and no earnout is paid to the target’s shareholders. However, for an EBIT of EUR
6mn the option is ITM, so target’s shareholders receive an earnout payment of EUR
0.5mn. Assuming an EBIT-multiple of 2, the target’s fair value is initially estimated as
being EUR 10mn, which is paid to the target’s shareholders as the initial consideration.
An EBIT of EUR 5mn in year one post-closing would confirm the expectations on which
the initial estimation was based. However, for an EBIT of EUR 4mn, the estimated fair
value would decrease to EUR 8mn, while an EBIT of EUR 6mn would increase the
estimation to EUR 12mn. So, the valuation decision changes as new information on the
target’s performance is available at the end of the earnout period. Obviously, the
acquirer’s downside risk of mispricing is limited to the initial consideration, while both
parties participate in the upside potential of stronger performance than expected.
Example B: Earnout premium like a binary option
At closing of the deal, the target’s estimated fair value and therefore the initial payment
to the target's shareholders is EUR 10mn. Additionally, the target’s shareholders receive
1mn if the Year One EBIT exceeds EUR 5mn and nothing below that level. Figure 2
shows the payment profile and further implications.
Figure 2: Payment profile and implications of earnout like a binary option
Source: Author’s figure
28
This example illustrates the simplest case of an earnout as a single lump sum payment if
the option expires ITM. The payment profile in this case is similar to a binary option. The
implications for the target’s estimated fair value and regarding the acquirer’s downside
risk of mispricing remain the same as in example A. It should be noticed, however, that
this earnout mechanism includes a cap of maximal payment of EUR 1mn. This implies,
that from the point on that the target’s performance generates a higher additional FV than
EUR 1mn, all the surplus accrues to the buyer.
Example C: Earnout premium like a call option with cap
At closing of the deal, the target’s estimated fair value and therefore the initial payment
to the target's shareholders is EUR 10mn. Additionally, the target’s shareholders receive
50% of the amount that the Year One EBIT exceeds EUR 5mn and nothing below that
level. The maximal earnout payment is capped at EUR 1mn. Figure 3 shows the payment
profile and further implications.
Figure 3: Payment profile and implications of earnout like a call option with cap
Source: Author’s figure
This case extends example A by adding a cap to the earnout mechanism. The payment
profile resembles a combination of a call and a put option. If the call option ends up ITM,
the target’s shareholders cash in a part of the additional generated FV such as EUR 0.5mn
for an EBIT of EUR 6mn. However, the earnout is capped at a maximum of EUR 1mn.
The put option therefore has a strike price of EUR 7mn, that is higher than the call option’s
strike price. Similar to the binary option, this mechanism limits the degree to which the
target’s shareholders participate in the upside potential and also the acquirer’s obligation
to pay if the target’s business performs outstandingly.
29
A more elaborate example of an earnout including a floor and a cap simultaneously is
shown in figure A1 in appendix 1.
All in all, these examples underline the similarity of earnouts and financial options. The
analysis carried out supports the view of researchers that options should therefore be
treated and valued as options. (Krishnamurti & Vishwanath, 2008; Bruner & Stiegler,
2001; Caselli, Gatti & Visconti, 2006) This result is extremely useful as we can justifiably
transfer knowledge about the dynamics and value of financial options to the case of
earnouts. Therefore and as a conclusion of the analytics carried out in this chapter, the
subsequent section matches earnout parameters to parameters of financial options.
3.1.2 Mapping the earnout parameters onto a financial call option
Theory commonly relies on the Black-Scholes Option Pricing Model in order to value
financial call options. (Berk & DeMarzo, 2014, p. 747) This valuation approach only
requires five input parameters to value an option: the stock price of the underlying (S),
the exercise price (X), the time to expiration (T), the uncertainty of the stock (𝜎), and the
risk-free interest rate (i). (Berk & DeMarzo, 2014, p. 748) If these parameters are also
available for a real option like an earnout, the same option pricing technique could be
transferred to the case of earnouts. In fact, practitioners point out the usefulness of option
pricing approaches such as the Black Scholes Model to value earnouts. (E.g. Thompson
& Schnorbus, 2010; American Appraisal, 2015)
Since earnouts resemble real options, the thesis maps an earnout parameters onto the
variables of a financial call option following the same approach how ROT maps
investment opportunities in general onto financial options. (See Luehrman, 1998, p. 52;
Berk & DeMarzo, 2014, p. 778 and) The match is illustrated in figure 4 and explained for
each parameter hereafter.
The current stock price of the underlying (S) is the difference between the acquirer’s
estimation of the target’s fair value at closing and the true fair value of the target at the
end of the earnout period. As the examples A-C illustrate, the actual performance of the
target during the earnout period can cause a positive or negative change in the target’s
estimated fair value. This deviation from the initial fair value estimation by the acquirer
is therefore labelled as “target’s unconsidered fair value”.
30
Figure 4: Mapping an earnout onto a financial call option
Source: Author's chart based on Luehrman (1998, p. 52)
The exercise price or strike price of the option (X) corresponds to the performance goal
written in the earnout contract. As we have seen from the payment profiles, only if the
performance goal is exceeded, the earnout ends up ITM. In this case the target’s true fair
value is higher than initially estimated and both parties share the additional value. The
degree to which the seller participates in a positive unconsidered fair value, i.e. the
earnout premium to be paid, is determined by formulas of varying complexity (e.g. an
EBIT-multiple as in examples A-C).
The time to expiration (T) perfectly corresponds to the earnout period. The uncertainty in
the target’s future performance (𝛔) corresponds to the standard deviation of returns on
the stock of the financial option. It is important to notice that this measure of uncertainty
is not related to information asymmetries between the acquirer and the target but to the
volatility in an industry. Finally, the time value of money (i) resembles the risk-free rate
of return of a financial option.
To sum it up, the earnout parameters are convincingly matched with the variables of a
common financial call option. On basis of this detailed analytical derivation, the thesis
31
now introduces the theoretical model on earnout design by Lukas, Reuer & Welling
(2012) from which hypotheses are derived subsequently.
3.2 A game-theoretic option pricing model on design of earnouts
3.2.1 Original model by Lukas, Reuer & Welling (2012)
In their model on earnouts in M&A deals, Lukas, Reuer & Welling (2012) examine the
optimal timing of M&A deals utilizing earnouts on the one hand and the design of the
earnout parameters on the other hand. Since this thesis aims at insights on the optimal
design of earnouts, the second stream of the model developed by Lukas, Reuer & Welling
(2012) is the focus of interest. The authors derive their model mathematically by means
of dynamic programming. However, for the purpose of describing the determinants of
earnout design, it is sufficient to explain the dynamics and consequences without referring
to its mathematical derivation.
First, this chapter describes how the model’s assumptions follow directly from the
information asymmetry hypothesis. Second, the chapter outlines how the presented model
incorporates implications from option pricing techniques. Finally, the chapter introduces
the game-theoretic dynamics of the model that lead straight to the hypotheses on earnout
design stated by Lukas, Reuer & Welling (2012). The following paragraphs refer to the
model dynamics as illustrated in figure 5.
Information asymmetry assumptions
Consistent with the information asymmetry hypothesis, Lukas, Reuer & Welling (2012)
describe earnouts as a useful tool to mitigate problems of information asymmetry. More
specifically, the authors outline an M&A deal in which the acquirer faces the necessity to
retain the target’s human capital. This is due to maintain market knowledge and
relationships with key customers that otherwise are at risk if the target’s key management
leaves post-closing. Furthermore, in order generate a value-enhancing effect from the
deal, the acquirer desires to create synergies (ϴ) in the post-closing phase. Synergies,
however, critically depend on the target’s cooperation (C) in the post-takeover phase.
Consequently, the acquirer faces two agency problems. On the one hand the success of
the business critically depends on the retention of management, on the other hand it
depends on not perfectly observable cooperation efforts by the target resulting in a moral
hazard problem. (Lukas, Reuer & Welling, 2012, p. 258) We can refer to chapter 2 and
32
conclude that an earnout serves as a solution to both of these problems. By making part
of the overall acquisition price contingent on future performance, the target is incentivized
to retain its key human capital. Moreover, the target is incentivized to cooperate during
the earnout period in order to more likely meet the performance goal.
Clearly, the potential benefits of an earnout agreement therefore critically depend on its
power to incentivize. Most importantly and as a novel aspect to the earnout literature the
authors point out, that this power in turn is varying with the likelihood (N) that the earnout
ends up ITM. Since the similarities of earnouts and financial options were adequately
analytically derived, we can turn towards implications from option pricing methodology
to identify determinants of the likelihood (N).
Implications from option pricing methodology
Following financial option pricing methodology, there are three determinants of the
probability of a financial option to end up ‘in the money’: (Berk & DeMarzo, 2014, p.
720; Hull, 2012, p. 215)
1. The higher the uncertainty of the underlying (σ), the higher the probability
2. The longer the time to expiration (T), the higher the probability
3. The lower the strike price (X), the higher the probability
As the thesis analytically derived before, each of these option parameters has its
counterpart in an earnout mechanism. Consequently, the model by Lukas, Reuer &
Welling (2012) transfers these implications from financial option methodology to the
likelihood (N) that an earnout ends up in the money.
1. The higher the uncertainty in the target’s future performance (𝛔), the higher is N
2. The longer the earnout period (T), the higher is N
3. The lower the required performance increase (𝛙) to reach the performance goal
(X), the higher is N
Similar research that uses option pricing methodology to value earnouts follow the same
logic. (E.g. Bruner & Stiegler, 2001; Caselli, Gatti & Visconti, 2006) Consequently, the
option pricing methodology allows to draw conclusions on how some of the earnout
parameters ultimately affect in which state the earnout is likely to expire.
33
Figure 5: Dynamics of model on earnout design by Lukas, Reuer & Welling (2012)
Source: Author's figure based on Lukas, Reuer & Welling (2012, p. 260)
Game-theoretic dynamics and hypotheses
The essential game-theoretic assumption that the authors state is that the acquirer and the
target firm have different interests post-closing. The acquirer seeks to create synergies
(ϴ) in order to make the deal value enhancing and is therefore dependent on the target’s
cooperation (C) in the post-closing phase. The more the acquirer faces agency problems
and consequently the necessity to incentivize the target, the larger he would determine
the contingent part of the acquisition price, i.e. the earnout ratio (EOR). For the target,
cooperation requires efforts and the target is only willing to invest these efforts as long as
it serves his goal to cash-in the earnout premium. Following game-theory, the higher the
likelihood (N) for an earnout premium to be paid, the less cooperation efforts the target
would invest. Lukas, Reuer & Welling (2012, p. 261) draw the conclusion that with
increasing N, the only way for the acquirer to secure cooperation and ultimately the
desired synergies is by increasing the incentive for cooperation by raising the earnout
ratio. (EOR)
However, as option pricing methodology suggests, the earnout parameters impact N.
