1

Make, Buy, or Ally?

Choice of and Payoff to Alternate Strategies for Innovations

Abhishek Borah and Gerard J. Tellis Abhishek Borah is Assistant Professor at the University of Washington, Seattle, WA: abhi7@uw.edu Gerard J. Tellis is Professor of Marketing and Management & Organization, Director of the Center for Global Innovation and Neely Chair of American Enterprise at the Marshall School of Business, University of Southern California. Address: P.O. Box 90089-1421, Los Angeles, California, USA. Tel: +1.213.740.5031, fax: +1.213.740.7828, E-mail: tellis@usc.edu. This project benefitted from a grant of Don Murray to the Center for Global Innovation.

2

Make, Buy, or Ally?

Choice of and Payoff to Alternate Strategies for Innovations

Abstract

Firms constantly grapple with the question of whether to make, buy, or ally for innovations.

The literature has not analyzed the choice of and payoff to these alternate routes to innovation for

the same firm. To address this issue, the authors collect, code, and analyze the choice of and payoff

to 3522 announcements of make, buy, and ally innovation for 192 firms across 108 industries over 5

years.

The authors find that make and ally generate positive and higher payoff than buy, which

generates a negative payoff. Nevertheless, firms continue to buy for two reasons. First, firms seem

to have no memory for the payoff to buy even though they have a memory for the payoff to make.

Second, firms tend to buy when they lack commercializations, even though the strategy seems not to

pay off. These results suggest that firms see buys as a quick fix for what may be a deep strategic

problem. Nevertheless, the negative returns to buy can be mitigated if the acquirer is experienced

and the target is related and offers high customer benefit. The authors offer explanations for and

implications of the results.

Keywords: Innovation, Announcements, Make, Buy, Ally, Payoff, Abnormal Returns, Content Analysis, Event Study

3

Innovations are critical for survival, growth, and success in today’s globally competitive

markets, especially when recession depresses sales of mature products. Making, buying, and allying

for innovations are the three most widely used strategies for expanding a firm’s innovation portfolio.

Firms spend billions of dollars each year in implementing these three strategies. Booz & Company

estimates that the top 1000 public firms investing in R&D spent $603 billion to make innovations in

2011 (Booz & Co. 2012). The Boston Consulting Group reports 22,700 mergers and acquisitions in

2011 worldwide with a value of $1.79 trillion (BCG 2012). Concurrently, Dealogic estimates firms in

2011 spent $12.1 billion in joint ventures in emerging markets alone (KPMG 2013).

Despite extensive research on make, buy or ally decisions across disciplines such as strategic

management, economics, marketing, and law (Geyskens, Steenkamp and Kumar 2006), several

aspects of the relative payoff to the make, buy, or ally decision are still not clear.

First, prior research has not compared the payoff to make, buy, and ally innovations of the

same firm. Researchers have evaluated making, buying, or allying for innovations in distinct studies

using separate samples of firms (See Table 1). Separate analysis may not be comparable because of

differences in samples or contexts. So, the current study compares the choice of and payoff to make,

buy, and ally for innovations within the same firm and time periods.

Second, prior studies have not examined whether firms learn from their past successes or

mistakes. That is, do past successes stimulate adoption of the successful strategy while past failures

lead to avoidance of the failing strategy? For example, HP recently announced a write down of $ 8.8

billion from an $11 billion acquisition of Autonomy announced earlier. Ironically, HP similarly

announced huge write-downs in two similar acquisitions, Palm and EDS.

Third, researchers have also not examined how a firm’s past commercialization of

innovations affects its announcement strategy for make, buy, and ally innovations. For example, is

4

buy used by firms to complement a high level of commercializations or to compensate for a low

level of commercializations? So, the current study seeks to answer the following questions:

What factors drive a firm to choose between announcing a make, buy or ally?

How do payoffs differ for make, buy, and ally for innovations of the same firm?

What factors affect the payoff from make, buy, or ally for innovations?

Do firms “learn” from the success or failure of their past strategies? That is, do payoffs from past strategies affect current strategies?

How does a firm’s prior commercialization of innovations affect the firm’s choice to make, buy, or ally for innovations?

To answer the questions posed above, we collect a unique dataset of 3522 announcements

of make, buy, and ally innovations for 192 firms across 108 industries for 5 years. We model the

choice of a make, buy, and ally using a Multinomial Logit model with correlated responses and firm

heterogeneity and then carry out a regression analysis of the payoff as a function of the choice of the

make, buy, and ally and other explanatory and control variables. We employ an event study to

estimate payoffs to announcement of make, buy, and ally innovation. Our models control for

selection bias, firm heterogeneity, repeated observations and other explanatory factors. Further, we

create novel, dynamic measures of innovation relatedness and customer benefit using content

analysis. To simply the terminology and for ease of exposition, we use the abbreviated terms

“make,” “buy,” and “ally” for announcements of make, buy, and ally for innovations.

The rest of the paper is organized as follows: The next four sections present the theory,

method, model, and results. The last section concludes with the findings, discussion of the findings,

implications, and the limitations.

Theory and Hypotheses

This section explains the conceptual framework and builds hypotheses for the drivers of the

choice of and payoff to make, buy, and ally. We begin with the definitions of the key terms and the

theory for the payoff metric employed in the study.

5

Definitions We define seven key terms relevant to the study: innovation project, initiation phase,

announcement, make, buy, ally, and choice. We define an innovation project as the total of a

firm’s activities in researching, developing, and introducing a new product or service (Sood and

Tellis 2009) (See Figure 1). We define the initiation phase as the start of an innovation project. An

innovation project comprises three phases: initiation, development and commercialization (Sood and

Tellis 2009). We define an announcement as the release of information by the firm directly or by

other sources about some event (e.g., make, buy, and ally) in the innovation project. We define

make as a firm’s announcement to develop a new product or service internally at the initiation phase

of the innovation project. We identify four types of make: starting a research and development

center or research laboratory or innovation center; starting a new process or system; starting a new

entity or division; starting a new project for developing a product or service. Note that make is a

promise or intention for future innovation. We use the term buy to refer to a firm’s announcement

to acquire a firm or a part of a firm explicitly for its innovation at the initiation phase. We exclude all

acquisitions undertaken for non-innovation reasons (e.g., cost considerations, reducing tax

obligations, economies of scale, etc.). We identify four types of buy: buying patents; buying software

or technology; buying research personnel; buying product or service that are modified or combined

with current innovations. Note that buy may also be a promise because in most cases the firm has to

integrate R&D, production, branding, marketing, and distribution, which is not guaranteed to be

successful. We define ally as the announcement of joining of two or more entities, for a specified or

unspecified period, to develop an innovation at the initiation phase. These alliances, which include

open innovations (Chesbrough 2003; West et al. 2006; Hagel III and Brown 2008; Lichtenthaler

2011; Bayus 2013) comprise: forming a joint venture with other companies to develop products, co-

developing products with firms, licensing of technology, hiring an expert on a contract basis to

6

answer a research problem, collaborating with universities or research institutes, and participating in

networks to develop innovations. Thus, alliances can be of three types: joint ventures, strategic

alliances, and licensing agreements. Note that ally may also be a promise because in most cases the

firm has to share R&D, production, branding, marketing, and distribution, with the partner or

partners, which is not guaranteed to be successful. We use the term choice to denote the choice

among make, buy, or ally.

The literature supports the consideration of make, buy, and ally as alternatives. Kreutzer

(2012) points out, that firms are aware that these three choices represent alternative means of

innovation. Dyer et al. (2004) find that 82% of firms view acquisitions and alliances as substitutes.

For example, Cisco Systems has one senior vice president responsible for acquisitions, alliances, and

internal development (Dyer et al. 2004). By placing all three functions under the same person, Cisco

checks the feasibility of each option starting with its internal capabilities. Cisco’s head of corporate

development states, “… we make the choice between internal development, acquisitions, or

alliances. At some point, I have to make the decision about what’s the right strategy for us”. Cisco

decides one among the three options using various criteria (Cisco Systems Canada 2009). Capron &

Mitchell (2010) argue that the typical firm choses one choice out of the three alternatives. They

provide Eli Lilly as an exemplar firm that choses one of the three choices at any given time based on

their existing capabilities and partnership characteristics.

Logic for Event Study Tracking the long term success of each of these choices (make, buy, or ally) is difficult if not

impossible because companies do not reveal the specific outcomes to each of these announcements

and the causes of the outcome. To resolve the information problem, we make use of the event

method. This method relies on the assumption that the stock price at a particular point in time fully

reflects all available information up to that point (Sharpe 1964; Fama 1998). The stock price relies

7

on trades by a market of thousands of investors the world over who track the company’s choices

and performance. The company’s announcement of a new decision reflects new information that

may affect the stock price positively, negatively, or not at all. The change in stock price, if any, on

the announcement of a decision reflects the discounted future cash flows that the market expects

from the decision, taking into account the past performance of the company, its future potential,

and its competition (Fama et al. 1969; Fama 1991). Thus, by definition, returns are future looking

and incorporate the long term (Campbell, Lo and MacKinlay 1997, Srinivasan and Bharadwaj 2004,

Srinivasan and Hanssens 2009). By this logic, the stock market returns to make, buy, or ally would

reflect the discounted expected returns in the future for that decision. By analyzing these returns

against the characteristics of these announcements, we can assess what is the payoff to make, buy, or

ally and what factors drive that payoff. The use of this method is increasing in marketing (e.g., Tellis

& Johnson 2007; Joshi & Hanssens 2008; Sood & Tellis 2009; Luo 2009; Wiles et al. 2010).

The Drivers of Choice This section proposes the conceptual framework for the drivers of choice and derives

several hypotheses. Based on the literature, three broad constructs drive choice: firm resources, firm

strategy, and firm outcomes (Jensen 1986; Haleblian et al. 2006; Mizik and Jacobson 2003; Sorescu

et al. 2007; Williamson 1975; Sood & Tellis 2009; Kelm et al. 1995). Figure 2 depicts the conceptual

framework. The most important firm resources are its managerial and financing capability. Within

firm strategy, the key constructs that drive choice are the firm’s innovation emphasis, value (creation

and appropriation) emphasis, and diversification emphasis. We measure a firm’s innovation

emphasis by its prior makes, buys, and allies. We measure a firm’s emphasis on value creation by its

research and development investments and on value appropriation by its marketing investments.

Within firm outcomes, the key constructs are based on outcomes of each phase of an innovation

project: initiation, development, and commercialization (Sood and Tellis 2009). We measure

8

outcomes of the initiation phase by payoff to prior make, buy, and ally. The most important

outcomes of the development and commercialization phases are the number of patents and the

number of commercializations, respectively.

We next build hypothesis on the effect of the outcomes of the initiation and

commercialization phase of an innovation project on choice. We focus on the firm outcome

construct as there is no prior research, which has examined how firm outcomes affect choice.

Performance feedback plays a crucial role in indicating if a firm needs to maintain or change its

management strategy. Firms can differ in their responses to performance outcomes (March 1981).

Payoff to Prior Make, Buy, and Ally The payoff to prior make, buy, and ally can affect choice (Cyert and March 1963; Levitt and

March 1988). Organizational learning theory states that firms are outcome-oriented (Levitt and

March 1988). A firm observes an outcome and links the outcome to routines (Levitt and March

1988). If a routine is associated with a successful outcome, the firm learns to continue that routine.