More specifically, higher uncertainty (𝛔), longer earnout periods (T) and a lower required
performance increase (𝛙) all increase N and therefore in turn also determine an increase
of EOR. Based on these essential interrelationships, the authors formulate the following
34
hypotheses regarding the appropriate design of earnouts in order to secure their power to
incentivize:
H1: Ceteris paribus, the higher the uncertainty of the target’s future performance, the
higher the earnout ratio.
H2: Ceteris paribus, the higher the earnout period, the higher the earnout ratio.
H3: Ceteris paribus, the lower the required performance increase in the target’s
performance, the higher the earnout ratio.
All in all, the authors convincingly combine insights from the information asymmetry
hypothesis with option pricing methodology and game-theoretic assumptions to arrive at
a model that explains the determinants of the earnout ratio. However, this model lacks the
determinants of the earnout period and the required performance increase. This thesis
therefore advances the model in order to explain all three earnout parameters in an
integrated approach.
3.2.2 Own advancements to the model
Lukas, Reuer & Welling (2012) explicitly state retaining management and mitigating
post-closing moral hazards as the reason for which the acquirer needs to incentivize the
target. The information asymmetry hypothesis further argues that mitigating adverse
selection problems is a purpose of using earnouts. However, the authors do not consider
this benefit within their model.
This thesis suggests that the value of earnouts as a signalling tool can perfectly be
integrated into the model on earnout design by Lukas, Reuer & Welling (2012). For an
earnout to serve as a credible signal for high quality targets, the acquirer should indeed
control for the likelihood that the contingent payment has to be paid in the end. If the
design of an earnout is such that the earnout premium is likely to be paid also lower
quality targets would tend to accept it and the assumption of a separating equilibrium is
violated. Consequently, this thesis suggests that only earnout design that controls for the
likelihood to expire ITM ensures the contract’s power as an effective signalling tool. With
this adjustment, all aspects of the question why an earnout is used in the first place would
also be reflected in deciding how to design an earnout in order to be effective.
The original model needs further adjustments since it does not comprise the determinants
of the earnout period and the required performance increase. Since the thesis is aiming at
35
describing the design of these parameters as well, it is necessary to adjust the model in
this regards. In fact, analysing the original model’s dynamics reveal two major disputable
issues.
First, the original model assumes that the acquirer faces an increase in N by increasing
the incentive for the target to cooperate through a higher EOR. To the contrary, the
acquirer could also face the increased likelihood caused by one of the independent
variables (T, 𝝈, or 𝝍) by shaping the remaining earnout parameters to offset the increased
likelihood. For instance, a higher uncertainty might be matched with a shorter earnout or
a higher required performance increase respectively. Consequently, the acquirer has two
options to react to an increased likelihood. Either the earnout ratio is increased to
encourage cooperation, or the likelihood of paying an earnout premium is decreased by
designing the remaining earnout parameters accordingly. Both ways would serve the goal
of mitigating the increased agency and adverse selection problems.
Second, the model by Lukas, Reuer & Welling (2012) lacks a systematic distinction
between exogenous and endogenous variables. Clearly, 𝝈 is an exogenous variable that
cannot be influenced by the parties of the deal. To the contrary EOR, T and 𝝍 are
endogenous variables of an earnout mechanism and are therefore designed actively.
This distinction allows for an important adjustment of the model dynamics, since all the
earnout parameters considered here should therefore be designed given a certain degree
of uncertainty in the M&A deal. Consequently, the empirical study presented in this paper
will be based on an advancement of the original model as depicted in figure 6.
36
Figure 6: Dynamics of advanced model on earnout design
Source: Author's figure
Additionally to the dynamic of choosing EOR (dynamic A), the acquirer has the option
to control N by choosing T (dynamic B1) and by choosing 𝜓 (dynamic B2). Given a high
degree of uncertainty, the acquirer could either raise the earnout ratio, decrease the
earnout period, or raise the required performance increase. These choices are not mutually
exclusive. As we have to think of the design of an earnout as a process of negotiation
between the buyer and the seller, it is more likely that the acquirer needs to adjust through
all three dynamics rather than shaping only one as he demands. Also, as Lukas, Reuer &
Welling (2012, p. 258) point out accountants and lawyers working on an M&A deal are
paid per rata of the earnout premium. Facing an increasing N solely by raising EOR
would therefore also increase the transaction costs for the buyer. From this perspective,
dynamic B1 and dynamic B2 present attractive alternatives. Consequently, we would
expect the exogenous variable of uncertainty to affect the earnout ratio, the earnout period
and the required performance increase at the same time.
All in all, this advanced model prescribes the acquirer to design the earnout contract
dependent on the degree of uncertainty within the M&A deal. The great upside of this
advanced model is that hypotheses can be derived stating the determinants for each
earnout parameter.
37
4 Hypotheses
The testable hypotheses H1-H3 regarding the earnout parameters are directly derived
from the advanced model on earnout design in 3.2.2. Accordingly, uncertainty about the
target’s future performance is the pivotal explanatory variable for each parameter.
H1: The higher the uncertainty of the target’s future performance, the higher the earnout
ratio.
H2: The higher the uncertainty of the target’s future performance, the shorter the earnout
period.
H3: The higher the uncertainty of the target’s future performance, the higher the
required performance increase.
The choice of the performance measure, however, is not directly linked to the game-
theoretic option pricing model. In order to test this remaining earnout parameter, a
hypothesis is therefore derived from empirical results available so far. Again, research
suggests that uncertainty is one key determinant for this choice.
As introduced before, Cain, Denis & Denis (2011) refer to the “informativeness principle”
of Holmström (1979) to explain the appropriate use of performance measures. They
suggest that in case of moral hazard problems an incentive tool such as the earnout should
be tied to observable and verifiable measures of the target’s efforts. In case of adverse
selection problems, the performance measure should be an observable and verifiable
signal of the target’s true value. Accordingly, Cain, Denis & Denis (2011, p. 162)
hypothesize that for targets with high growth opportunities but low profitability today, a
non-financial or a measure of sales are most appropriate as these measures present drivers
of future profitability. Indeed, their empirical results indicate that for targets with high
growth opportunities the performance measure is more likely to be sales or a non-financial
measure rather than income. Furthermore, they find that for targets from highly volatile
industries a measure of sales is more likely to be used. The authors argue that sales in
uncertain environments are easier to verify than income measures which might be
affected by arbitrary cost allocation.
Since this logic just like the hypotheses 1-3 follows directly from information asymmetry
problems and since uncertainty again seems to be an important determinant of the earnout
38
parameter performance measure, this thesis will test the results reported by Cain, Denis
& Denis (2011) through the following hypotheses:
H4a: For high uncertainty in the target’s future performance, the performance measure
for the earnout agreement is more likely to be sales.
H4b: For high uncertainty in the target’s future performance, the performance measure
for the earnout agreement is less likely to be a measure of income.
H4c: For targets with high growth opportunities, the performance measure for the earnout
agreement is more likely to be sales or a non-financial measure.
5 Data sample creation
This chapter describes the three step data selection process carried out to create a sample
of data required to test the hypotheses H1-H4. First, deals that have used an earnout
agreement are identified through the Zephyr database. As the second step, information on
the earnout parameters (the dependent variables) is retrieved from Zephyr and
Investegate, a database of company announcements. Third, two alternative uncertainty
measures (the independent variables) are generated for each deal through the use of Orbis
and Datastream. Only those deals for which both independent and at least one dependent
variable could be retrieved remain in the final data sample that is briefly presented at the
end of this chapter. The data sample creation and how each selection step reduces the
sample size can be tracked in table A1 in appendix 2.
5.1 Deal search
The initial sample of M&A deals employing earnouts is created by a customized search
in the Bureau van Dijk database Zephyr. As the first step, a general search strategy
including 8 search criteria was set up. (1) As deal types only mergers and acquisitions
were considered that (2) were completed at the time of the deal search. (3) The acquirer
must be a listed company since these firms are expected to provide more detailed
disclosure of earnout specific information than unlisted firms. (4) Most importantly, the
method of payment has to include an earnout. (5) To ensure a significant principal-agent
relationship between acquirer and target, the final stake owned by the acquirer has to be
at least 50%. (6) To ensure significant potential valuation risk in the deal, the deal value
has to be at least EUR 10mn while the deal value is capped at a maximum of EUR 1bn to
39
exclude extremely dominant large deals. (7) The deal has to be completed between
01.01.2000 and 30.06.2015. (8) Finally, like in previous studies the acquirers had to
originate from the same territory. The search strategy was then run for acquirers from
Germany, Scandinavia, UK and US separately. The individual search strategy results are
shown in figures A2-A5 in appendix 2.
For acquirers from Germany, Zephyr reported only 38 deals while for Scandinavia 73
deals were identified. Both sample sizes do not allow a meaningful empirical analysis and
were therefore not considered. For the UK, Zephyr reported 550 deals while for the US
883 deals were identified. Since potentially each single deal has to be analysed for earnout
specific information, the US sample was considered as too large to handle in the scope of
this thesis. Therefore, the UK sample was chosen as the data basis.
5.2 Collecting data on earnout parameters
Information on the earnout parameters was retrieved from the Zephyr database and the
Investegate database. Investegate is an online database to search for announcements by
UK listed companies, including announcements on M&A events. Therefore, for each deal
identified in Zephyr, the corresponding announcement in Investegate was retrieved and
analysed for parameter-specific information. (See table A2 in appendix 1 for an example)
In order to systemize the search for data on the earnout parameters, the exact definition
of each parameter had to be standardized.
The earnout ratio expresses the portion of the overall deal value that in the databases was
classified to be contingent on performance and paid out as an earnout in the databases
used. The measure is therefore expressed in percentages.
The earnout period was defined to be the overall length of time in which the performance
of the target determines some part of the contingent payment. Consequently, the earnout
period is measured in years. While some earnouts are paid for instance based on the
target’s performance of the last 3 years, other earnouts are paid in three instalments at the
end of each of this three years according to sub-performance goals. Still, in this thesis
both earnouts are considered to have an earnout period of three years since the final
purchase price is ultimately determined at the end of the third year. Moreover, this
simplification was necessary as the information from Zephyr and Investegate did not
allow for a systematic separation of these cases as frequently only the start and end date
of the earnout period were reported.
40
The required performance increase was defined as the required increase in the target’s
performance of the last fiscal year prior to the acquisition in order to reach the
performance goal. This measure is expressed in percentage and is usually expected to be
positive. In order to generate this measure, data on the target’s pre-closing performance
and the performance goal is required. However, the analysis in Zephyr and Investegate
could not reveal this type of information to a large extent. In fact, only for 17 deals data
on the specific performance goal could be retrieved. The creation of a large enough data
sample would therefore had required further analysis in additional databases. Due to time
limitations, this task could not be fulfilled in the scope of this thesis. Consequently, H3
cannot be empirically tested.
The hypotheses regarding the performance measure (H4a-c) required a classification of
performance measures as ‘sales’, ‘income’ and ‘non-financial measures’. This distinction
is also common in previous studies. (Kohers & Ang, 2000; Cain, Denis & Denis, 2011)
In general, measures that are deemed to be easily verifiable and not diluted by decisions
on accounting regulations were determined to be sales-like measure, such as revenues or
share price. To the contrary, measures that are to be found lower at the income statement
or depend on a deeper valuation analysis were determined to be income-like measures,
such as EBITDA or asset valuation. Non-financial measures turned out to be easily
identifiable, such as milestones in R&D projects or operational conditions. The
classification of all measures reported in the deals is shown in table A3 in appendix 2.
As a result of collecting data on the earnout parameters, a total of 40 deals had to be
excluded. For 34 of these deals the search yielded no information on earnout parameters
at all. Further 6 of these deals in fact utilized a different payment method than earnouts
and were therefore wrongly listed in the Zephyr database.
5.3 Generating the explanatory variable “uncertainty”
As the next step, the remaining 510 deals in the data sample were analysed for the
availability of an uncertainty measure. Research agrees to use the target’s standard
deviation in daily returns as a proxy measure for uncertainty in the target’s future
performance. (E.g. Cain, Denis & Denis, 2011; Lukas & Heimann, 2014) The analysis of
the target sample, however, reveals that in 508 deals the target was unlisted, while in 2
deals the target was delisted from the stock after the transaction. Therefore, no data on
the target’s daily standard return on the stock market is available. Consequently, a proxy
41
measure is required. Since the quality of the independent variable is essential to the
empirical analysis, two alternative measures of uncertainty as defined hereafter were
generated to serve as a robustness check. It is a strict requirement that both uncertainty
measures are available for a particular M&A deal in order to remain in the data sample.
Uncertainty of a single proxy target
As a first uncertainty measure, this study utilize the standard deviation of daily returns
for the median firm operating in the same industry as the target, thereby following
previous research. (Cain, Denis & Denis, 2011, p. 158; Lukas & Heimann, 2014, p.486)
The standard deviation of daily returns was calculated over a one year period prior to the
announcement date of each particular deal. Consistent with prior studies, the US SIC
industry codes are used as the industry classification. (E.g. Barbopoulos & Sudarsanam,
2012; Lukas & Heimann, 2014) In the Bureau von Dijk database Orbis, UK-listed
companies with the same SIC code as the target were searched. In order to identify the
median firm, the reported list of companies were ranked according to their cash-flows in
the fiscal year prior to the respective deal. Cash-flows are considered to appropriately
reflect a firm’s business activity.1 However, standardized and comparable financial
figures are only available through Orbis from 2005 onwards. Consequently, only for deals
that were announced in 2006 or later the firms in the same industry could be ranked
according to their prior year’s cash-flows. For all deals announced in 2005 or earlier the
required data was not available. Consequently, 103 deals between 2000 and 2005 had to
be excluded from the sample. Furthermore, 28 deals had to be excluded since no listed
proxy target could be identified in the respective industry.
For the remaining 379 deals, the standard deviation of daily returns for the proxy company
(median firm) in the one year prior to the announcement of the acquisition was calculated.
The required data on daily returns was retrieved through Datastream and the standard
deviation was calculated subsequently. (For detailed formula see appendix 2) In the
subsequent empirical analysis, this measure of uncertainty is referred to as uncert_proxy.
Uncertainty of the proxy target’s industry
An alternative measure of uncertainty was derived in order to carry out a robustness
check. Instead of using the uncertainty of a single proxy target, this alternative approach
1 As an exception, for the insurance industry profit was chosen as the criteria since cash flows are not available.
42
derives an uncertainty measure for the entire industry this proxy target operates in.
Thomson Reuters Datastream offers industry indices for UK based companies.
Consequently, for each proxy target daily return data was retrieved for its particular
industry index one year prior to the respective deal announcement. Again, the uncertainty
measure was finally calculated as the standard deviation in daily returns of these indices.
(For detailed formula see appendix 2) In the subsequent empirical analysis, this measure
of uncertainty is referred to as uncert_ind.
After all, 2 more deals had to be excluded from the sample since the acquirer was listed
primarily at a US stock exchange which was revealed in a further data check. This final
data clean-up reduced the sample to 377 deals.
5.4 Final data sample
As a result of the different data collecting steps described above, the final total sample
consists of 377 earnout deals. For all of these deals both measures of uncertainty are
available. However, not every deal provides information regarding all earnout parameters
simultaneously.
The earnout ratio is known for 353 deals, while 279 deals show information on the earnout
period, and 220 deals for the performance measure. These sample sizes allow for powerful
empirical analysis of H1, H2 and H4. The performance goal, however, could only be
retrieved for 17 deals as described before. Consequently, this sample size does not allow
for a meaningful empirical analysis of the required performance increase and H3 has to
be dropped. Finally, 172 deals comprise data on earnout ratio, earnout period and
performance measure simultaneously. This subsample therefore is labelled the full
information sample. As a further robustness check, all hypotheses are tested on both the
total sample and the full information sample. The accompanying CD to this thesis
includes an overview of both samples with all required deal specific data.
43
6 Methodology
6.1 Control variables for information asymmetry
The theoretical model at the centre of this thesis describes the earnout as a solution to
problems of information asymmetry. Furthermore, this approach models earnouts as real
options and consequently derives hypotheses on the design of its parameters according to
option pricing methodology. The variable of uncertainty therefore is the pivotal
determinant of earnout design since all earnout parameters are expected to be adjusted
according to the degree of uncertainty in the deal.
However, from the review of theory and empirical evidence on earnout design in chapter
2 we know that alternative explanations of the earnout design exist. Although lacking
consent, most of the scholars argue that not only the choice to use earnouts in the first
place but also the design of its parameters is determined by indicators of information
asymmetry. (E.g. Kohers & Ang, 2000; Barbopoulos & Sudarsanam, 2012) Therefore, it
is of great interest to not only test the hypotheses H1-H4 as derived from the option
pricing model. Rather, we would like to control for the opposing view that information
asymmetry indicators in fact directly determine the earnout design, thereby ignoring the
‘likelihood-concept’.
Therefore, the regression models developed to test H1-H4 as described in the subsequent
chapters are supplemented by common indicators of information asymmetry. Referring
back to chapter 2, research agrees that private targets, targets from the high-tech and
service industry, cross-industry deals and cross-country deals are indicators of high
information asymmetry between the acquirer and the target. For simplicity reasons, a
target is considered private if Zephyr reports it as being unlisted. Furthermore, research
argues that larger and older acquirers better cope with information asymmetries in M&A
deals and that the deal size in turn affects the magnitude of the valuation risk present in a
transaction. Thus, also acquirer’s size, acquirer’s age and deal size act as control
variables.
Table 3 below shows the control variables, the mnemonic used in the regression models,
the variable type, the exact definition and the database from which it is sourced.
44
Table 3: Definition of control variables for information asymmetry
Variable Mnemonic Type Definition Database
Private target private Dummy 1 = unlisted;
0 = listed
Zephyr
Cross-industry
deal
cross_ind Dummy 1 = different two-digit SIC
code for acquirer and target
0 = same two-digit SIC code
Zephyr
Cross-country
deal
cross_count Dummy 1 = different country code
for acquirer and target
0 = same country code
Zephyr
High-tech target high_tech Dummy 1 = SIC code matches high-
tech classification2
0 = no match
Zephyr
Service industry
target
serv_ind Dummy 1 = SIC code matches
service industry
classification3
0 = no match
Zephyr
Acquirer’s age acqu_age Continous Number of years between
acquirer’s incorporation and
deal’s announcement year
Zephyr
Acquirer’s size acqu_size Continous Acquirer’s market value of
equity 4 weeks prior to deal
announcement in GBP mn
Datastream
Deal size deal_size Continous In GBP mn Zephyr
Source: Author’s table
6.2 Regression models on earnout parameters
This chapter explains and explicitly states the regression models used to analyse the
hypotheses H1, H2 and H4a-c. Consistent to previous research, the earnout ratio is
examined by means of a Tobit regression, the earnout period by means of an OLS
regression and the choice of the performance measure is structured as a logit model. The
regression models and test statistics are all run in the econometric software Eviews. As
mentioned in the chapter on data, the methodology includes two robustness checks.
First, each model is run for the two uncertainty measures (uncert_proxy and uncert_ind)
separately. Since the hypotheses state uncertainty to be the central determinant of the
earnout design, ensuring the validity of this measure is of high importance. The regression
models stated below for simplicity reasons only show the case for the uncertainty measure
uncert_proxy. Second, each hypothesis is tested on the total sample and the full
information sample. As the latter might possess higher data quality compared to deals for
which the earnout parameters are only partly known.
2 High-tech industry classification by SIC codes according to a meta-study on optimal high-tech classifications by Kile & Phillips (2009). See table A4 in appendix 3 for details. 3 Own classification according to SIC code logic. See table A5 in appendix 3 for details.
45
The significance of the estimated coefficients of the explanatory variables will be tested
by simple t-tests. As both the quality of the regression models to provide effective
estimators and the power of the t-tests depend on normality assumptions, the common
Gauss-Markov assumptions (A1-A4) are tested to hold: (Verbeek, 2012, p. 18)
A1: Error terms (𝜀𝑖) have mean zero
A2: All error terms are independent of all explanatory variables
A3: All error terms have the same variance (homoscedasticity)
A4: The error terms are mutually uncorrelated (no autocorrelation)
As described along with the empirical results, the models and test statistics are adjusted
if one assumption is found to be violated. For details on the t-test see appendix 3.
6.2.1 Tobit regression model of the earnout ratio
Following previous studies, a Tobit regression is used to examine the determinants of the
earnout ratio (EOR) by means of maximum likelihood estimation. (Kohers & Ang, 2000;
Cain, Denis & Denis, 2011) The use of a Tobit model accounts for the fact that the earnout
ratio is a continuous variable but constrained to a range from 0 to 1. (Verbeek, 2012, p.
238) Consequently, all values below 0 and above 1 are mapped to 0.
The Tobit model tries to explain the earnout ratio for each deal i (𝐸𝑂𝑅𝑖) by a regression
on a constant (𝛼), the uncertainty measure (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) and the control variables.