If a routine is associated with an unsuccessful outcome, firms learn to avoid the routine (Cyert and

March 1963). If firms achieve a positive make, buy, or ally outcome, it serves as a positive

reinforcement to a firm’s current strategy. This prompts firms to continue with the strategy.

Substantial evidence exists that outcomes affect firm behavior: good outcomes lead to organizational

persistence (e.g., Miller and Chen, 1994) while bad outcomes lead to organizational change

(Haleblian et al. 2006). This line of reasoning suggests:

H1a: High payoff to make in the prior year encourages firms to make H1b: High payoff to buy in the prior year encourages firms to buy H1c: High payoff to ally in the prior year encourages firms to ally

Number of Commercializations Firms vary in their ability to commercialize innovations (Chandy et al. 2006).

Commercializations are more relevant to marketing scholars and most important for marketing

managers than any intermediate output (Prabhu et al. 2005). Indeed, Hauser, Tellis and Griffin 2006

9

point out that “a firm’s overall profitability results from the portfolio of products it commercializes

over time and across product lines”. We posit that commercialization intensity is a factor that affects

buy and ally negatively and make positively for the following reasons. First, for firms with low

commercializations, buy allows firms to gain access to the ideas, talent, and tacit and codified

knowledge of targets, which eventually translate into commercializations. Second, buy and ally allow

a relatively fast means of obtaining commercializations. On the other hand, make takes a long time

to materialize into a commercialization. Much research has shown that it is desirable to bring

products to market quickly (Kessler and Chakrabarti 1996). Buying innovative firms or partnering to

obtain innovations allows firms to quickly gain access to the target’s commercializable products,

trusted channel relationships, and loyal customer base. This line of reasoning suggests:

H2a: High number of commercializations in the prior year encourages firms to make H2b: Low number of commercializations in the prior year encourages firms to buy H2c: Low number of commercializations in the prior year encourages firms to ally

The Drivers of Payoff to Make, Buy, or Ally Figure 3 shows our conceptual framework for the drivers of payoff to make, buy, or ally. We

suggest two broad constructs that drive the payoff to make, buy, or ally: firm resources and firm

strategy. Under firm resources, the key construct that drives payoff is the firm’s financing ability.

Under firm strategy, three key constructs that drive payoff are the firm’s focal innovation emphasis,

value (creation and appropriation) emphasis, and diversification emphasis. We measure a firm’s focal

innovation emphasis by the make, buy, or ally choice and the customer benefit, relatedness, location,

and riskiness of the innovation.

We do not have specific hypotheses about the payoff to make, buy and ally, because the

literature contains arguments for both a positive payoff and a negative payoff. Instead we merely

summarize the reasons proposed for positive and negative payoff and propose to address these

issues as research questions.

10

First, make can lead to positive payoff because of internalization of capabilities, ownership

of intellectual property, and full capture of returns (Doukas and Switzer 1992; Kanter 1999; Cefis

and Marsili 2006; Kreutzer 2012). On the other hand, make can lead to negative payoff because of

uncertainty, huge investments, and long development periods (Nakamura 2001; Lev 2001; Erickson

and Jacobson 1992; Griffin 2002; Rothaermel and Hess 2010; Kreutzer 2012). This study will seek to

empirically examine whether make leads to a positive or negative payoff.

Second, buy can lead to positive payoff because of acquisition of knowledge such as new

technology, talent, and intellectual property (Ranft & Lord, 2000; Mayer and Kenney 2004). On the

other hand, buy can lead to negative payoff because of disturbance in conventional routines, high

transaction costs, low reversibility, risk of overpayment and cultural clashes (Williamson 1975;

Nahavandi and Malekzadeh 1988; Hitt et al. 1996; Hitt et al. 2009; Rothaermel and Hess, 2010;

Kreutzer 2012). This study will seek to empirically examine whether buy leads to a positive or

negative payoff.

Third, ally can lead to positive payoff because of shared risk, access to knowledge with low

transaction costs, and high flexibility to enter, commit, or exit (Gomes Casseres 2000; Kreutzer

2012; Capron and Mitchell 2012). On the other hand, ally can lead to negative payoff because of lack

of management attention, relationship risks, threat of opportunistic behavior, potential competition

between partners, and shared returns (Parkhe 1993; Bleeke and Ernst 1995; Ybarra and Turk 2011;

Kreutzer 2012; Capron and Mitchell 2012). This study will seek to empirically examine if ally leads to

a positive or negative payoff.

Control Variables We use the following control variables in our choice model: Managerial capability, financing

capability, number of patents, prior number of make, prior number of buy, prior number of ally,

marketing investments, R&D investments, diversification levels, and industry. Please see Table 2 for

11

the reasons and Online Appendix A1 (Panel A) for the measures of for these variables. We use the

following control variables in our payoff model: Relatedness of the innovation, customer benefit of

the innovation, marketing investment, R&D investments, financing capability, diversification, prior

risk to make, prior risk to buy, prior risk to ally, location of make, location of buy, location of ally,

types of make, types of buy, types of ally, industry, and competition. The rationale for the inclusion

of these control variables are in Table 2 and the measures are in Online Appendix A1 (Panel B).

Method

We test the hypotheses by assembling data from 192 firms across 108 industries. We collect

these data using the historical method (Golder and Tellis 1993; Golder 2000). Below we detail the

sample selection, data collection, and the measures of the focal variables.

Sample Selection We use four different samples to minimize any sample selection bias and maximize the

generalizability of results. The four samples exclude firms not listed on the American Stock

exchanges and financial institutions because they experienced considerable turmoil during our

study’s timeframe. Moreover, financial innovations are inherently risky and complex.

Sample 1. This sample is drawn from a list of the most innovative firms in the world. We

include this sampling frame because the most innovative firms have an ample number of makes. We

use the 2008 BusinessWeek and Boston Consulting Group’s list of the 50 most innovative firms in the

world for the sampling frame. The list includes firms from various industries. Although the list has

50 firms, we could only include 36 firms. We drop 14 firms because they are either financial

institutions or not listed on the American stock exchanges.

Sample 2. We ensure that we have a sizeable number of buy by selecting the 36 most

acquisitive firms in the world1. We rely on Securities Data Company’s (SDC) mergers and

1 We use the period from 2002 to 2007 to select the top 36 acquirers. Financial institutions are dropped.

12

acquisitions database to extract the list. The SDC database comprehensively covers all worldwide

mergers and acquisitions from 1985.

Sample 3. We randomly select another 64 firms from the Fortune 500 list in 2008. The

Fortune 500 is a list compiled by Fortune magazine ranking the top 500 public corporations of the

United States of America as measured by their gross revenue. This sampling frame allows us to

select the largest publicly held firms in US.

Sample 4. We randomly select another 64 firms from Fortune’s 501-1000 list in 2008. This

sampling frame allows us to select relatively smaller and publicly held firms in US. Samples 3 and 4

include firms, which have low to high number of make, buy, and ally.

As a result, we have 192 firms in our sample. This sample selection strategy enables us to

compile a substantial number of make, buy, and ally from various industries. Firms range from large

to small, innovative to non-innovative, with low to high make, buy, and ally rates, and include

marketers of products and services. This sampling strategy leads to a broad, representative set of

firms and is important for the generalizability of results.

There is an overlap of 8 firms between Samples 1 and 22. The list of the 192 firms is in

Online Appendix A2. For time sampling, we chose the period from July 1st 2002 to June 30th 2007.

We focus on this 5 year period since the Fama-French factors were available latest till June 2007

when we started data collection in 2008. The availability of the Fama-French factors is intrinsically

important for our study.

Data Collection We use a number of sources for collecting the make, buy, and ally. We identify

announcements using four respected syndicated sources (Capital IQ, Factiva, SDC database, and

LexisNexis® Academic) plus other sources (company websites, news websites, company blogs, and

2 We exclude firms from sample 1 and 2 for the random selection process for samples 3 and 4.

13

independent blogs). As such, we use a variety of sources to collect all the information related to the

innovation. We code an announcement as a make, buy, or ally if it occurs in at least two or more of

the sources above. Details about data sources are in Online Appendix A3.

We use Capital IQ, Factiva, and Lexis-Nexis to identify and control for seven different types

of confounding events around the -1 to +1 window. The procedure of the data collection is in

Online Appendix A4. We identify 1174 make, 1331 buy, and 1017 ally innovations, which total 3522

announcements. All these announcements are at the level of a specific innovation project of a firm.

We eliminate 1671 events due to confounding events from the 3260 announcements, which we use

in our Multinomial Logit model. Thus, our sample for the event analysis and payoff regression

model comprises 441 make, 754 buy, and 394 ally for a total of 1589 events. This sample size is

higher than all other marketing studies, which use the event study method and also run a check for

confounding events3. Overall, collecting, reading, and coding all the different types of data used in

the study consumed around 3100 man-hours.

Measures This section describes the measures and rationale of the key variables in the hypotheses. The

details and definition of the measures for all the other variables are in Online Appendix A1.

Payoff to Prior Make, Buy, and Ally We measure this variable by averaging the returns to the firm’s make, buy, and ally per year

prior to the date of the current event. We use the -1 to +1 day window for calculating the returns for

each event. We measure this variable over the prior one year but also test for longer windows in the

robustness checks. Our position for using this measure of prior payoff is the following. When

considering whether to make buy or ally, managers are very likely to consider how the firm fared

when it made the decision in the past (Cyert and March 1963; Levitt and March 1988). In other

3 Wiles et al. (2012) use 880 events; Chen et al. (2011) use 606 events; Karniouchina et al. (2011) use 928 events; Swaminathan & Moorman (2009) use 230 events; Tellis & Johnson (2007) use 421 events; Gielens et al. (2002) use 98 events; Some studies use a higher number of events but they don’t control for confounding events (Sood & Tellis 2009; Elberse 2007); Meta-analyses for 127 acquisition studies report average sample sizes of 221 events (King et al. 2004).

14

words, they are likely to learn from past experience with these decisions (Cyert and March 1963;

Levitt and March 1988). The most recent payoff to these decisions in the prior year is likely to be the

most important factor influencing a manager’s learning from the past (Haleblian et al. 2006).

Number of Commercializations We measure this variable by the number of new product launches per year prior to the date

of the current event. We create this measure with the following formula:

(1) ShareCommit =

where NPAit stands for new product launch announcement for firm i on day t. We sum all new

product launch announcements for a firm i prior to day t up to one year and divide the same by the

total of all new product launch announcements for all firms prior to t up to one year. We rely on the

Capital IQ database for this particular variable. We read each entry under the category of “Product-

Related Announcements” within the Key Developments feature to ascertain a new product launch.

Since the database has complete coverage only from January 2002 for new product launch

announcements, we use a moving window of 1 year for announcements from January 1st 2003 to

June 30th 2007 and the maximum time available (6 months to 1 year) for announcements from July

1st 2002 to December 31st 2002.