For the purpose of an easier interpretation of the resulting coefficients, the continuous
variables 𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖, 𝑎𝑐𝑞𝑢_𝑎𝑔𝑒𝑖, 𝑎𝑐𝑞𝑢_𝑠𝑖𝑧𝑒𝑖 and 𝑑𝑒𝑎𝑙_𝑠𝑖𝑧𝑒𝑖 are transformed to
logarithmic variables. The model therefore indicates the absolute change in the earnout
ratio caused by a ceteris paribus percentage change of these log variables. (Verbeek, 2012,
p. 75) The regression model to be run in Eviews therefore is:
𝐸𝑂𝑅𝑖 = 𝛼 + 𝛽1 ∗ log (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) + 𝛽2 ∗ 𝑝𝑟𝑖𝑣𝑎𝑡𝑒𝑖 + 𝛽3 ∗ 𝑐𝑟𝑜𝑠𝑠_𝑖𝑛𝑑𝑖 + 𝛽4
∗ 𝑐𝑟𝑜𝑠𝑠_𝑐𝑜𝑢𝑛𝑡𝑖 + 𝛽5 ∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 + 𝛽6 ∗ 𝑠𝑒𝑟𝑣_𝑖𝑛𝑑𝑖 + 𝛽7
∗ log (𝑎𝑐𝑞𝑢_𝑎𝑔𝑒𝑖) + 𝛽8 ∗ log (𝑎𝑐𝑞𝑢_𝑠𝑖𝑧𝑒𝑖) + 𝛽9 ∗ log (𝑑𝑒𝑎𝑙_𝑠𝑖𝑧𝑒𝑖) + 𝜀𝑖
with 𝐸𝑂𝑅𝑖 = 0 for all 𝐸𝑂𝑅𝑖 ≤ 0 and ≥ 1
46
6.2.2 OLS regression model of the earnout period
The determinants of the earnout period (EOPer) are examined by means of a cross-
sectional linear regression employing the method of ordinary least squares (OLS) since
this continuous variable is not subject to any upper limit and cannot be negative. The OLS
regression model tries to explain the earnout period for each deal i (𝐸𝑂𝑃𝑒𝑟𝑖) by a
regression on a constant (𝛼), the uncertainty measure (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) and the control
variables. Again, for the purpose of an easier interpretation of the resulting coefficients,
the continuous variables 𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖, 𝑎𝑐𝑞𝑢_𝑎𝑔𝑒𝑖, 𝑎𝑐𝑞𝑢_𝑠𝑖𝑧𝑒𝑖 and 𝑑𝑒𝑎𝑙_𝑠𝑖𝑧𝑒𝑖 are
transformed to logarithmic variables. The regression model to be run in Eviews therefore
is:
𝐸𝑂𝑃𝑒𝑟𝑖 = 𝛼 + 𝛽1 ∗ log (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) + 𝛽2 ∗ 𝑝𝑟𝑖𝑣𝑎𝑡𝑒𝑖 + 𝛽3 ∗ 𝑐𝑟𝑜𝑠𝑠_𝑖𝑛𝑑𝑖 + 𝛽4
∗ 𝑐𝑟𝑜𝑠𝑠_𝑐𝑜𝑢𝑛𝑡𝑖 + 𝛽5 ∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 + 𝛽6 ∗ 𝑠𝑒𝑟𝑣_𝑖𝑛𝑑𝑖 + 𝛽7
∗ log (𝑎𝑐𝑞𝑢_𝑎𝑔𝑒𝑖) + 𝛽8 ∗ log (𝑎𝑐𝑞𝑢_𝑠𝑖𝑧𝑒𝑖) + 𝛽9 ∗ log (𝑑𝑒𝑎𝑙_𝑠𝑖𝑧𝑒𝑖) + 𝜀𝑖
6.2.3 Binary choice model of the performance measure
H4 a-c seek to explain the probability of the performance measure to be either a measure
of sales, a measure of income, or a non-financial measure. The pivotal explanatory
variables again are the measures of uncertainty. The dependent variable, i.e. the use of
the performance measure in question, is therefore limited to the two possible outcomes
“yes” or “no”, i.e. two discrete alternatives. Therefore, a binary choice model is required
to model these dependencies. Probit and logit models are most common in studies to
examine these types of relationships. (Verbeek, 2012, p. 208) While the probit model
assumes a standard normal distribution, the logit model assumes a standard logistic
distribution of the residuals. Therefore, in case the probit model results report non-
normality characteristics such as a high skewness in its residuals, the logit model is the
more appropriate approach. (Verbeek 2012, p. 208) As described later in the chapter on
empirical results and as shown on the accompanying CD, the probit model reports high
skewness and therefore a logit model was chosen for H4a-c as depicted below.
To test H4a, the dependent binary variable is defined as 𝑠𝑎𝑙𝑒𝑠𝑖 with the two discrete
alternative outcomes:
47
𝑠𝑎𝑙𝑒𝑠𝑖 = 1 , if the performance measure is a measure of sales
𝑠𝑎𝑙𝑒𝑠𝑖 = 0 , if the performance measure is not a measure of sales
The logit model tries to explain the probability of 𝑠𝑎𝑙𝑒𝑠𝑖 to be the performance measure
in a deal by a constant (𝛼), the uncertainty measure (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) and the control
variables. The model to be run in Eviews therefore is:
logit(𝑠𝑎𝑙𝑒𝑠𝑖 = 1) = 𝐿(𝛼 + 𝛽1 ∗ log (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) + 𝛽2 ∗ 𝑝𝑟𝑖𝑣𝑎𝑡𝑒𝑖 + 𝛽3 ∗
𝑐𝑟𝑜𝑠𝑠_𝑖𝑛𝑑𝑖 + 𝛽4 ∗ 𝑐𝑟𝑜𝑠𝑠_𝑐𝑜𝑢𝑛𝑡𝑖 + 𝛽5 ∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 + 𝛽6 ∗
𝑠𝑒𝑟𝑣_𝑖𝑛𝑑𝑖 + 𝛽7 ∗ log (𝑎𝑐𝑞𝑢_𝑎𝑔𝑒𝑖) + 𝛽8 ∗ log (𝑎𝑐𝑞𝑢_𝑠𝑖𝑧𝑒𝑖) +
𝛽9 ∗ log (𝑑𝑒𝑎𝑙_𝑠𝑖𝑧𝑒𝑖) + 𝜀𝑖)
with L being the standard logistic distribution function
To test H4b, the dependent binary variable is defined as 𝑖𝑛𝑐𝑜𝑚𝑒𝑖 with the two discrete
alternative outcomes:
𝑖𝑛𝑐𝑜𝑚𝑒𝑖 = 1 , if the performance measure is a measure of income
𝑖𝑛𝑐𝑜𝑚𝑒𝑖 = 0 , if the performance measure is not a measure of income
The logit model tries to explain the probability of 𝑖𝑛𝑐𝑜𝑚𝑒𝑖 to be the performance measure
in a deal by a constant (𝛼), the uncertainty measure (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) and the control
variables. The model to be run in Eviews therefore is:
logit(𝑖𝑛𝑐𝑜𝑚𝑒𝑖 = 1) = 𝐿(𝛼 + 𝛽1 ∗ log (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) + 𝛽2 ∗ 𝑝𝑟𝑖𝑣𝑎𝑡𝑒𝑖 + 𝛽3 ∗
𝑐𝑟𝑜𝑠𝑠_𝑖𝑛𝑑𝑖 + 𝛽4 ∗ 𝑐𝑟𝑜𝑠𝑠_𝑐𝑜𝑢𝑛𝑡𝑖 + 𝛽5 ∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 + 𝛽6 ∗
𝑠𝑒𝑟𝑣_𝑖𝑛𝑑𝑖 + 𝛽7 ∗ log (𝑎𝑐𝑞𝑢_𝑎𝑔𝑒𝑖) + 𝛽8 ∗ log (𝑎𝑐𝑞𝑢_𝑠𝑖𝑧𝑒𝑖) +
𝛽9 ∗ log (𝑑𝑒𝑎𝑙_𝑠𝑖𝑧𝑒𝑖) + 𝜀𝑖)
with L being the standard logistic distribution function
To test H4c, the dependent binary variable is defined as 𝑛𝑜𝑛_𝑓𝑖𝑛𝑖. (The determinants of
a measure of sales are already tested in H4a) The two discrete alternative outcomes for
𝑛𝑜𝑛_𝑓𝑖𝑛𝑖 are:
𝑛𝑜𝑛_𝑓𝑖𝑛𝑖 = 1 , if the performance measure is a non-financial measure
𝑛𝑜𝑛_𝑓𝑖𝑛𝑖= 0 , if the performance measure is not a non-financial measure
48
The logit model tries to explain the probability of 𝑛𝑜𝑛_𝑓𝑖𝑛𝑖 to be the performance
measure in a deal by a constant (𝛼), the uncertainty measure (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) and the
control variables. The model to be run in Eviews therefore is:
logit(𝑖𝑛𝑐𝑜𝑚𝑒𝑖 = 1) = 𝐿(𝛼 + 𝛽1 ∗ log (𝑢𝑛𝑐𝑒𝑟𝑡_𝑝𝑟𝑜𝑥𝑦𝑖) + 𝛽2 ∗ 𝑝𝑟𝑖𝑣𝑎𝑡𝑒𝑖 + 𝛽3 ∗
𝑐𝑟𝑜𝑠𝑠_𝑖𝑛𝑑𝑖 + 𝛽4 ∗ 𝑐𝑟𝑜𝑠𝑠_𝑐𝑜𝑢𝑛𝑡𝑖 + 𝛽5 ∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 + 𝛽6 ∗
𝑠𝑒𝑟𝑣_𝑖𝑛𝑑𝑖 + 𝛽7 ∗ log (𝑎𝑐𝑞𝑢_𝑎𝑔𝑒𝑖) + 𝛽8 ∗ log (𝑎𝑐𝑞𝑢_𝑠𝑖𝑧𝑒𝑖) +
𝛽9 ∗ log (𝑑𝑒𝑎𝑙_𝑠𝑖𝑧𝑒𝑖) + 𝜀𝑖)
with L being the standard logistic distribution function
7 Empirical results
This chapter presents the results of the empirical analysis. First, the most interesting
descriptive statistics on the data sample are presented. In the second part of the chapter,
the results from the regression models on the determinants of the earnout parameters are
presented in detail.
7.1 Descriptive statistics
This section presents descriptive statistics on the final data samples on which the
hypotheses were tested. It thereby serves as a reference point to conclude on differences
between the total sample and the full information sample which is a robustness check.
Also, descriptive statistics allow to compare the data sample of this thesis with previous
empirical studies on earnouts. This helps to conclude on consistency with previous
research or on possible limitations to which degree the results of the thesis can be
integrated into the current state of research.
Table 4 below compiles descriptive statistics on the total sample as compared to the full
information sample for which all relevant earnout parameters are known. The observed
variables are presented as their mnemonic and clustered into the categories “uncertainty
measures”, “earnout parameter”, “deal characteristics” and “acquirer characteristics”.
49
Table 4: Descriptive statistics on data sample
Variables Total sample Full information sample
Number of deals:
377
172
Un
certa
inty
mea
sure
s
uncert_proxy: min 0.0024 0.0024
Ø 0.0251 0.0251
max 0.1050 0.1050
rel. dispersion 59% 67%
uncert_ind: min 0.0049 0.0049
Ø 0.0131 0.0129
max 0.0424 0.0373
rel. dispersion 37% 38%
Ea
rno
ut
pa
ram
eter
s
EOR (in %): min 1% 2%
Ø 38% 40%
max 100% 100%
rel. dispersion 60% 55%
EOPer (in years): min 0.25 0.25
Ø 2.5 2.6
max 8 8
rel. dispersion 53% 55%
Performance measure
sales N 45 34
income N 146 123
non-fin N 48 28
Dea
l ch
ara
cter
isti
cs deal_size: min 6.8 6.8
(in GBP mn) Ø 44.4 47.2
max 750.5 679.4
cross_ind N / (in % of total) 130 (34%) 67 (39%)
cross_count N / (in % of total) 183 (49%) 88 (51%)
private N / (in % of total) 375 (99%) 171 (99%)
high_tech N / (in % of total) 129 (34%) 55 (32%)
serv_ind N / (in % of total) 198 (53%) 93 (54%)
Acq
uir
er c
ha
ract
eris
tics
acqu_age (in years)
min 1 1
Ø 29 30
max 169 169
acqu_size (in GBP mn)
min 2 2
Ø 1,661 821
max 84,597 10,989
Source: Author’s analysis
50
For the total sample, the statistics report a significant difference between the two
alternative uncertainty measures. The relative dispersion coefficient shows that the proxy
target’s standard deviation is significantly stronger dispersed around its mean than the
measure of the target’s industry in relative terms.4 Clearly, this result reflects that the
uncertainty of an industry is calculated on a larger sample of companies what counteracts
more extreme measures that might be found if solely looking at single proxy targets.
Furthermore, the sample includes earnout ratios across the entire possible range of values
from 1% to 100% and reports a mean of 38%. This result is perfectly in line with a recent
study from the UK (38%) by Barbopoulos & Sudarsanam (2012, p. 685) and close to
figures for the US (33%) reported by Cain, Denis & Denis (2011, p. 155). Consequently,
there is evidence that the earnout premium on average represents a significant part of the
overall maximum acquisition price.
The earnout period ranges from 0.25 years to a maximum of 8 years, reporting a mean of
2.5 years. This finding is consistent with previous studies from the US and UK that also
reported an earnout period of around 2.5 years on average. (See Barbopoulos &
Sudarsanam, 2012, p. 683; Cain, Denis & Denis, 2011, p. 157) The relative dispersion
proves considerable variability for the earnout ratio and the earnout period within both
samples. We can therefore conclude that the design of the two parameters is strongly
adjusted to each deal’s requirements.
Income is by far the most frequently used performance measure, while sales and non-
financial measures occur with almost equal frequency. The study by Cain, Denis & Denis
(2011) supports that income measures are most often used, however they report a higher
usage of sales than non-financial measure indicating a difference between US and UK
based samples.
The deal characteristics reveal a rather small average deal size of GBP 44mn. This finding
is to some extent backed by the study of Kohers & Ang (2000) for US earnout deals that
reports an average deal size of USD 44mn. Relatively small deals are not surprising taking
into account that 99% of the target companies in the deal sample are private targets.
Consequently, we will exclude the variable private from the subsequent regression
4 The relative dispersion coefficient shows the extent of variability in relation to the mean of the population and is defined as the ratio of standard deviation to the mean (σ/µ). It is a measure of dispersion in a sample that is independent from the sample’s scale and allows for comparison across different scaled samples.
51
analysis as it describes the almost entire sample of observations. Similarly, in their
study of UK deals using earnouts, Barbopoulos & Sudarsanam (2012, p. 683) report that
99% of the deals involve non-public targets and same evidence is available for the US by
Cain, Denis & Denis (2011).5 Consequently, we find strong support for the dominant
hypothesis in literature that earnouts are best suited for private targets.
Also, the high percentage of cross-industry deals (34%), cross-country deals (49%) and
targets from the high-tech (34%) and service industry (53%) gives support to the
argument that earnouts are favourably chosen in situations that indicate high information
asymmetry. Finally, the acquirer characteristics report that earnouts are used by
established (av. age of 29 years) and rather large acquirers (av. GBP 1.6bn market
capitalization).
For the full information sample, the statistics do not have to be reviewed in detail since
they strongly resemble the ones from the total sample. As an exception, the acquirer’s
size differs significantly. The statistic reports acquirers in the full info sample that are on
average only half as large in terms of market capitalization than in the total sample (GBP
1.6bn vs. GBP 0.8bn). Taking a closer look at the data sample reveals that these results
are driven by the top 7 deals in terms of deal size in the total sample that are all not
included in the full info sample due to missing information on the earnout parameters.
Exclusive of these 7 deals, the average acquirer’s size of the total sample is GBP 1.06bn
and therefore much closer to the mean of GBO 0.8bn in the full info sample.
Most importantly, we can conclude that the statistics regarding the earnout parameters
are in line with previous studies. The data collection and the parameter definitions
therefore seem consistent with research so far. Furthermore, the comparison of the two
samples considered here reveals high similarity. The full information sample therefore is
representative for the total sample and should serve as a good robustness check.
7.2 Results from the regression models
This chapter finally presents and interprets the empirical results from the regression
models. This section therefore aims to identify the determinants of the earnout ratio (H1),
the earnout period (H2) and the performance measure (H4a-c). As described before, each
5 The authors differentiate non-public targets as private targets or subsidiaries, however both studies are consistent with only finding 1% of public targets.
52
hypotheses is tested on both the total sample and the full information sample as a
robustness check. Furthermore, for each hypothesis a regression model is run including
the uncert_proxy variable and another model is run including the uncert_ind variable
separately. The first is referred to as Model 1 while the latter is referred to as Model 2.
Both models include all control variables as defined before. Since the two uncertainty
measures can be expected to be highly correlated, a separate analysis avoids the problem
of heteroscedasticity in the explanatory variables to this extent. Also, a separate
examination allows for a better comparison of the explanatory power of both models.
Detailed results and analysis steps to test the normality assumptions can be found on the
accompanying CD.
7.2.1 Determinants of the earnout ratio
Before interpreting the results of the regression model, the Gauss-Markov assumptions
A1-A4 were tested for both models in both samples. Autocorrelation is not expected to
be an issue since no time series data is included in the model. Also, there is no reason to
expect that the error terms are dependent on the explanatory variables. However,
heteroscedasticity can be expected to be present in the data sample throughout the entire
analysis in several ways.
First, the high-tech and service industry variable are overlapping in terms of SIC codes
and therefore show a correlation. Second, the acquirer’s age and the acquirer’s size in
terms of market capitalization can be expected to correlate as well. Third, we can expect
larger acquirers to close deals of larger size. In general, heteroscedasticity is frequently
occurring in cross-sectional regressions including many dummy variables, as it is the case
here. (Verbeek, 2012, p. 98) In order to deal with this issue, the Tobit model is adjusted
for heteroskedastic-consistent White standard errors. (White, 1980)
Finally, the sample was analysed for normally distributed residuals. For a sample to be
normally distributed its error terms should be mean zero. For a continuously observed
variable, such as the earnout ratio in our case, testing for normality should also include
the check for skewness and excess kurtosis. (Verbeek, 2012, p. 202) The statistics given
at the bottom of table 5 prove that the residuals are very close to mean zero for both
models in both samples. Also, the Kurtosis measure is close to 3 indicating a close to
normal distribution. (Verbeek, 2012, p. 202) The skewness measure is slightly positive,
implying that the residuals are slightly right-skewed and not perfectly symmetrical
53
distributed around zero. However, all in all we accept that the residuals are close to normal
distributed. Also, relying on the law of large numbers the distribution is considered to be
asymptotically normal distributed. (Verbeek, 2012, p. 34) The regression model and the
t-statistic are therefore considered to have high power.
Table 5 presents the results of the OLS regression for H1 for both model 1 and 2 based
on the total sample and the full information sample separately. For each explanatory
variable the respective coefficient and t-ratio is reported. Coefficients that are significant
at the common 1%, 5% or 10% levels are marked by asterisk (***), (**) or (*)
respectively. The levels indicate the marginal significance level for which the null
hypothesis can be rejected. (Verbeek, 2012, p. 31)
Table 5: Results from Tobit model on determinants of earnout ratio
Source: Author’s analysis
First of all, we should compare the explanatory power of the different models to conclude,
which one reports the most reliable and valid results. There is no consensus in academia
on a goodness of fit measure for Tobit models like the R² measure for OLS regressions.
(Veall & Zimmermann, 1994) However, we can rely on the Akaike information criterion
(AIC) and the Schwartz information criterion (BIC) as means for model selection. Models
with a lower AIC or BIC are preferable. (Verbeek, 2012, p. 66) Both measures report
lower values for the total sample and therefore consistently suggest that the models of the
total samples are better specified than those in the full information sample. More
54
specifically, model 2 in the total sample reports the most negative values that implies the
best available specification. We therefore tend to rely the most on the reported
coefficients therein.
From H1, we expect a positive effect of the uncertainty measures on the earnout ratio. In
the total sample, the uncert_proxy coefficient shows an unpredicted slightly negative
effect on the earnout ratio. In contrast to that, in model 2 the alternative measure of
uncertainty uncert_ind shows the predicted positive coefficient. Clearly, these results
fundamentally differ. However, since model 2 reports preferable AIC and BIC values and
since the t-ratio for uncert_ind implies a considerably higher significance than
uncert_proxy we should really rely more on the uncert_ind measure. Consequently, as
predicted in H1 an increasing uncertainty would also increase the earnout ratio. A 10%
larger uncertainty, ceteris paribus, would therefore raise the earnout ratio by 0.004 points.
Besides this increase being very low, the coefficient for uncert_ind is still not significant
for standard significance levels. The insignificance of the uncertainty measures is
consistent also across samples. All in all, we cannot find strong evidence for uncertainty
to be a significant determinant of the earnout ratio and therefore have to reject H1.