Model This section describes the Carhart Four Factor model for returns, the model for choice, and

the model for payoffs. Because we have firms’ decisions that take the form of announcements of

make, buy, and ally with known time stamps, we employ the event method to gauge the impact of

make, buy, and ally on returns (Srinivasan and Hanssens 2009).

i

t

tt

it

t

tt

it

NPA

NPA

365

1

365

1

15

Four-Factor Model for Computing Returns The normal return to a stock is the price of a stock on a day minus that on the prior day

divided by the price on the prior day. The expected return of a stock on a day is the return of a stock

that can be predicted for that day based on a general market index such as the S&P. The abnormal

returns to a stock due to an event is the normal returns minus the expected returns for the same day.

As such, abnormal returns control for fluctuations in price that affect the whole market (not due to

the particular event of a particular firm) as revealed in movements of a market index. We use the

Fama-French and Carhart Four-Factor model to calculate the abnormal returns. We use the

cumulative average abnormal returns (CAR) in the (-1, 1) window as our payoff metric. As the

model has been used in prior research (e.g., Wiles et al. 2010; Tirunillai and Tellis 2012), we skip the

details for brevity. The interested reader can refer to Online Appendix B for the details.

Model for Choice Since we have three unordered announcements (make, buy, ally), we estimate a Multinomial

Logit Choice Model. Our specification allows for correlated choices. For example, a firm’s make

choice might be correlated with its buy choice. Our specification also allows for firm heterogeneity.

For example, two firms with similar make, buy, or ally experience could develop different abilities of

managing make, buy, or ally. In equation form, the Multinomial Logit model with random

intercepts is:

(2) iccij

1ij

ijcux

'log , c= make, ally, buy

i stands for firm and there exists I (i=1,2,…….I) firms. The ith firm has ni observations where j

denotes the jth observation at each unique time t (which in our case, is a day). = P( Yij = c) are

response probabilities where Yij denotes the jth response for firm i (j=1,2,…..ni). This response is

from one of c (c=make, ally, buy) choices. xijdenotes a column vector of p explanatory variables for

the jth observation for firm i. xij include the hypothesized variables: prior make, buy, and ally

p ijc

16

li (ac ,bc ,S ) =exp(ac + xij

' bc +uic )

exp(am + xij' bm +uim )

m=1

3

å

é

ë

êêêê

ù

û

úúúú

I (Yij=c)

j=1

ni

Õ

æ

è

ççççç

ö

ø

÷÷÷÷÷

-¥

¥

ò X fu(ui ,S )dui

payoffs, number of commercializations, and numerous control variables defined later (see Online

Appendix A1). Whenever possible, we use a window of 1 year prior to the announcement date for

calculating the explanatory variable. Else we use the prior year’s value. This helps us tackle any

endogeneity issues (Boulding and Staelin 1995). are constant terms and the effects of the p

explanatory variables are assessed through bc = (b1c,b2c,.........bpc )'. The ac

and bc are considered

as fixed effects. uic are considered as random effects. We assume a multivariate normal distribution

for uic with an expectation of 0 and an unstructured covariance matrix S i.e., for ui = (ui1,ui2,ui3)' ,

we have ui ~ N(0,S ) . For reasons of identification, we have a1 = 0,b1 = 0, andui1 = 0. This

identification scheme results in the interpretation of parameters with reference to the first category

and S to be a 2 by 2 matrix. In our model, the first category is make (c=1 where 1 stands for make).

Ally is indicated by c=2 and buy by c=3. Thus, the likelihood contribution of the ith firm is:

(3)

Here, fu(ui ,S ) is the multivariate normal density and I()is the indicator function. The overall

likelihood function is the product of the likelihood contributions for each firm i.e., li . This likelihood

function consists of a product of I integrals, where each of these cannot be solved in closed form.

Thus, maximum likelihood estimation of the parameters is not possible. We thus resort to estimating

the model using numerical integration, more specifically by adaptive Gaussian quadrature. We use

SAS to program and estimate the model (Kuss and McLerran 2007)4. Note our estimation does not

4 Following Kuss and McLerran, we parameterize the covariance of the random effects such that (1) the estimate of the variance is positive-definite, (2) the components of the covariance are constructed as a correlation multiplied by the root of the product of the components of the variance, and (3) the correlation is constrained to be between −1 and 1.

ac

17

suffer from the Independence of Irrelevant Alternatives assumption because we assume the random

intercepts to be correlated across choice occasions.

Model for Payoff To specify the payoff model we first test for sample selection bias in panel data following the

procedure suggested by Verbeek and Nijman (1992). If we don’t find evidence of selection, we

estimate the model without controlling for selection bias. If we find evidence for such bias, we

resort to procedure suggested by Wooldridge (1995).The details of the procedure and the results are

in the Online Appendix C. We find no evidence of sample selection bias. So, we run a random

effects panel regression analysis of payoff as a function of the independent variables, combining

make, buy, and ally in one model. In equation form, we estimate the following model:

(4) 4

aitbitiit εγ*controlsβ*allyβ*buyαpayoff ++++=

for i=1,2,…….I firms, t denotes the time (which in our case, is a day) when firm i makes a choice,

iα are random firm specific effects and assumed to be IID , 4

itε is an idiosyncratic error, itbuy is an

indicator variable where 1 indicates that the firm chose buy, itally is an indicator variable where 1

indicates that the firm chose ally. The control variables such as R&D investment, marketing

investment, financing capability, and prior risk are measured prior to the announcement date to

control for endogeneity (Boulding and Staelin 1995). Our estimation method takes care of the

potential problem of unobserved heterogeneity because we use a random effects model (Greene

2003). We use Stata’s xtreg command and use the vce option to obtain firm-robust standard errors.

Results

All our announcements are within each project of a firm. Within a project, the vast majority

of announcements (N=3260, 92.5%) are for only a make or ally or buy. We call these pure strategies.

A minority of projects have a combination of two or more strategies. We call these mixed strategies.

18

They amount to 7.4% (N=262) of the announcements. Our subsequent analysis is divided into two

parts: first that of pure strategies within projects and then that of mixed strategies within projects.

Analysis of Pure Strategies Within Projects This section presents the results of the descriptive analysis, choice analysis and multivariate

analysis of payoffs.

Descriptive Analysis The descriptive statistics of all variables are in Online Appendix D. At the firm level, on

average, a firm in our sample has 5.3 makes, 6.6 buys, and 5 allys. One third of the firms (34% of

firms) predominantly use only one strategy. We define a predominant strategy if a firm uses the

strategy for 90% or more of its innovations. We find that within firms that predominantly use one

strategy, more firms use buy as a predominant strategy than firms that use either make or ally as a

predominant strategy: 23% of firms predominantly buy, 5.5% of firms predominantly make, and

4.9% of firms predominantly ally. 68% of firms chose one strategy for more than 50% of its

innovations. Thus, most firms tend to favor one of the three strategies.

Next we analyze the payoff to various strategies using the (-1, 1) window. We use this

window because we control for confounding events around this window. Table 3A shows the mean

results of the payoff to make, buy, and ally. Note that the payoff to buy is strongly negative and

significant while that to make and ally are strongly positive and significant.

The first objection to these results is that they are across all observations, including firms

who use buy sparely and those who buy intensely. The argument can be made that firms that use the

buy strategy intensely may be more adept at it and may be able to earn higher returns. Table 3B

shows the payoff to buy based on the intensity of use of the strategy: firms that mainly buy,

moderately buy, minimally buy, and never buy. Note that the payoff is negative and significant both

for firms that use buy moderately and those that mainly use the strategy.

19

The next objection to these results could be that firms that use a mixed make, buy and ally

strategy would do best because they can combine the advantages of each of these strategies:

developing innovations internally when they have the expertise, allying where they lack expertise,

and buying when allying and making are insufficient. Table 3C shows the mean payoff for mixed

strategies. Note that mostly ally or mostly make is better than any type of mixed make, buy, and ally.

The next objection could be that the above strategies do not take into account relatedness of

the target and the innovativeness of the target. Table 3D shows payoff of buy under various

conditions. Note that buy does best when the target is related, or when the benefits to customers are

high, or both. However, the mean results even in these circumstances are negative. This analysis

indicates that even though buy in general yields negative returns, those returns can be mitigated but

not eliminated by buying targets that are related and that have substantially better customer benefit.

A final objection could be that the above analyses are all descriptive and mostly univariate.

They do not take into account characteristics of the firm, selectivity bias in each strategy, and prior

risk from the strategies. To control for these other factors, we proceed to the multivariate analysis of

the payoff to strategies conditioned on the multinomial analysis of the choice of these strategies.

Analysis of Choice Table 4 contains the results of the Multinomial Logit Model with random intercepts. The

dependent variable is a nominal variable where the reference category is the choice of make that is

coded as 1. The choice of ally and buy is coded as 2 and 3 respectively. To assess the simultaneous

effect of the explanatory variables on the probabilities of make, buy, and ally,5 Table 4 reports the

marginal effects. The estimated coefficients of Equation 2 are in the Online Appendix E. We report

the marginal effects at the mean. The estimated parameters in Table 4 show the effect of the

explanatory variables on the probability of undertaking the innovation choice.

5 The formula for the marginal effect of an explanatory variable xp on choice c for firm i is: Pic * ( bpc - Pic

c=2

3

å *bpc )

20

We first turn to the effects of prior payoff to each strategy. Payoff to prior make is positively

and significantly (3.053, p<.05) associated with make. This result supports Hypothesis H1a.

However, both the marginal effects of payoff to prior buy on buy choice and of payoff to prior ally

on ally choice are not significant. Thus, we do not find support for both Hypothesis H1b and H1c.

These results imply that firms have a memory for the payoff to make but no memory for the payoff

to buy and ally6. We suggest possible explanations for this result in the discussion section.

The number of commercializations is positively and significantly associated with make (2.6,

p<.05) and significantly and negatively associated with buy (-4.76, p<.01). However, the number of

commercializations does not affect ally significantly. Thus, we find support for Hypotheses H2a and

H2b but not for Hypothesis H2c. This result suggests that firms with low commercializations have a

high likelihood to buy than make. Given the importance of new products and the long lead time to

produce them, buy seems like a quick fix for what may be a deep strategic problem.

A firm’s number of patents has no significant effect on the innovation choices. Thus, an

increasing number of patents do not seem to affect the likelihood of make, buy, or ally. Prior

number of make is positively and significantly associated with make (0.236, p<.05). Thus, firms with

a focus on make continue to make. Financing capability is positively and significantly associated with

buy (0.499, p<.05). This result indicates that cash rich firms have a propensity to buy. Highly

diversified firms have a negative and significant relation with ally (-0.07, p<05). This result suggests

that highly diversified firms do not use ally to learn new knowledge and diversify into new industries.

Prior number of ally and buy do not significantly affect ally and buy choice respectively. This result

along with the prior result of the non-significant effect of prior payoffs of ally and buy suggests that

6 The effects of prior make, buy, and ally on choice may differ if these outcomes are measured for longer time intervals.

We thus re-estimate the model for choice using longer time intervals that range from 2 to 5 years. Online Appendix F has the estimated coefficients of Equation 2 for 2 and 3 year intervals. Our results remain the same both in terms of direction and significance. Our results remain same using the value-weighted CRSP index (Online Appendix G).

21

firms forget both their ally and buy announcements and their payoffs. Multicollinearity is not an

issue in the model. The results of stepwise build-up are in Online Appendix H1.