The indicators of information asymmetry cross_ind, cross_count and serv_ind all report
positive coefficients for model 2 in the total sample while only serv_ind is highly
significant at a 1% level. Ceteris paribus, the target being from the service industry
therefore increases the earnout ratio by 0.08 points. Basically, the earnout ratio is
therefore found to be higher in situations that imply a high degree of information
asymmetry between acquirer and target. This result suggests that in situations that imply
adverse selection and agency problems to be severe, the contingent portion is increased
to provide a stronger signalling tool on the one hand and a stronger management retention
incentive on the other hand. Surprisingly, the high-tech variable shows a negative
coefficient which runs counter this argumentation. Still, its effect is highly insignificant
and does not change the overall conclusion.
The coefficients acqu_age, acqu_size and deal_size all show slightly negative
coefficients. Only the coefficient for the acquirer’s size is close to a 10% significance
level, suggesting that larger acquirers tend to use smaller earnout ratios. This might be
due to the fact that larger acquirers are less exposed to the valuation risk and therefore
face less need to shift the risk towards the target.
55
A cross-sample comparison reveals that both models in both samples yield consistent
results in terms of the sign and to a large extent in terms of the significance of the
coefficients. Therefore, the results from the Tobit regression model are robust across
samples and indicate an unbiased data selection process.
7.2.2 Determinants of the earnout period
With regards to test the normality assumptions, autocorrelation and the independence of
the error terms was not considered an issue. However, the Breusch-Pagan-Godfrey test
for heteroscedasticity reported heteroscedasticity for both models in both samples. This
issue was dealt with by adjusting for heteroskedastic-consistent White standard errors. In
contrast to the model for H1, the residuals for the total sample this time showed excess
Kurtosis and were significantly right-skewed. The data sample therefore was checked for
outliers. In order to arrive at approximate normal distribution properties to ensure high
power of the test statistic and consistency of the OLS estimators, the six most positive
outlying observations were removed from the sample. To ensure consistency between
the two samples, the same outliers were removed from the full information sample as
well. Detailed results of the normality tests are shown on the accompanying CD.
Table 6 presents the coefficients and t-ratios for each explanatory variable of the
regression. Significant coefficients are market by asterisk.
The explanatory power of the models set up to test H2 differs significantly, with a
considerably higher adjusted R² measures for the full information sample and model 2 in
particular. We therefore tend to rely more on the results from this regression, noting,
however, that the fit of the linear regressions to model the dependent variable is still low.
56
Table 6: Results from OLS model on determinants of earnout period
Source: Author’s analysis
According to H2, we would expect a negative coefficient for the uncertainty measures
within the regression results. Model 1 for the full information sample returns the expected
result for uncert_proxy, however the coefficient is highly insignificant. Model 2 in the
sample and both regressions for the total sample to the contrary report a positive
coefficient for the respective uncertainty measure. Due to these inconsistent results and
since the coefficient of uncert_proxy changes its sign between the two samples, a negative
correlation between earnout period and uncertainty remains highly questionable. The
most significant result presents uncert_ind for the total sample, but still it cannot be
considered a valid result in terms of common significance levels. Again, uncert_ind yields
the more significant coefficients implying a better specification to measure uncertainty
than uncert_proxy. After all, we cannot find strong evidence that would support a negative
impact of uncertainty on the earnout period and consequently have to reject H2.
According to the information asymmetry theory, a longer earnout period might resolve
the information asymmetries over time which is expected valuable especially in case of
targets with high growth opportunities. However, the results for the high_tech variable,
which is supposed to indicate high-growth opportunities, runs counter this rationale since
its coefficient suggests a negative but insignificant effect for the full information sample.
To the contrary, there is strong evidence for the variable cross_count to be a significant
57
and strongly positive determinant of the earnout period. While in the total sample it is
significant on a 5% level, the more powerful full information sample reports a
significance level of 1%. The coefficient shows that in case of a cross-country deal, the
earnout period increases by 0.6 years. Positive coefficients are also found for cross_ind
and serv_ind that, however, lack significance. We might still conclude in terms of
information asymmetry theory, that acquirers desire longer earnout periods in these
situations to resolve cultural differences in cross-country deals, asymmetries in market
knowledge in cross-industry deals or simply to incentivize target’s key human capital to
remain longer with the firm such as for target from the service industry that typically
heavily rely on human capital.
Furthermore, there is some evidence that the acquirer’s size has a negative impact on the
earnout period. No study so far offers a possible explanation to this finding. However,
larger acquirers are expected to cope better with information asymmetries due to their
longer experience in M&A deals and better access to information on the target.
Consequently, they might have less need for additional information that is revealed during
the earnout period in order to mitigate valuation risks. Still, this rationale is speculative.
7.2.3 Determinants of the performance measure
As the first step, the distribution of the residuals in the H4a-c models was tested in order
to conclude on the fit of the probit and logit model respectively. Running the binary
choice models reported non-normal statistics of the residuals. (See CD) Therefore, the
logit model was opted for as the estimation model with better fit to test H4a-H4c. Also,
since heteroscedasticity among the explanatory variables should be expected, the logit
model was adjusted by White standard errors. In general, the coefficients from logit
models can only be interpreted as how they impact the probability of choosing the
performance measure in question by looking at their signs and significance levels. A
positive coefficient increases this probability, while a negative one lowers it accordingly.
However, the exact size of the coefficient cannot be interpreted meaningfully. (Verbeek,
2012, p. 208)
Determinants for the choice to use sales as performance measure
Table 7 presents the coefficients and t-ratios that result from examining the probability
that sales is used as the performance measure in a logit model:
58
Table 7: Results from logit model on determinants of sales measure
Source: Author’s analysis
First of all, the explanatory power of the models is compared by means of Mc Fadden R²,
which is a goodness-of-fit measure especially adjusted to binary choice models.
(Verbeek, 2012, p. 212) The models of the total sample report the highest explanatory
power, implying that up to 9% of the variation in the performance measure sample is
explained by the model. The subsequent analysis is therefore focused on the results from
the total sample.
From H4a, uncertainty is expected to show a positive impact on the probability that sales
is used as the performance measure. Both uncertainty coefficients show an unpredicted
negative impact on the probability of choosing sales as the performance measure. Still,
there is no evidence for a significant impact of uncertainty on the probability that sales is
chosen as the performance measure. This finding is also consistent across both samples
and both models. Therefore, both the signs of the coefficients and their significance levels
tell us to reject H4a.
There is strong evidence that in deals with high-tech targets the performance measure is
more likely sales as indicated by the significant positive coefficient for high_tech. This
finding is consistent across models and samples. For these firms that typically carry high
growth opportunities, a measure of sales might serve best as a driver measure of future
profitability and thereby reveals the most valuable information for targets of this kind.
Consequently, we cannot reject H4c in this regard.
59
To the contrary, the coefficients for cross_ind and serv_ind suggest a negative impact on
the probability of the performance measure being sales. Although their coefficients are
not significant up to a 10% level in any of the reported models, these results run counter
the general assumption that in case of high information asymmetry the performance
measure is more likely to be an easier verifiable sales measure.
Moreover, there is some evidence for acqu_size and deal_size being significant
determinants of the choice for sales. While the acquirer’s size is negatively related, the
size of the deal increases the probability. Literature suggests that larger acquirers better
cope with information asymmetry what might indicate that these firms not have to rely
on easily verifiable measure such as sales. Regarding the deal size, however, the theory
on information asymmetry does not offer a possible explanation.
Determinants for the choice to use income as performance measure
Table 8 presents the coefficients and t-ratios that result from examining the probability
that income is used as the performance measure in a logit model:
Table 8: Results from logit model on determinants of income measure
Source: Author’s analysis
Again, the models within the total sample carry higher explanatory power as indicated by
Mc Fadden R², while model 1 and model 2 perform almost equally.
According to H4b, uncertainty is expected to negatively affect the probability that income
is used as a performance measure. The coefficients for uncert_proxy and uncert_ind for
60
the total sample reveal a positive relationship between uncertainty and the choice of
income as the performance measure. This finding runs counter the expectations that
increasing uncertainty would make it less probable for the performance measure to be
income. Also, the results are at odds with previous empirical evidence that income is
chosen for targets from less volatile industries. (Cain, Denis & Denis, 2011) Again, this
inconsistency to previous studies might indicate a misspecification in the uncertainty
measures. All in all, we strongly have to reject H4b.
The coefficients for cross_count and high_tech both suggest a negative impact on the
probability that income is used. In case of cross-country deals this result is plausible as a
measure of income is potentially affected by unfamiliar accounting standard in other
countries and not easily verifiable for the acquirer. In case of high-tech targets this result
is intuitive as we have shown that for high growth-targets a driver measure of future
profitability such as sales is favourable.
The coefficients for cross_ind and serv_ind run counter the argumentation that income is
less likely a measure for deals with high information asymmetry. However, their
coefficients are insignificant and therefore do not change the overall conclusion.
Finally, larger acquirers tend not only to use less likely sales but also less likely income.
This result contradicts the possible explanation given for H4a and does not allow for a
consistent conclusion.
Determinants for the choice to use a non-financial performance measure
Table 9 presents the coefficients and t-ratios that result from regressing the probability
that a non-financial performance measure is used in a logit model.
The models of both samples report very similar explanatory power in terms of Mc Fadden
R². To be consistent throughout the analysis of H4a-c, the total sample is focused on.
Again, the uncertainty measures have no significant impact. However, the focal point in
the analysis of the results is the question whether for targets with high growth
opportunities the performance measure is more likely to be non-financial.
61
Table 9: Results from logit model on determinants of non-financial measure
Source: Author’s analysis
As predicted in H4c, the indicator for targets with high growth opportunities high_tech
shows a positive coefficient. Consequently, the results suggest that for targets that heavily
rely on reaching milestones in development projects and that require high investments for
future profits, a non-financial measure is the best driver measure for future profit and
therefore most appropriate. Still, the coefficient is not very significant and therefore we
have to reject H4c in this regard. As shown before, there is significant evidence though
that for high-tech targets a measure of sales is more likely to be used. Consequently, the
empirical results do suggest that for these targets an easy verifiable measure like sales or
a non-financial driver measure is the appropriate choice since profitability measures
would only hinder current growth targets.
Furthermore, for cross-country deals, cross-industry deals and targets from the service-
industry the probability that non-financial measures are opted for increases. Although
none of these coefficients is significant, it indicates that in situations of high information
asymmetry the non-financial measure conveys more valuable information.
62
8 Evaluation and avenues for further research
This chapter first evaluates the most interesting empirical findings in the context of the
theoretical model and its assumptions and compares the results to previous studies.
Furthermore, it points out the limitations the thesis faces in terms of literature, data and
methodology. Finally, the last subchapter is devoted to possible avenues for further
research.