Multivariate Analysis of Payoff We pool the payoffs for make, buy, and ally into a single model to analyze their differences

after controlling for other independent variables (see Table 5). For this purpose, we estimate

Equation 4 using a random effects model instead of the fixed effects model because we cannot

reject the null hypothesis that the random effects model is consistent and efficient7 (Hausman 1978).

In this model, the coefficient of the included buy and ally indicators denote the difference in payoff

from make (excluded level) after controlling for all other effects. We include interactions of buy with

other independent variables to test if these other effects vary by type of buy.

The coefficient of the buy indicator is negative and significant (-2.83%, p<.01, see Table 5).

Thus, buy leads to a negative and smaller payoff than make even after controlling for all other

variables, confirming the simple descriptive analysis. We find the coefficient of the interaction of

buy indicator and innovation relatedness is positive and significant (0.16%, p<.01). This result

indicates that firms can improve the payoff to buy if the target is related to it. We find the coefficient

of the interaction of the buy indicator and customer benefit is positive and significant (0.21%,

p<.05). This result indicates that firms can improve the payoff to buy if they buy targets with high

customer benefit. We find the coefficient of the interaction of the buy indicator and prior number of

buy is positive and significant (1.71%, p<.05). This result indicates that firms can achieve positive

payoffs to buy if they have prior buy experience. We find the coefficient of buying a target’s research

personnel is negative and significant (-0.58%, p<.05). This result indicates that firms should be wary

of buying only research personnel as integrating such personnel into the acquired firm may be

difficult due to the employee’s felt loss of independence.

7 Hausman Test :Chi-Square value (33) =31.27, p-value=0.5533

22

As a robustness check, we include the amount paid to acquire the target in the regression

model as it can affect the buy payoff. For example, firms may pay more than the target was worth to

them (Eccles et al. 1999). We could only obtain the buy amount for 320 out of 754 buys. Thus, we

drop 434 observations. Moreover, we could not calculate the acquisition premium for more than

80% of the buys as either the market value or the amount paid was not available. The results are in

the two rightmost columns of Table 5. We get similar results as our main model. Multicollinearity is

not an issue in the model. Results of the stepwise build-up are in Online Appendix H2. We find

similar results using Propensity Score Matching, which is an alternative to our modeling framework.

Analysis of Mixed Strategies within Projects The prior analysis deals entirely with pure strategies of make, buy, or ally within a project.

The argument could be made, that firms do best when they carry out a mix of make, buy, or ally

within a project. Do such within-project mixed strategies do better? We next address this question.

Each announcement in our data belongs to a unique innovation project. To identify mixed

strategies, we need to find out combinations of make, buy, and ally for the same project. Also, to

reduce left censoring8 in the identification of mixed strategies, we need to find out if another

announcement for the same project preceded the announcements in our sampling window. To

reduce left censoring at least to some extent, we collect announcements for 18 months before our

main window of time sampling. We then identify multiple announcements of make, buy, or ally

within a project.

We find that in 7.4% of cases (262 announcements), firms did use some mixture of make,

buy, and ally within a single project. These 262 announcements relate to 87 projects. Within these 87

projects, we identify strings of related announcements of make, buy, and ally. Note, as far as the

market is concerned, the first announcement in the string for a particular project, does not appear as

8 We need to be concerned only about left censoring because the market would know about any announcements within

a project prior to our sampling time. However, announcements after our sampling time would be unknown to the

market and would not affect returns within our sampling time.

23

a mixed strategy. Only the 2nd, 3rd, etc., announcements can be perceived as part of a mixed strategy.

We next proceed to analyze the payoffs to these subsequent announcements conditional on the first

announcement within a project. We use the Fama-French Carhart model to calculate the abnormal

returns and report the abnormal returns of the (0, 0) window.

Table 6 (Panel A) presents the number of announcements involving mixed strategies and the

average payoff of each combination. For any mixed strategy, except for one, we don’t find payoff to

subsequent announcements of make, buy and ally significant conditional on the first announcement.

The one exception is this, the payoff to make is 0.46% (p<.05, N=58) when a mixed strategy comes

first. Also, the last row of Table 6A shows that when make is part of a mixed strategy, returns are

significantly positive and better than if buy or ally were part of the mixed strategy.

Finally, we combine the pure strategies analyzed earlier with the mixed strategies analyzed

here and present the results in Panel 6B. Note, that our primary descriptive results remain the same

for all strategies as for the pure strategies: namely that make and ally yield significantly positive

returns while buy yields significantly negative returns (See Table 6, Panel B).

We do not pool the mixed strategies with the pure strategies and redo the logit and

regression analysis, because the number of mixed strategies is relatively small, the main descriptive

results do not change, and the analysis of the pooled data becomes extremely complex due to a vast

number of possible combinations.

Discussion

Firms constantly grapple with the question of whether to make, buy, or ally. They widely

pursue these strategies spending trillions of dollars. This study seeks to find out the pattern of make,

buy, and ally, the factors that drive this pattern, and the factors that drive the payoff to make, buy,

and ally. This section summarizes the findings, discusses some key issues, suggests implications for

practice, and lists some limitations.

24

Summary of Findings The key findings of the study are the following:

Make, buy, and ally are widely used as strategies to obtain innovations but buy is the most

prevalent followed by make and ally. Despite buy’s higher prevalence, make and ally generate a

significantly positive and much higher payoff than buy. Buy leads to a significantly negative

payoff of -0.28%. Make and Ally have a significantly higher payoff than buy even after

controlling for several explanatory variables, different estimation methods, and various

combinations of pure and mixed strategies.

Firms do not use their prior payoffs to buy as a factor in their subsequent buy choice. However,

they do use their prior payoffs to make in their subsequent make choice. This result suggests

that firms do not have any memory of or “learn from” their prior payoffs to buy but do

remember or “learn” of their prior payoffs to make.

The number of commercializations of innovations is negative and significantly associated with

buy choice. This result suggests that firms buy to compensate for a low level of

commercializations rather than to complement a high level of commercializations.

The negative returns to buy can be mitigated if targets are related to acquirers and have high

customer benefit and if acquirers have prior buy experience.

Discussion of Key Issues This section addresses 3 key questions emerging from the results: Why are make and ally

consistently better than buy? Why do not firms “learn” from prior payoffs to buy? Why do firms

buy when they lack commercializations?

Why Are Make and Ally Consistently Better Than Buy? Buy does worse than make or ally for several reasons. First, make and ally are relatively more

reversible and flexible than buy. Firms starting out on an internal development project can stop their

venture and change plans with little trouble. Similarly, firms in an alliance can easily exit from an

25

alliance if the partnership is not going well. However, buy is less flexible to change. When firms buy,

they incur high financial and management costs. These costs increase the firm’s separation pain that

translates to a potential lock-in (Kreutzer 2012). Second, compared to make and ally, buy has

numerous post-acquisition problems such as changes in the conventional routines of the acquirer,

clash of cultures between the acquirer and target, difficulties in integrating the financial and

information systems of the two firms, and inability to retain employees (Nahavandi and Malekzadeh

1988; Hitt et al. 2009). Third, by the time a firm determines the target is a good buy, many other

firms also make a similar determination and the price of the target increases to match or exceed its

potential value. Fourth, the expectations of buy might be lower because firms might buy to preempt

competition where the costs of not competing can be higher than the loss incurred. The success rate

of acquisitions is estimated to be around 17% (KPMG 1999). Some recent examples of acquisition

failures are HP’s $11 billion acquisition of Autonomy, eBay’s $2.6 billion acquisition of Skype and

Microsoft’s $6.3 billion acquisition of online-advertising company aQuantive (Duncan 2012; Reilly

2012). HP admitted that 87% of its investment in Autonomy was vacuous and wasted; eBay

struggled to profit from spending billions on Skype’s Voice over Internet Protocol (VOIP)

innovation while Microsoft took a $6.2 billion write-down on the aQuantive purchase (Reilly 2012).

Why Do Not Firms “Learn” from Prior Payoffs To Buy? Firms may not learn from their prior payoffs for several reasons.

First, firms might not consider payoffs to buy because the firm’s management might believe

that the costs of not buying the firm, before their rivals, can be higher than the potential loss.

Second, absence of commercializations may create pressure for firms to buy. This pressure

may lead firms to ignore or downplay their prior buys which led to poor payoffs. Indeed, our result

for prior commercializations shows that firms tend to buy when they lack commercializations, even

26

though the strategy seems not to pay off. These results suggest that firms see buy as a quick fix for

what may be a deep strategic problem.

Third, both theoretical and empirical evidence show that more often than not, buys are

followed by layoffs in both the parent and the acquired firm (Krishnan and Park 2002). Researchers

find that acquiring firms carry out workforce reductions to eliminate redundancies and realize

synergy gains from a buy (O’Shaughnessy and Flanagan 1997; Krishnan et al. 2007). Staff turnover

in both firms could facilitate forgetting as executives involved in prior buy may leave the firm. Such

forgetting usually has negative consequences. For example, in 1991, AT&T suffered major losses in

its acquisition of NCR. These losses were due to substantial overpayment and clashing corporate

cultures that eventually led to the spin-off of NCR (Ziegler 1996; Vijayan and Jacobs 1997).

However, it seems that AT&T did not apply its prior learning from the NCR buy (Haleblian and

Finkelstein 1999). In 1993, AT&T acquired McCaw Cellular possibly overpaying for a firm that

again had a different corporate culture (Holt 1993; Kufner 1994). Similarly, HP’s missteps in its

acquisitions could be attributed to its revolving door for leadership (Carey 2012).

Fourth, managers may forget payoffs to prior buys due to their vested interests. Top

executives may behave opportunistically and not care about the payoff to buy due to personal gains

(Trautwein, 1990). As acquisitions increase firm size, they often have a positive effect on a top

executive’s compensation and influence (Hitt et al. 2009). Moreover, if acquisitions are used as a

diversification strategy, it may also reduce the top executive’s employment risk (Hitt et al. 2009).

Malmendier and Tate 2008 find that overconfident CEOs overestimate their ability to generate

acquisition returns. The authors find that the acquisition returns are significantly more negative for

overconfident CEOs than for non-overconfident CEOs. Psychologists suggest that individuals are

particularly overconfident about outcomes, which are under their control and to which they are

highly committed, such as buys (Langer 1975; Weinstein 1980). These CEOs may neglect the prior

27

payoffs to buy to portray their decisions to be correct. A CEO’s integrity may be questioned if one is

not behaving in line with one’s beliefs (Steele 1988). Thus, the CEO may trivialize or deny the

negative payoffs in a process of self-justification (Festinger 1957).

Why Do Firms Buy When They Lack Commercializations? Commercializations are the culmination and probably the most important output of a firm’s

innovation project. A firm’s overall earnings depend on its commercializations (Hauser et al. 2006).

Firms with low commercializations view buy as a quicker means of obtaining commercializations

than make. Moreover, buying innovative firms enables the acquirer to gain access to the target’s

portfolio of current and future commercializable products. We posit that firms with low

commercializations most likely carry out a problemistic search (March and Simon 1958; Cyert and

March 1963). Problemistic search arises as a response to a problem the firm faces. It occurs when

the performance is below an aspiration level, is initially done near the problem symptom, is

dependent on the current state of the firm, and is goal oriented (Cyert and March 1963; Greve

2003). Thus, firms with low commercializations first search for solutions within. Because such firms

may not have resources to develop innovations, they buy innovative firms.

Implications for Managers This study has five implications for managers.

First, for innovations, firms should emphasize and announce both internal development and

alliances. Even though make involves sizeable investments with the payoff delayed and uncertain

and alliances involve partners’ behaving opportunistically, a firm’s make and ally are rewarded while

buy is punished. Consider the success of Apple using make. After Steve Jobs took the helm in 1997,

he revived innovation within the firm with a series of big, successful innovations. Apple entered the

markets of music, phone, and books outside its core area of computers. Notably, it did so primarily

if not solely through make (Time 2007; Yoffie and Slind 2008). Consider the success of Procter &

Gamble using ally. In 2001, when P&G was in a stage of declining growth, P&G realized that the

28

key to increase its rate of innovation success was through collaboration. Thus, P&G launched the

Connect+Development initiative that was based on and committed to partnership. P&G’s goal was

to ensure that 50% of its innovation would contain a substantial amount of external collaboration

(Huston and Sakkab 2006). The Connect+Develop (C+D) program comprises a global team that

hunts for solutions to its businesses needs by partnering with academia, small and medium

enterprises, global companies, individuals, NGOs and government labs. In 2009, P&G achieved

70% higher than average Net Present Value from its C+D projects. The total dollar value for a make

on average is USD $165 million and the total dollar value for an ally on average is USD $62 million.

Many firms turn to buy due to lack of commercializations. Consider AT&T’s strategy to

move from the traditional landline phone business into the promising broadband Internet and cable

business in the late 1990s. AT&T’s regular overpayment for its targets led the firm to high debt

loads and eventually a spin-off off both its wireless business and cable assets. A recent example is

HP’s dependency on acquisitions to obtain innovations. HP has lost billions of dollars in value by

purchasing Palm for its mobile software, EDS for its Information Technology services, and

Autonomy for its textual software. When screening Autonomy, HP had a choice between ally and

buy. However, it chose buy that eventually led to a mammoth write-down of USD $8.8 billion

(Worthen & Scheck 2013). The total dollar value for a buy on average is a negative USD $42 million.

Second, when considering a buy, firms should keep in mind their prior payoffs to buy.

Moeller et al. 2005 find that during the period 1980 to 2001, acquiring firms lost over USD $220

billion in dollar stock market returns over the (-2, +1) day window. Yet, firms have continued to buy

over the last decade (BCG 2012). Our empirical analysis suggests that firms extensively pursue buy

probably because they do not “learn” from prior negative payoffs. HP’s headlong dive into the

Autonomy deal and the subsequent debacle could be attributed to the firm’s forgetfulness of its

prior acquisition failures. In additional analysis, we find that the average abnormal stock returns in

29

the 3-day period surrounding HP’s three acquisition announcements of Compaq, EDS, and Palm is

-10.35%. In absolute dollar terms, HP lost on average around $6.6 billion dollars for these 3 buys.

Other scholars have echoed our explanation. Day’s (1994) quote nicely summarizes the

situation firms might find themselves in if they don’t develop adequate memory systems,

“Organizations without practical mechanisms to remember what has worked and why will have to

repeat their failures and rediscover their success formulas all over again”. De Holan and Philips 2004

propose that until knowledge is stored in organizational memory systems, it can be easily forgotten.

Thus, firms need to find the routines associated with the buy that led to the negative payoff. Some

successful acquirers are already using innovative methods to inquire and capture the lessons from

their prior buys. For example, a global chemical firm set up post-acquisition reviews such as open-

ended surveys and one-on-one interviews with the key players in the acquisition (Heimeriks et al.

2008). Another example concerns a global industrial conglomerate which uses a wiki-style "deal

room" to discuss and store prior acquisition processes (Heimeriks et al. 2008). A third example is

Dell’s strategy of developing an acquisition guidebook, which documents not only its prior

acquisition successes and failures but also best practices and mistakes of other firms (Enderle 2011).

If firms are ignoring prior payoffs to buy due to empire building executives, a firm’s board

needs to incorporate pay-for-performance compensation schemes tied to the stock price of the firm.

Such incentive schemes can make them more sensitive to the returns of the firm (Jensen and

Murphy 1990). As for overconfident CEOs, standard incentive contracts are unlikely to correct their

decision making as these CEOs think they are maximizing the firm’s value. Thus, independent

directors of firms could play an active role in the selection of targets (Malmendier and Tate 2008).

If firms are ignoring prior payoffs to buy due to the pressure to quickly commercialize

innovations, firms need to realize that rush to market may be a bad policy (Golder and Tellis 1993).

30

Indeed, one of Steve Jobs’ guiding principles was never to rush a product to market, but to get the

quality right. The iPad was not commercialized until it met his standard for quality (Golvin 2011).

Third, if a buy is warranted, firms can maximize the payoff if they acquire innovations

related to their current capabilities. When Nokia bought Sega.com in August 2003, Nokia had built

substantial capabilities in multiplayer gaming. The relatedness of SEGA’s innovation to Nokia,

allowed Nokia to garner USD $540 million in the 3-day window surrounding the announcement.

Fourth, if a buy is warranted, firms can maximize the payoff if they acquire innovations with

high customer benefit. Consider Parker Hannifin’s buy of Airtek in January 2007. Airteks’ drying

and filtration equipment for compressed air enabled Parker Hannifin’s global filtration

business customers to enjoy a complete compressed air treatment package starting from the

compressor to the point of use. The customer benefit of Airtek’s innovation allowed Parker

Hannifin to garner USD $319 million in the 3-day window surrounding the announcement.

Fifth, firms can obtain higher payoffs to buy by developing acquisition experience.

Experience enables firms to learn from their prior successes and failures (Levitt and March, 1988).

Consider the success of Cisco. Cisco’s success in its buys is attributed to its team of well-practiced

executives and well-tuned acquisition screening, selection, and integration process (Goldblatt 1999).

Limitations This study has several limitations which can be the basis of future research. First, we study

the announcements about making, buying, and allying innovations and the not the events per se. Indeed,

there could be events in these strategies that are not announced. It is extremely challenging and

complicated to obtain information about each make, buy, and ally and its outcome. So, a stream of

research in marketing, strategy, economics, and finance, does treat the announcement of make, buy, or

ally as equivalent to the event (see Table 1). Second, the data does not include firms, which are not

listed on the American Stock exchanges. Future research might explore whether the same results

31

hold for such firms. Third, we do not control for confounding events due to competitors’ activities.

No previous studies in marketing, management and finance have controlled for competitor events.

Although this limitation increases the noise in the data it does not bias any estimated coefficient.

Fourth, the primary driver for a make, buy or ally, may be the infeasibility of one or more of the

other options. Though, we do not explicitly model this situation, we partly account for it by

including prior choices and outcomes as independent variables in current choices. Fifth, we assume

the Efficient Market Hypothesis (EMH) for our empirical analysis. We acknowledge that the EMH

accepts that the amount of publicly available information about a firm or project can be limited and

vary across firms and projects. Sixth, though ally includes strategic alliances, joint ventures, and

licensing agreements, we concede that our definition of ally is a good though not a comprehensive

measure of open innovation. Seventh, we do not know whether all three alternatives appear in the

firm’s consideration set while making the choice and if the consideration set varies across decision

points. Data limitations prevent us from measuring consideration sets. We assume that our panel of

firms considers these choices at each point. Thus, our estimates may be considered as conservative.

32

Table 1: Papers on Make, Buy, Ally in 16 top business and economics journals* from Jan 2001-April 2013

* JM, JMR, Marketing Science, Management Science, SMJ, AMJ, Org. Science, Administrative Science Quarterly,

Journal of Management, Journal of Finance, Journal of Financial Economics, Review of Financial Studies,

American Economic Review, Journal of Political Economy, Review of Economic Studies, Quarterly Journal of

Economics

No. Author Year Journal Focus

of

Study

Event Derived

Measure

Is Event =

Announcement?

1

Louis K. C. Chan, Josef

Lakonishok, and

Theodore Sougiannis 2001

Journal of

Finance Make

Research and

Development

Expenditures

Buy and Hold

Returns No

2

Allan C. Eberhart,

William F. Maxwell

and Akhtar R. Siddique 2004

Journal of

Finance Make

Unexpected

R&D increases

Long Term

Abnormal Returns No

3

Namgyoo K. Park, John

M. Mezias and Jaeyong

Song 2004

Journal of

Management Ally

Technological

Alliances

Cumulative

Abnormal Returns No

4

Matthew J. Higgins and

Daniel Rodriguez 2006

Journal of

Financial

Economics Buy

New product-

focused

biotechnology

industry

acquisitions

Cumulative

Abnormal Returns No

5

Kartik Kalaignanam,

Venkatesh Shankar and

Rajan Varadarajan 2007

Management

Science Ally

New Product

Development

Alliances

Cumulative

Abnormal Returns No

6

David Benson and

Rosemarie H. Ziedonis 2008

Organization

Science Buy

Acquisitions of

Technology

Startups

Cumulative

Abnormal Returns No

7

Joanne E. Oxley,

Rachelle C. Sampson,

and Brian S. Silverman 2009

Management

Science Ally

Research and

development

alliances

Cumulative

Abnormal Returns No

8 Ashish Sood and

Gerard J. Tellis 2009

Marketing

Science

Make,

Buy and

Ally

grouped

into one

category

called

Alliance

s

Alliance

formation

including

acquisitions;

Make only

includes new

development or

manufacturing

facilities

Cumulative

Abnormal Returns No

9

Sam Ransbotham and

Sabyasachi Mitra 2010

Management

Science Buy

Acquisitions of

innovation Abnormal Returns No

10

Akbar Zaheer, Exequiel

Hernandez and Sanjay

Banerjee 2010

Organization

Science Buy

High-

technology

acquisitions

Cumulative

Abnormal Returns No

11

Joshua Sears and Glenn

Hoetker 2013

Strategic

Management

Journal Buy

Technological

acquisitions

Cumulative

Abnormal Returns No

33

Table 2. Rationale for Inclusion of Control Variables

Table 3A: Abnormal Returns to an Average Event by Strategy for (-1, +1) window

Strategy N Returns Sig. Level

Make 441 0.25% < 0.01

Buy 754 -0.28% < 0.05

Ally 394 0.32% < 0.05

Table 3B: Abnormal Returns to an Average Event by Strategy of Buy for (-1, +1) window Strategy Definition N Returns Sig. Level

Mainly Buy Firms that use buy as a strategy in 75% to

100% of their announcements

361 -0.09% < 0.01

Moderately Buy Firms that use buy as a strategy in 25% to

74% of their announcements

747 -0.1% < 0.01

Minimally Buy Firms that use buy as a strategy in < 25% of

their announcements

462 0.18% < 0.01

Never Buy Firms that don’t use buy as a strategy 19 1.68% < 0.05

Variable Rationale for Inclusion

Model

for

Choice

Managerial Capability Firms with capable managers may manage the post-acquisition process better than

firms with non-capable managers.

Financing Capability Firms with financing capability may spend the cash on acquisitions instead of offering

the cash to shareholders (Jensen 1986).

Number of Patents Firms with low number of patents can quickly and easily gain an entire portfolio of

patents and pending patent applications using buy and ally.

Prior No. of Make,

Buy, Ally

Firms may follow the same strategy used in the past due to inertia (Szulanski 1996).