8.1 Discussion of the empirical results
In order to discuss the empirical results of this thesis, we should refer back to the advanced
model on earnout design. The quality of the model can be concluded on by discussing the
implications of the findings for the model’s assumptions.
Basically, the advanced model on earnout design expects two possible dynamics if the
likelihood for an earnout premium to be paid increases due to increased uncertainty. To
ensure the effectiveness of the earnout contract as an incentive and signalling tool, the
acquirer could on the one hand choose to raise the earnout ratio in order to stronger
incentivize the target for cooperative behaviour and sort low from high quality targets.
Referring back to figure 6, this was labelled dynamic A. On the other hand, the acquirer
could also choose to shorten the earnout period (B1) or to increase the performance goal
(B2) in order to reduce the increased likelihood. While results for the earnout ratio
basically test dynamic A, the results for the earnout period and the required performance
increase would consequently test for the advanced model’s dynamics B1 and B2.
For the earnout ratio, the empirical analysis reports the predicted positive signs in the
coefficients for the better specified uncertainty measure. However, its impact on the
earnout ratio seems insignificant. Cain, Denis & Denis (2011) report uncertainty to be a
strong and significant driver of the earnout ratio. Therefore, in context of recent literature
there is indeed some evidence that acquirers tend to raise the earnout ratio as a reaction
to increased agency and adverse selection problems caused by an increased likelihood of
the earnout to end up in the money. This conclusion is in support of the dynamic A that
was developed in the original model by Lukas, Reuer & Welling (2012). Furthermore,
the control variables in the thesis’ analysis show strong evidence that the earnout ratio is
higher in situations of severe information asymmetry. Especially for targets from the
service industry the ratio tends to increase while cross-industry and cross-country deals
63
report the same tendency. Similar to the case of an increased likelihood, higher earnout
ratios in case of high information asymmetry reflect the acquirer’s need for a strong
signalling and incentive tool to mitigate related problems. In so far, the thesis contributes
strong support to theory and previous empirical results. (Kohers & Ang, 2000; Beard,
2004; Cain, Denis & Denis, 2011)
Regression results suggest that the length of the earnout period is not determined by the
uncertainty about the target’s future performance. Consequently, we would have to deny
that acquirers seek to control the likelihood of an earnout to be paid through the shape of
this parameter. To the contrary, Cain, Denis & Denis (2011) report the predicted
significant negative relationship between uncertainty and earnout periods. Moreover,
Lukas & Heimann (2014) in their event-study show that acquirers benefit from shorter
periods. The authors argue that a shorter time frame reduces the impact of “noise” in the
performance measure which is in line with the option-model logic that longer earnout
periods would simply increase the likelihood of the performance gaol to be reached. Their
results therefore give some evidence for the theory that acquirers consider the likelihood
when designing the earnout period (B1) although these studies were not referring to the
theoretic model developed here. After all, the thesis fails to replicate these findings.
Our findings regarding the control variables of information asymmetry remain unclear.
Firms with high growth opportunities do not show a positive impact on the length of the
earnout contract. To the very contrary, other indicators of information asymmetry such
as cross-country and cross-industry deals are highly significant and positive drivers of the
earnout period. We might argue that longer earnout periods in these cases are favourable
to resolve the information asymmetry over time. This argumentation, however, runs
counter to Cain, Denis & Denis (2011) who report longer earnout periods only for high
growth firms. The authors, in line with Lukas & Heimann (2014), to the very contrary
suggest that longer earnout periods in general are unfavourable since they increase the
risk of disputes about the performance measure and the performance goal between the
parties. The empirical results presented here cannot clarify the contradictions regarding
the impact of information asymmetry.
Obviously, the dynamic regarding the required performance increase (B2) could not be
tested due to data limitations. A substantial part of the advanced model on earnout design
therefore remains untested.
64
Although not directly linked to the theoretical model, the thesis additionally examined
the choice of the performance measure. Uncertainty is not found to be a significant
determinant of this choice. To the contrary, indicators of information asymmetry are
reported to drive this decision instead. For targets with high growth opportunities, there
is strong evidence that the performance measure is more likely to be sales or a non-
financial measure rather than income which confirms findings by Cain, Denis & Denis
(2011). However, the empirical findings do not allow to extend this argumentation to all
types of information asymmetry as the remaining indicators show inconsistent results.
Still, the results allow the conclusion that in situations of higher information asymmetry,
income is not the appropriate choice as the performance measure and thereby confirms
the overall conclusion of Cain, Denis & Denis (2011).
To sum it all up, the thesis in combination with previous research finds evidence that the
earnout ratio is an important control lever to the acquirer to face increased adverse
selection and agency problems. The rationale that the acquirer could instead also control
for the likelihood of an earnout premium to be paid finds some promising prove in
previous studies at least for the case of earnout periods. However, this thesis fails to yield
supportive empirical prove. This might be due to some limitations of this study as outlined
in the next subchapter. It remains unknown if the same dynamics apply to the required
performance increase, i.e. the performance goal. Taking all into consideration, the current
state of research, although offering encouraging results, does not allow for a final
conclusion on the value of the option-based model presented in this thesis. Further
research as outlined in 8.3 is strongly encouraged.
8.2 Limitations to the study
The thesis faces limitations with regards to the body of literature available, the data
sample and the theoretical model that should be taken into account in the evaluation.
First, there is only a very limited body of literature available regarding the design of
earnouts to source from. The research that can serve as reference points to evaluate the
empirical results is actually limited to five relevant studies and heavily relies on the most
comprehensive one by Cain, Denis & Denis (2011). Consequently, research is not yet
offering reliable and robust results and lacks consensus. We can therefore only conclude
that previous studies offer promising results for the option-based model on earnout
design, but we can for sure not take the limited body of literature as a proof for its
65
adequateness. The theoretical option pricing model itself can rely on several scholars that
identified the similarities between earnouts and options. However, only the work by
Lukas, Reuer & Welling (2012) transfers this approach into a systematic model. This in
turn, to the author’s knowledge, lacked empirical testing. The advanced model presented
in this thesis and its empirical test are therefore a novel work and should be considered to
some extent explorative.
Second, the earnout specific data selection is a challenging task. Especially the disclosure
of targets’ performance pre-closing is not directly observable. As the most obvious
limitation to this study, hypotheses 3 regarding the required performance increase had
consequently to be dropped. Unfortunately, one parameter that is considered to influence
the likelihood of an earnout payment thereby remains unobserved. Also, since collecting
data on the earnout parameters requires a time-consuming, in-depth analysis of primary
sources such as public deal announcements only a limited sample size could be handled
in this thesis. Consequently, this leads to a loss in representativeness and power of test
statistics. In order to standardize data for empirical analysis several simplifications were
required. The earnout deals in the data sample were assigned a single measure of the
earnout period although they were comprising possible earnout payments in several
instalments over the years. Also, no distinction was made between earnouts with lump
sum payments or more complex formulas determining the exact earnout premium.
Instead, only the maximal possible earnout payment was considered to be able to state an
earnout ratio. These simplifications possibly causes a lack of accuracy.
Third and maybe most importantly, the quality of the empirical test of the theoretical
model is essentially dependent on the quality of its measure of uncertainty. As the
empirical analysis indicates, the results are sensitive to the exact definition of this
measure. The definition via a single proxy target based on the criteria ‘cash-flows’ seems
oversimplified and misspecified. Mostly, this measure yielded highly insignificant
coefficients. The measure for the target’s industry instead can be considered to provide a
reliable measure of inter-industry differences in volatility. Throughout the empirical
analysis, this alternative uncertainty proxy yielded more significant coefficients.
However, a more accurate multi-dimensional matching of the actual target with a
comparable listed proxy would deliver more accurate results than the simple industry
average. Since information on the mostly privately-held targets is not easy to access, this
matching would again be a complex and time-consuming task. In contrast to the
66
uncertainty measure, the dependent variables earnout ratio, earnout period and
performance measure are less likely misspecified since the descriptive statistics proved a
high similarity as compared to data samples of previous studies.
Apart from the definition of uncertainty, the model heavily relies on the assumption that
earnout parameters determine the likelihood of an earnout pay-out. While the hypotheses
tested here are based on this assumption, no analysis was carried out to examine if a
certain design of earnout parameters de facto results in more frequent earnout payments.
Such an analytical approach is so far missing throughout literature.
Consequently, there are several possible avenues for further research on earnout design
that could overcome these limitations.
8.3 Further research
After all, the model on earnout design based on option pricing methodology is still
considered a comprehensive and promising theory. Further research is therefore
encouraged to run additional empirical analyses that overcome the limitations of this
thesis as outlined before. Priority should be given to find a most appropriate and accurate
measure of uncertainty and to collect data on the required performance increase as this
parameter remains the most under-studied part of an earnout contract. Furthermore, it
would be meaningful to test the model on a deal sample only comprising single lump sum
earnouts and exclude any contracts with several instalments paid over the years since
these are suspicious to dilute the option pricing analogy of earnouts.
Another approach to test, if the likelihood of earnouts to end up in the money is correctly
modelled by reference to option pricing techniques, is a long-run study. Data should be
collected on the actual outcome of earnout contracts, i.e. if an earnout had to be paid at
the end of the earnout period. Scholars then could examine if those earnouts that
according to option pricing methodology were designed to more likely end up in the
money actually resulted in more frequent earnout payments.
If future research finds more evidence for the option-based hypotheses to be valid, these
hypotheses at the same time could serve as “rules” for the optimal earnout design. Those
deals that apply them in their contract design would be labelled as optimal and a larger
wealth effect for the acquirer would be expected from these deal announcements. This
thesis did not proceed with this analysis since the determinants of the earnout parameters
67
so far remain uncertain. However, the appendix 3 includes a possible methodological
approach to this problem.
9 Conclusion
This thesis investigates the research question of what factors determine the design of
earnouts in M&A deals. Thereby in contributes to the yet limited research on earnout
design both theoretically and empirically. In order to arrange the empirical results into
the broader context it is therefore worth to revisit the theoretical contributions in brief.
Through a review of earnout literature the two main motives to use earnouts in M&A are
identified. First, to mitigate adverse selection problems by serving as a signalling tool for
high quality targets and second, to mitigate agency problems by serving as an incentive
tool to target’s management in the post-closing phase.
As the main theoretic contribution to research, the thesis develops a model on earnout
design that explains how the common earnout parameters are shaped to ensure the
earnout’s effectiveness as a signalling and incentive tool. The basic assumption of this
model states, that the effectiveness of an earnout depends on the likelihood that the target
will receive an earnout premium in the end. In reference to game-theory, we expect that
a higher likelihood is associated with less efforts by the target in the post-closing phase
and with even lower quality firms tending to accept the earnout. Consequently, as the
likelihood for an earnout payment increases, the earnout is expected to lose its power as
a signalling and incentive tool. The model suggests two ways for the acquirer to react to
an increased likelihood for an earnout premium to be paid. On the one hand by increasing
the earnout ratio, i.e. the contingent part of the overall acquisition price, in order to
stronger incentivize the target for post-closing cooperation and to motivate only high
quality targets to accept the deal and on the other hand by shaping the earnout parameters
such that the likelihood is controlled for.