Marketing Investments We use marketing investments to measure value appropriation. Value appropriating

firms may put more focus on ally and buy than make. R&D Investments We use R&D investments to measure value creation. Value creating firms may put

more focus on make than buy and ally. Diversification Levels Firms may use acquisitions as a diversification strategy.

Industry Choice of make, buy, and ally may differ by industry.

Model

for

Payoff

Relatedness Relatedness of the innovation to a firm’s current capabilities allows firms to recognize, assimilate and apply new information (Cohen and Levinthal 1990).

Customer Benefit Innovations that benefit customers may lead to higher sales, cash flow, earnings.

Marketing Investments High payoffs due to better advertising, branding, pricing, and distribution.

R&D Investments High payoffs due to better in-house R&D infrastructure and expertise.

Financing Capability High payoffs due to firm’s ability to finance, maintain, and finish projects.

Diversification High payoffs as diversified firms can absorb innovations unrelated to their core skills.

Prior Risk to Make,

Buy, and Ally

Risks involved in make, buy, and ally are different (e.g., buy is less risky than make because there are many factors involved in the success of a buy).

Location of make, buy,

ally

Location of innovation may affect payoff (e.g., setting up an R&D center in emerging markets may lead to lower payoffs due to the lack of established institutional systems). such activities. Type of Make, Buy, ally Payoffs might be influenced by the different types of make, buy, and ally.

Industry Payoff to strategies might differ by industry.

Competition Some firms’ may achieve higher payoff to buy since they use it to pre-empt competitors who are also interested in the same target (Dyer et al. 2004).

34

Table 3C: Abnormal Returns to an Average Event by Strategy Group for (-1, +1) window

Table 3D: Abnormal Returns to an Average Event by Type of Buy for (-1, +1) window Type of Buy Returns Sig. Level

Low Relatedness -0.60% <0.05

High Relatedness -0.03% <0.05

Low Customer Benefit -0.53% <0.05

High Customer Benefit -0.01% <0.05 We use median split to determine low and high relatedness/customer benefit.

Table 4: Results of Multinomial Model of Choice - Marginal Effects

The symbols * and ** denote statistical significance at the 0.05 and 0.01 levels; Log Likelihood Value = -3090.75

Strategy Group Definition N Returns Sig. Level

Mostly Ally Firms that use ally in > 25% of their announcements & use

each of the other two in < 25% of their announcements

74 1.74% < 0.01

Mostly Make Firms that use make in > 25% of their announcements & use

each of the other two in < 25% of their announcements

60 0.5% < 0.01

Mostly Make and

Ally

Firms that use make & ally in > 25% of their

announcements & use buy in < 25% of their announcements

347 0.19% < 0.05

Mostly Make and

Buy

Firms that use make & buy in > 25% of their

announcements & use ally in < 25% of their announcements

241 0.07% < 0.05

Mostly Triple Play Firms that use each of the three strategies in > 25% of their

announcements

234 -0.001% < 0.05

Mostly Buy Firms that use buy in > 25% of their announcements & use

each of the other two in < 25% of their announcements

408 -0.09% < 0.05

Mostly Buy and

Ally

Firms that use buy & ally in > 25% of their announcements

& use make in < 25% of their announcements

225 -0.33% < 0.05

Make Buy Ally

Independent Variables Coefficient t-val Coefficient t-val Coefficient t-val

Prior Make Payoff 3.053* 2.51 -3.17** 2.9 0.125 0.19

Prior Buy Payoff 0.679 1.1 -0.193 0.29 -0.485 1.12

Prior Ally Payoff 0.768 1.19 -0.57 0.83 -0.192 0.54

Number of Patents -0.001 0.07 -0.005 0.56 0.006 1.29

Number of Commercializations 2.956* 2.02 -4.76** 2.78 1.722 1.82

Managerial Capability -0.005 0.65 0.006 0.59 -0.001 0.11

Financing Capability -0.323 1.41 0.499* 1.98 -0.174 1.27

Prior Number of Make 0.236* 2.15 -0.139 1.69 -0.096 1.18

Prior Number of Buy -0.05 0.5 0.125 0.98 -0.067 0.98

Prior Number of Ally -0.156 1.46 0.142 1.22 0.014 0.24

Marketing Investment 0.038 0.85 -0.019 0.39 -0.017 0.55

R&D Investment 0.012 0.53 0.009 0.35 -0.021 1.2

Related-Diversified Firms -0.006 0.1 0.012 0.16 -0.006 0.24

Unrelated-Diversified firms 0.094 1.23 -0.062 0.8 -0.03 0.96

High Diversified Firms -0.005 0.07 0.083 0.96 -0.07* 2.01

B2B goods 0.7 1.46 -0.05 1.45 0.011 0.77

B2C goods 0.279** 2.65 -0.3** 2.67 0.132 1.91

B2C services 0.178* 2.52 -0.31** 2.89 0.14* 2.03

Intercept -0.372* 2.31 0.426** 3.02 -0.05 0.72

35

Table 5: Results of Regression Model of Payoff (Dependent Variable: Returns)

The symbols * and ** denote statistical significance at the 0.05 and 0.01 levels; Note our results regarding the main

effects and interactions remain same when we impute the missing value of the target value with the grand mean.

Results of the Propensity Score Matching, an alternative estimator, are in Online Appendix I.

Model without Target

Value

Model with Target Value

Independent Variables Coefficient z-stat Coefficient z-stat

Buy Indicator -2.83%** -3.46 -4.48%** -4.37 Ally Indicator 0.16% 0.35 0.10% 0.23

Innovation Relatedness -0.03% -0.67 -0.02% -0.54 Customer Benefit -0.02% -0.55 -0.02% -0.48

Prior Number of Buy -1.18% -1.4 -1.33% -1.64 Buy Indicator*Innovation Relatedness 0.16%** 2.66 0.26%** 3.14

Buy Indicator*Customer Benefit 0.21%* 2.46 0.31%* 1.96 Buy Indicator*Prior Number of Buy 1.71%* 1.96 2.29%* 2.09

Marketing Investment 0.12% 0.84 0.16% 0.87 R&D Investment 0.05% 0.69 0.04% 0.47

Financing Capability 0.24% 0.25 1.20% 1.05 Related-Diversified Firms 0.03% 0.15 -0.04% -0.17

Unrelated-Diversified firms 0.00% -0.02 -0.07% -0.33 High Diversified Firms -0.20% -1.15 -0.29% -1.26

Prior risk to Buy 0.11% 0.46 0.25% 0.74 Prior risk to Ally -0.30% -0.88 -0.32% -0.84

Prior risk to Make 0.21% 0.67 0.04% 0.12 Buy– Emerging Markets 0.17% 0.78 0.29% 0.78

Ally – Emerging Markets -0.31% -0.84 -0.30% -0.82 Make – Emerging Markets -0.09% -0.43 -0.09% -0.41

Buy_Target_Product -0.50% -1.28 -0.15% -0.25 Buy_Target_Software -0.04% -0.19 -0.37% -0.98

Buy_Target_Research_Personnel -0.58%* -1.99 -1.90%** -3.15 Buy_Target_Patent -0.77% -1.35 -0.56% -0.53 Ally_Joint_Venture -0.61% -1.43 -0.56% -1.31

Ally_Licensing -0.55% -0.95 -0.51% -0.88 Make_R&D_center 0.02% 0.04 0.02% 0.04 Make_New_Project -0.05% -0.1 -0.07% -0.15 Make_New_Entity -0.35% -0.86 -0.36% -0.89

B2B goods -0.46% -1.42 -0.52% -1.12 B2C goods -0.58% -1.73 -0.69% -1.56

B2C services -0.74%* -1.97 -0.72% -1.45 Competition -0.06% -0.22 -0.27% -0.81

Target Value 0.00% -1.6 Intercept 1.42% 1.89 1.61% 1.83

Fit Statistics Overall R-Square: .043 Overall R-Square: .056 N Make=441, Buy =754,

Ally=394

Make=441, Buy =320,

Ally=394

36

Table 6: Analysis of Mixed Strategies within Projects

Panel A: Average Abnormal Returns# to Subsequent

Announcements on Event Day Conditional On Make Buy Ally

Make comes first -0.44% (N=14) -0.15% (N=9) -0.69% (N=9)

Buy comes first 0.34% (N=34) 0.62% (N=5) -0.95% (N=2)

Ally comes first 0.03% (N=25) 0.86% (N=2) -0.52% (N=5)

Mixed+ comes first 0.46%* (N=58) 0.79% (N=4) -0.25% (N=10)

Total 0.25%* (N=131) 0.33% (N=20) -0.51% (N=26)

Panel B: Average Abnormal Returns# to

Announcements on Event Day Projects Make Buy Ally

All (Pure and Mixed) 0.09%* (N=1174) -0.08%* (N=1331) 0.13%* (N=1017)

Pure only 0.09%* (N=1022) -0.1%* (N=1276) 0.14%* (N=962) Legend:

N is the number of announcements

* and ** denote statistical significance at the 0.05 and 0.01 levels;

+ mixed strategy means two different types of announcements preceded the target announcement

# for the (0,0) event window

Figure 1: Events during Initiation, Development, and Commercialization Activities of Innovation Project

37

Figure 2: Conceptual Framework: Drivers of Choice of Make, Buy, or Ally

Figure 3: Conceptual Framework: Drivers of Payoff to Make, Buy, or Ally

38

References Ahuja, G., Katila, R. 2001. Technological Acquisitions and the Innovation Performance of Acquiring Firms:

A Longitudinal Study. Strategic Management Journal, 22(3): 197-220 Bayus, B. 2013. Crowdsourcing New Product Ideas Over Time: An Analysis of the Dell IdeaStorm

Community. Management Science, 59 (January), 226-244. Booz & Co. 2012. The 2012 Global Innovation 1000 Study. Accessed Oct 20, 2012, booz.com. BCG 2012. How M&A Can Grow Portfolio Value. Plant and Prune. Annual M&A report. Bleeke, J., & Ernst, D. 2002. Is your strategic alliance really a sale. Strategy: critical perspectives on business and

management. Vol. 4, 4, 214. Boulding, W., Staelin, R. 1995. Identifying Generalizable Effects of Strategic Actions of Firm Performance:

The Case of Demand-Side Returns to R&D Spending, Marketing Science,14, 222-3 Campbell, J. Y., A. W. Lo, A. C. McKinley. 1997. The Econometrics of Financial Markets. Princeton University

Press, Princeton, NJ. Capron, Laurence, Will Mitchell. 2010. Finding the right path. HBR, 88.7-8, 102 --- , 2012. Build, Borrow, Or Buy: Solving the Growth Dilemma. Harvard Business Press. Carey, P. 2012. HP's acquisition misstep far from the first. San Jose Mercury News. Carhart, M. 1997. On the Persistence of Mutual Fund Performance. Journal of Finance 52, 57–82 Cefis, E., Marsili, O. 2006. Survivor: The Role of Innovation in Firm’s Survival. Res. Policy, 35, 626 Chandy R, Tellis GJ. 2000. The incumbents curse: incumbency, size, and radical product innovation. Journal of

Marketing; 64:1–17. Chandy, R., Hopstaken, B., Narasimhan, O., & Prabhu, J. 2006. From invention to innovation: Conversion

ability in product development. Journal of Marketing Research, 43(3), 494−508 Chen, Y., Liu, Y., Zhang, J. 2012. When do third-party product reviews affect firm value and what can firms

do? The case of media critics and professional movie reviews. Journal of Marketing, 76(2), 116-134. Chesbrough, H. 2003. Open Innovation: The New Imperative for Creating and Profiting from Technology. Harvard

Business School Press: Boston, MA. Cisco Systems Canada Co. (2009). Build, Buy or Ally Growth Strategies for Business. Presentation by Joe Deklic,

Vice President, Strategic Investments Group, Cisco Systems Canada Co. Retrieved 25th June, 2013 from http://www.slideshare.net

Cohen, W., D. Levinthal 1989. Innovation and learning: The two faces of R&D. The Econ. Journal, 99 p569Cyert, R.M., March, J.G. 1963. A behavioral theory of the firm. Englewood Cliff, NJ: Prentice Hall.

Day, G. 1994. The Capabilities of Market-Driven Organization. Journal of Marketing, 58 p37 De Holan, M. P., N. Phillips. 2004. Remembrance of Things Past? The Dynamics of Organizational

Forgetting. Management Science 50(11): 1603–1613. Doukas, John, Lorne Switzer 1992. The Stock market’s Valuation of R&D Spending and Market

Concentration. Journal of Economics and Business, 44, 95-114 Duncan, G. 2012. It’s been a year: What has Microsoft done with Skype? Accessed Jan 23, 2013.

http://www.digitaltrends.com Dyer, J. H., Kale, P., Singh, H. 2004. When to ally and when to acquire. HBR, 82(7-8), p108. Elberse, A. 2007. The power of stars: Do star actors drive the success of movies? J. of Marketing, 102-120. Enderle, R. 2011. Dell's Success vs. Microsoft's and Google's Shared Problem. Datamation Erickson G., R. Jacobson 1992. Gaining comparative advantage through discretionary expenditures: The

returns to R and D and advertising. Management Science 38 (9), pp. 1264– 1279 Fama, E., L. Fisher, M. C. Jensen, R. Roll. 1969. The adjustment of stock prices to new information.

International Economic Review 10(1) 1–21. ---, 1991. Efficient Markets: II. Journal of Finance, 46 1575-1617. ---, K. French 1993. Common risk factors in the returns on stocks and bonds. Jrnl. of Fin. Economics 33: 3-56. --- 1998. Market efficiency, long-term returns, and behavioral finance. Journal of Finance, 49, 283. Festinger, L. 1957. A theory of cognitive dissonance. Oxford, England: Row, Peterson. Geyskens, Inge, Jan-Benedict E.M. Steenkamp, Nirmalya Kumar 2006. Make, buy, or Ally: A transaction cost

theory meta-analysis. Academy of Management Journal 49 (3), p519

39

Geyskens, I., Gielens, K., & Dekimpe, M. G. 2002. The market valuation of internet channel additions. Journal of Marketing, 102-119.

Goldblatt, H. 1999. Forty-two acquisitions and counting: Cisco's secrets. Fortune, 177-184 Golder, P. 2000. Historical Method in Marketing Research with New Evidence on Long-term Market Share

Stability. Journal of Marketing Research 37 (May) 156-72 Golder, P., Tellis, G.J., 1993. Pioneering Advantage: Marketing Fact or Marketing Legend. Journal of Marketing

Research, 30 (May), 158-170 Golvin, C. 2011. Product Strategists: Strive To Be Like Steve. Accessed Jan 7th 2012. http://blogs.forrester.com/ Gomes-Casseres, B. 2000. Alliances and risk: securing a place in the victory parade. Financial Times. pp. 6-7 Greene, W.H. 2003. Econometric analysis, 5th edition. Prentice Hall, NJ Greve, Henrich R. 2003. Organizational Learning from Performance Feedback: A Behavioral Perspective on Innovation

and Change. Cambridge, UK: Cambridge University Press Griffin, Abbie 2002. Product Development Cycle Time for Business-to-Business Products. Industrial Marketing

Management, 31:2, 291-304, March Hagel III, J., Brown, J.S. 2011. Creation nets: harnessing the potential of open innovation. Journal of Service

Science. Vol 1, Issue 2, Pages 27-40 Haleblian, J., Finkelstein, S. 1999. The Influence of Organization Acquisition Experience on Acquisition

Performance: A Behavioral Learning Perspective. ASQ, 44, 29-56 Haleblian, J., Kim, J., Rajagopalan, N.2006. Routine-based learning, performance-based learning, and their

interaction: evidence from the U.S. Commercial banking industry. Acad. of Management Journal, 49. 357-370. Hauser, J., Tellis, G. J., Griffin, A. 2005. Research on Innovation: A Review and Agenda for Marketing

Science. Marketing Science, 25. Hausman, J. 1978. Specification Tests in Econometrics. Econometrica, vol. 46(6), 1273-1291 Heckman, J.J. 1979. Sample Selection Bias as a Specification Error. Econometrica 47, p153 Heimeriks, K., S. Gates, M. Zollo 2008. The secrets of successful acquisitions. Sloan Management Review,

Business Insight (WSJ), R9: September Hitt, M. A., Hoskisson, R. E., Johnson, R. A., Moesel, D. D. 1996. The Market for Corporate Control and

Firm Innovation. Academy of Management Journal, 39(Oct), pp. 1084-1119 Hitt, M.A., King, D., Krishnan, H., Makri, M., Schijven, M., Shimizu, K., Zhu, H., 2009. M&A: Overcoming

Pitfalls, Building Synergy and Creating Value. Business Horizons, 2009, 52-523-529 Holt, H. 1993. AT&T: Dialing for dollars. Financial World, 162: 17-18. Huston, L., Sakkab, N. 2006. Connect and Develop: Inside Procter & Gamble’s new model for innovation.

Harvard Business Review, 84(3): 58–66. Jegadeesh, N., Titman, S. 1993. Returns to Buying Winners and Selling Losers: Implications for Stock Market

Efficiency. Journal of Finance, 48, pp. 65–91 Jensen, M. C. 1986. The Agency Costs of Free Cash Flow: Corporate Finance and Takeovers. AER, 76 (2) Jensen, M., Murphy, K., 1990. Performance pay and top-management incentives. J. of Political Econ. 98, p225 Joshi, A. M., Hanssens, D. M. 2009. Movie advertising and the stock market valuation of studios: a case of

“great expectations?” Marketing Science, 28(2), 239-250. Kanter, R. M. 1999. From Spare Change to Real Change: The Social Sector as Beta Site for Business

Innovation. Harvard Business Review, 77(3): 123:132 Karniouchina, E. V., Uslay, C., Erenburg, G. 2011. Do marketing media have life cycles? The case of product

placement in movies. Journal of Marketing, 75(3), 27-48. Kelm, K. M., Narayanan, V. K., & Pinches, G. E. 1995. Shareholder value creation during R&D innovation

and commercialization stages. Acad. of Management Jrl, 38(3), 770-786. Kessler, E. H., Chakrabarti, A. K. 1996. Innovation speed: A conceptual model of context, antecedents, and

outcomes. Academy of Management Review, 21: 1143-1191.

King, D. R., Dalton, D. R., Daily, C. M., & Covin, J. G. 2004. Meta‐analyses of post‐acquisition performance: Indications of unidentified moderators. Strategic Management Journal, 25(2), 187-200.

KPMG 1999. Unlocking shareholder value: The keys to success. Report. KPMG 2013. Strategic alliances: Joint ventures as a way into new markets. Report. Kufner, Andrew 1994. AT&T's 12 billion dollar cellular dream. Fortune, 130: 100-1 06.

40

Kreutzer, M. 2012. Selecting the right growth mechanism: The choice between internal development, strategic alliances, and mergers & acquisitions. Balanced Growth. Springer, 77-94.

Krishnan H.A., Hitt M.A., Park D. 2007. Acquisition Premiums, Subsequent Workplace Reductions and Post-Acquisition Performance. Journal of Management Studies 44: p709

__, Park D. 2002. The impact of work force reduction on subsequent performance in major mergers and acquisitions: an exploratory study. Journal of Business Research. 55(4), p285

Kuss, O., McLerran, D. 2007. A note on the estimation of the multinomial logistic model with correlated responses in SAS, Computer Methods and Programs in Biomedicine. 87, p 262

Langer, E., 1975. The illusion of control. Jrnl. of Personality and Social Psychology 32, p311 Lee, L. 1982. Some Approaches to the Correction of Selectivity Bias. Review of Econ. Std., 49, 355 Lev B. 2001. Intangibles-management, measurement and reporting. Brookings Institution Press, DC Levitt B., March J. G. 1988. Organizational learning. Annual Review of Sociology, 14, p319 Lichtenthaler, U., 2011. Open innovation: past research, current debates, and future directions. Academy of

Management Perspectives, 25(1), pp. 75–93. Luo, X. 2009. Quantifying the long-term impact of negative word of mouth on cash flows and stock

prices. Marketing Science, 28(1), 148-165. March, J.G. 1981. Footnotes to organizational change. ASQ, 26 563-577 March, J. ,Simon H.A. 1958. Organizations. Wiley-Blackwell. 1st edition Malmendier U., Tate G. 2008. Who Makes Acquisitions? CEO Overconfidence and the Market’s Reaction.

Journal of Financial Economics, vol. 89(1), pp. 20-43 Mayer, D., Kenney, M. 2004. Ecosystems and Acquisition Management: Understanding Cisco's Strategy.

Industry and Innovation 11, (4): 299-326 McClusky, M. 2009. The Nike Experiment: How the Shoe Giant Unleashed the Power of Personal Metrics. Accessed

Dec 7th 2012. http://www.wired.com/ McWilliams, A., Siegel, D. 1997. Event Studies in Management Research: Theoretical and Empirical Issues.

Academy of Management Journal, Vol. 40, No. 3, June, pp. 626-657. Miller, D., Chen M.J. 1994. Sources and Consequences of Competitive Inertia: A Study of the U.S. Airline

Industry Administrative Science Quarterly Vol. 39, No. 1. (1994), pp. 1-23. Mizik, N., Jacobson R. 2003. Trading Off between Value Creation and Value Appropriation: The Financial

Implications of Shifts in Strategic Emphasis. Journal of Marketing, 67, p63 Moeller SB, Schinglemann FP, Stulz RM. 2005. Wealth destruction on a massive scale? A study of acquiring-

firm returns in the recent merger wave. Journal of Finance , 60, p757 Nahavandi, A. & Malekzadeh, A.R. 1988. Acculturation in mergers and acquisitions. AMR, 13(1): 79-90 Nakamura, L. I. 2001. What is the US gross investment in intangibles? (At least) one trillion dollars a year!