Based on an in-depth analysis of the similarities between earnouts and financial call
options, the thesis derives that the same forces that determine a financial option to expire
in the money also determine the likelihood of an earnout agreement to result in an earnout
premium paid to the target. In reliance on option-pricing techniques a higher uncertainty
about the target’s future performance, a longer earnout period and a lower performance
goal are all expected to increase the likelihood for the earnout to expire in the money.
68
Consequently, the hypotheses are derived that in case of high uncertainty, i.e. increased
likelihood, the acquirer reacts with an increased earnout ratio, shorter earnout periods or
higher performance goals. Thereby, the thesis successful answers the states research
question from a theoretical perspective.
The thesis further contributes to research by testing these hypotheses empirically on a
sample of 377 earnout deals from UK acquirers between 2006 and 2015. Results from
tested regression models document weak evidence that the acquirer choose higher earnout
ratios in case of high uncertainty, while previous studies even show strong support for
this relationship. Further, this thesis reports strong evidence that in case of high
information asymmetry between acquirer and target, like in cross-country and cross-
industry deals and for targets from the service industry, the earnout ratio increases. Again,
this result is in line with previous studies. Consequently, there is strong evidence that the
earnout ratio is used as the primary “control lever” if the acquirer faces the need to design
strong signalling and incentive tools in case of high uncertainty and information
asymmetry.
In contrast, the regression models regarding the earnout period yield inconsistent and
insignificant results. Previous research, however, finds that acquirers indeed choose
shorter earnout periods if uncertainty and consequently the likelihood is high. In context
of this yet limited literature, the rationale that acquirers shape the earnout period in order
to offset increased likelihood due to high uncertainty at least remains reasonable.
Unfortunately, due to data limitations the determinants of the performance goal could not
be empirically examined. A substantial part of the advanced model on earnout design
therefore remains untested and no previous study offers supplementary evidence.
Thereby, the thesis to some extent fails to successfully answer the research question from
an empirical perspective.
Though not related to the theoretical model, the thesis also provides empirical evidence,
that the appropriate performance measure in situations of high information asymmetry is
not income but rather sales or non-financial, while uncertainty is no significant
determinant.
After all, as the main contribution to earnout research the thesis develops a model that
explains that not only the contingent part of the overall acquisition price, i.e. the earnout
ratio, but also the remaining earnout parameters play a role in ensuring an earnout’s power
69
to serve its purpose of solving information asymmetry problems. The real option view on
earnouts proves to be a useful framework to understand the implications of different
designs on the earnout’s value as a risk-reducing instrument. While empirical evidence is
only partly available through this thesis and previous studies, this avenue of research is
promising. Future research that overcomes some of the data limitations of this thesis, that
utilizes more elaborate measures for uncertainty in the target’s future performance and
that conducts a long-run analysis of the likelihood of earnout premiums to be paid
eventually is strongly encouraged to pave the way towards the definition of optimal
earnout design.
70
References
Akerlof, G.A. 1970, ‘The market for “lemons”: Quality uncertainty and the market
mechanism’, Quarterly Journal of Economics, vol. 84, no. 3, pp. 488-500.
American Appraisal 2015, Earn-Outs and Contingent Consideration: Valuation,
viewed 07 May 2015, <http://www.americanappraisal.com/US/Library/Articles/
EarnOutsandContingentConsideration.htm?PrintPage=yes&FB_Values=&&&#>.
Asquith, P., Bruner, R.F. & Mullins, D.W. 1983, ‘The gains to bidding firms from
merger’, Journal of Financial Economics, vol. 11, no. 1, pp. 121-139.
Barbopoulos, L. & Sudarsanam, S. 2012, ‘Determinants of earnout as acquisition
payment currency and bidder’s value gains’, Journal of Banking & Finance, vol. 36, no.
3, pp. 678-694.
Bargeron, L.L., Schlingemann, F.P., Stulz, R.M. & Zutter, C.J. 2008, ‘Why do private
acquirers pay so little compared to public acquirers?’, Journal of Financial Economics,
vol. 89, no. 3, pp. 375-390.
Beard, D.R. 2004, ‘The mitigation of asymmetric information through the use of
earnouts’, Dissertation, Louisiana State University, Louisiana.
Berk, J. & DeMarzo, P. 2014, Corporate Finance – Global Edition, 3rd edn, Pearson,
Harlow.
Bruner, R.F., Stiegler, S. 2001, ‘Technical note on structuring and valuing incentive
payments in M&A: Earnouts and other contingent payments to the seller‘, Working Paper
(Excerpt), Darden Graduate School of Business Administration, University of Virginia.
Cain, M.D., Denis, D.J. & Denis, D.K. 2011, ‘Earnouts: A study of financial contracting
in acquisition agreements’, Journal of Accounting and Economics, vol. 51, pp. 151-170.
Caselli, S., Gatti, S. & Visconti, M. 2006, ‘Managing M&A risk with collars, earn-outs,
and CVRs’, Journal of Applied Corporate Finance, vol. 18, no. 4, pp. 91-104.
Chang, S. 1998, ‘Takeovers of privately held targets, methods of payment, and bidder
returns’, The Journal of Finance, vol. 53, no. 2, pp. 773-784.
71
Craig, B. & Smith, A. 2003, ‘The art of earnouts’, Strategic Finance, vol. 84, no.12, pp.
44-47.
Datar, S., Frankel, R. & Wolfson, M. 2001, ‘The effects of adverse selection and agency
costs on acquisition techniques’, Journal of Law, Economics, & Organization, vol. 17,
no. 1, pp. 201-238.
Del Roccili, J.A. & Fuhr, J.P. 2001, ‘The pros and cons of earnouts‘, Journal of Financial
Service Professionals, vol. 55, no. 6, pp. 88-93.
Eckbo, B.E. 2009, ‘Bidding strategies and takeover premiums: A review’, Journal of
Corporate Finance, vol. 15, no. 1, pp. 149-178.
Fuller, K., Netter, J. & Stegemoller, M. 2002, ‘What do returns to acquiring firms tell
us? Evidence from firms that make many acquisitions’, The Journal of Finance, vol. 57,
no. 4, pp. 1763-1793.
Holmström, B. 1979, ‘Moral hazard and observability’, The Bell Journal of Economics,
vol. 10, no. 1, pp. 74-91.
Hull, J.C. 2012, Options, Futures, and other Derivatives – Global Edition, 8th edn,
Pearson, Harlow.
Kile, C.O. & Phillips, M.E. 2009, ‘Using industry classification codes to sample high-
technology firms: Analysis and recommendations’, Journal of Accounting, Auditing &
Finance, vol. 24, no. 1, pp. 35-58.
Kohers, N. & Ang, J. 2000, ‘Earnouts in mergers: Agreeing to disagree and agreeing to
stay’, Journal of Business, vol. 73, no. 3, pp. 445-476.
Krishnamurti, C. & Vishwanath, S.R. 2008, ‘Real options analysis in mergers and
acquisitions’, in Krishnamurti C. & Vishwanath S.R. (ed.), Mergers, acquisitions and
corporate restructuring, SAGE Publications India, New Delhi, pp. 115-145.
Luehrman, T.A. 1998, ‘Investment opportunities as real options: Getting started on the
numbers’, Harvard Business Review, vol. 76, no. 4, pp. 51-67.
Lukas, E. & Heimann, C. 2014, ‘Technological-induced information asymmetry, M&As
and earnouts: Stock market evidence from Germany’, Applied Financial Economics, vol.
24, no. 7, pp. 481-193.
72
Lukas, E., Reuer, J.J. & Welling, A. 2012, ‘Earnouts in mergers and acquisitions: A
game-theoretic option pricing approach’, European Journal of Operational Research,
vol. 223, pp. 256-263.
Mantecon, T. 2009, ‘Mitigating risks in cross-border acquisitions’, Journal of Banking
& Finance, vol. 33, no. 4, pp. 640-651.
McDonald, R. & Siegel, D. 1986, ‘The value of waiting to invest’, The Quarterly Journal
of Economics, vol. 101, no. 4, pp. 707-728.
Moeller, S.B., Schlingemann, F.P. & Stulz, R.M. 2004, ‘Firm size and the gains from
acquisitions’, Journal of Financial Economics, vol. 73, no. 2, pp. 201-228.
Myers, S.C. & Majluf, N.S. 1984, ‘Corporate financing and investment decisions when
firms have information that investors do not have’, Journal of Financial Economics, vol.
13, no. 2, pp. 187-221.
Quinn, B.J.M. 2012, ‘Putting your money where your mouth is: The performance of
earnouts in corporate acquisitions’, University of Cincinnati Law Review, vol. 81, pp.
127-172.
Ragozzino, R. & Reuer, J.J. 2009, ‘Contingent earnouts in acquisitions of privately held
targets’, Journal of Management, vol. 35, no. 4, pp. 857-879.
Reuer, J.J., Shenkar, O. & Ragozzino, R. 2004, ‘Mitigating risk in international mergers
and acquisitions: The role of contingent payouts’, Journal of International Business
Studies, vol. 35, no. 1, pp. 19-32.
Spence, M. 1973, ‘Job market signaling’, Quarterly Journal of Economics, vol. 87, no.
3, pp. 355-375.
Thompson, C.J. & Schnorbus L.A. 2010, M&A facilitators: The value of earnouts,
viewed 19 May 2015, <http://www.srr.com/article/ma-facilitators-value-earnouts>.
Travlos, N.G. 1987, ‘Corporate Takeover Bids, Methods of Payment, and Bidding Firms'
Stock Returns’, Journal of Finance, vol. 42, no. 4, pp. 943-963.
Trigeorgis, L. 1991, ‘Anticipated competitive entry and early preemptive investment in
deferrable projects’, Journal of Economics and Business, vol. 43, no. 2, pp. 143-156.
73
Veall, M.R. & Zimmermann, K.F. 1994, ‘Goodness of fit measures in the Tobit model’,
Oxford Bulletin of Economics and Statistics, vol. 56, no. 4, pp. 485-499.
Verbeek, M. 2012, A guide to modern econometrics, 4th edn, John Wiley & Sons,
Chichester.
White, H. 1980, ‘A heteroscedasticity-consistent covariance matrix estimator and a direct
test for heteroscedasticity’, Econometrica, vol. 48, no. 4, pp. 817-838.