Technical report, Federal Reserve Bank of Philadelphia No. 01-15 Nijman, T.E., Verbeek, M.J.C.M. 1992. Testing for selectivity in panel data models. International Economic

Review, 33(3), 681-703. O'Shaughnessy K. C., Flanagan, David J. 1998. Determinants of layoff announcements following M&As: an

empirical investigation. Strategic Management Journal. 19, 10, p989 Parkhe, A. 1993. Strategic alliance structuring: A game theoretic and transaction cost examination of interfirm

cooperation. Academy of Management Journal, 36, 794–829. Prabhu, Jaideep, Chandy, Rajesh, Ellis Mark. 2005. Acquisition and Innovation in High-Tech Firms: Poison

Pill, Placebo, or Tonic? Journal of Marketing, 69(1), 114-130. Ranft, A., Lord, M. 2000. Acquiring new knowledge: The role of retaining human capital in acquisitions of

high-tech firms. Journal of High Technology Management Research, 11: 295 Reilly, D. 2012. 'Tis No Season for Goodwill to Investors. Wall Street Journal. Nov 24, 2012, B16. Roberts, E., Berry, C. 1985. Entering new businesses: Selecting strategies for success. Sloan Management Review, vol. 36 no. 3, pp. 3-17. Rothaermel, F., & Hess, A. 2010. Innovation strategies combined. MIT Sloan Management Review, 51(3). Shaver, J. M. 1998. Accounting for Endogeneity when Assessing Strategy Performance: Does Entry Mode

Choice Affect FDI Survival. Management Science, 44(4): 571-585

41

Sharpe, W.F. 1964. Capital asset prices: A theory of market equilibrium under conditions of risk. J. of Finance 19, 425-442.

Sood, A., Tellis. G. J. 2009. Do Innovations Really Payoff? Total Stock Market Returns to Innovation. Marketing Science May/Jun, 28 (3) p. 442 (16 pages).

Sorescu, A., Chandy, R., Prabhu, J.2007. Why Some Acquisitions Do Better than Others: Product Capital as a Driver of Long-term Stock Returns. Journal of Marketing Research, 44(1), 56-72

Srinivasan, R., Bharadwaj. S. 2004. Event Studies in Marketing Strategy Research. Assessing Marketing Strategy Performance (eds.) C. Moorman , D. Lehmann, Marketing Science Institute

Srinivasan, S., Hanssens. D. M., 2009. Marketing and Firm Value: Metrics, Methods, Findings, and Future Directions. Journal of Marketing Research, Volume 46, Number 3, June.

Steele, C. M. 1988. The psychology of self-affirmation: Sustaining the integrity of the self. In L. Berkowitz (Ed.), Advances in experimental social psychology 21, 261–302. NY: Academic Press.

Swaminathan, V., Moorman, C. 2009. Marketing alliances, firm networks, and firm value creation. Journal of Marketing, 73(5), 52-69.

Szulanski, G. 1996. Exploring internal stickiness: Impediments to the transfer of best practice within the firm. Strategic Management Journal, 17: 27–42.

Tellis, G., Johnson, J. 2007. The value of quality. Marketing Science, 26 (6), 758-73 Tellis, G. J., Prabhu, J. C., Chandy, R. K. 2009. Radical innovation across nations: The preeminence of

corporate culture. Journal of Marketing, 73(1), 3–23. Time Magazine 2007. The Apple of your ear. January 22, Vol. 169 No. 4 Trautwein, F. 1990. Merger motives and merger prescriptions. Strategic Mgmt. Journal, 11(4) p 283 Varadarajan, R., Vasudevan, R. 1987. Diversification and performance: A reexamination using a new two

dimensional conceptualization of diversity in firms. Academy of Mgmt. Journal, 30, p 380 Vella, F. 1998. Estimating Models with Sample Selection Bias: A Survey. Journal of Human Resources, vol. 33, no. 1, 127-167 Verbeek, M., & Nijman, T. 1992. Testing for selectivity bias in panel data models. International Econ Review,

681-703. Vijayan, J., Jacobs, A. 1997. Warning signs of a merger gone bad. Computer World, 31: 17-18. Weinstein, N. 1980. Unrealistic optimism about future life events. J. of Pers. & Soc. Psych. 39, 806 Wiles, M. A., Morgan, N. A., & Rego, L. L. 2012. The effect of brand acquisition and disposal on stock

returns. Journal of Marketing, 76(1), 38-58. Worthen, B., Scheck, J. 2013. Inside H-P's Missed Chance to Avoid a Disastrous Deal. Wall Street Journal.

Accessed Online on January 24th 2013, http://online.wsj.com West, J., Vanhaverbeke, W. , Chesbrough, H. 2006. Open innovation: a research Agenda. In: Chesbrough, H.,

Vanhaverbeke, W. and West, J. (eds), Open Innovation: Researching a New Paradigm. Oxford: Oxford University Press, pp. 285–307.

Wiles, M. A., Jain, S. P., Mishra, S., Lindsey, C. 2010. Stock market response to regulatory reports of deceptive advertising: the moderating effect of omission bias and firm reputation. Marketing Sci, 29(5), 828

Williamson, O.E.1975. Markets and Hierarchies. Free Press, NY Wooldridge, J. M. 1995. Selection Corrections for Panel Data Models under Conditional Mean Independence

Assumptions. Journal of Econometrics, 68, 115-132 Ybarra, C.E., Turk, T.A. 2011. Strategic Alliances with Competing Firms and Shareholder Value. Journal of

Management and Marketing Research, Vol. 6. Yoffie, D. B., M. Slind 2008. Apple Inc., 2008. Harvard Business School Case Studies, 708-480, 2-32 Ziegler, R. 1996. NCR gets another new lease on life. Wall Street Journal, December 30: B1.

42

Online Appendix A – Details of Method Section

A1: Operationalization of Control Variables

Panel A: Variables in both the Model for Choice and Payoff

Variable Source Definition/Operationalization Managerial Capability Compustat Tobin’s q: Ratio of the market value of a firm to its total

assets. Market value of the firm equals the market value of

common equity plus the book value of preferred stock plus

the book value of total debt (t-1).

Financing Capability Compustat Free Cash Flow: Operating cash flow minus capital

expenditures. We normalize this measure by dividing the

free cash flow by the total assets (t-1). Number of Patents National Bureau of

Economic

Research patent

database (Hall et al.

2001)

Number of patents granted in the year prior to the

announcement date. As we have a range of industries, we

standardize this measure by industry. We use the NBER

(National Bureau of Economic Research) patent database

to collect patents for all firms for every 4 digit SIC

(Standard Industrial Classification) code in our sample.

Prior Number of Make,

Prior number of Buy,

Prior Number of Ally

Capital IQ, Factiva

and Lexis-Nexis

Number of make decisions per year prior to the date of the

current event. We use a window of one year for calculating

this variable. In order to control for size, we normalize this

measure by the total sales in year t-1. We use the same

method as above to measure prior number of buy and ally.

Marketing Investments Compustat Ratio of selling and general administrative expenses to the

total assets (t-1) standardized by each SIC code. It is set to

0 when SG&A expense is missing.

R&D Investments Compustat Ratio of research and development expenses to the total

assets (t-1) standardized by each SIC code. It is set to 0

when R&D expense is missing.

Diversification Levels:

Low Diversified Firms,

Related-diversified firms,

Unrelated-diversified

firms, & High diversified

Firms. Base case: Low

Diversified Firms

Compustat Four different categories of diversification based on two

broad patterns of diversification (Varadarajan and

Ramanujan 1987). Broad spectrum diversification (BSD):

number of two-digit SIC codes in which a firm operates.

Mean narrow spectrum diversification (MNSD): number

of four digit SIC codes in which a firm operates divided by

number of two-digit SIC categories in which it operates.

Industry Occupational

Safety and Health

Administration

(OSHA)

Use four-digit SIC codes to categorize firms into Business

to Business (B2B) goods, B2B services, Business to

Consumer (B2C) goods, B2C services (Palepu 1985;

Srinivasan et al. 2011)

43

Panel B: Variables only in Model for Payoff Innovation Relatedness Capital IQ, Factiva

and Lexis-Nexis,

SDC

Schema of innovation relatedness expressing increasing

relatedness, on a 10 point scale ranging from 1 to 10. Scale

coded by two research assistants. The research assistants did

not know the study’s objectives. (See Online Appendix B8.1

below for Schema). (Kappa = .66) Customer Benefit Capital IQ, Factiva

and Lexis-Nexis,

SDC

Schema of customer benefit expressing increasing customer

benefit, on a 10 point scale ranging from 1 to 10. Scale

coded by two research assistants. The research assistants did

not know the study’s objectives. (See Online Appendix B8.2

below for Schema). (Kappa = .72)

Prior risk to Make, Prior

risk to Buy, Prior risk to

Ally

CRSP, Kenneth

French’s website Coefficient of i1β in equation (2), i.e., the Carhart Four-

Factor model for computing payoffs. First, we estimate prior

event i1β for each firm using 265 days of daily returns

ending 1 day before the event day. Next, we estimate post

event i1β for each firm using 265 days of daily returns

starting 1 day before the event day. The change in systematic

risk ( i1βΔ ) attributed to the make announcement is the

difference between the pre and post event systematic risk.

We measure this variable by averaging the systematic risks to

the firm’s make, buy, and ally per year prior to the date of

the current event. For robustness, we also use alternative

end and start day of 10 days before the event day.

Competition Compustat For each firms’ primary SIC industry, we square each firm’s market share and take the sum over all firms. We next subtract this sum from 1 (Fang et al. 2008).

Emerging Markets Capital IQ, Factiva

and Lexis-Nexis,

SDC

If the announcement indicated that the innovation was going to be carried out in an emerging market. We used the summary list provided in Wikipedia (http://en.wikipedia.org/wiki/Emerging_markets). We identified a country as en emerging market if the country appeared in the “Summary list”.

Types of Buy – Target’s

product, Target’s

software, Target’s

research personnel,

Target’s patent

Capital IQ, Factiva

and Lexis-Nexis,

SDC

We read the text of the announcement to identify if the target’s product, software, research personnel, patents were bought in the acquisition. We indicated the presence of a buy type by a dummy variable.

Types of Make – R&D

center, New process,

New entity, New

product

Capital IQ, Factiva

and Lexis-Nexis,

SDC

We read the text of the announcement to identify if the innovation involved the development of an R&D center, the incorporation of a new process, the start of a new entity to develop product or the start of a new project to develop a product. We indicated the presence of a make type by a dummy variable.

44

Types of Ally – Strategic

Alliance, Joint Venture

Capital IQ, Factiva

and Lexis-Nexis,

SDC

We read the text of the announcement to identify if the innovation was a strategic alliance or a joint venture. We indicated the presence of a joint venture by a dummy variable.

Licensing Agreement Capital IQ, Factiva

and Lexis-Nexis,

SDC

We read the text of the announcement to identify if the firm in our sample licensed an innovation from another firm. We only categorize alliances as licensing agreements if the firm in our sample was a licensee rather than a licensor. We indicated the presence of a licensing agreement by a dummy variable.

Target Value Capital IQ, Factiva

and Lexis-Nexis,

SDC

We read the text of the announcement to capture the amount the acquirer paid to the buy the target. We use the US $ to determine the value. Thus, target value in currencies other than US $ were converted to US $ using the exchange rate on the acquisition date. We used http://www.oanda.com/currency/historical-rates/ for the conversion.

A2. List of Firms in Study

The list is in the url: http://bit.ly/16LHhfp

A3. Database Descriptions

Capital IQ, a division of Standard & Poor, contains comprehensive information on firms, industries, and markets worldwide. Capital IQ has a Key Developments feature. This feature provides categorized news and corporate event data. Experienced data analysts continuously monitor, aggregate and tag information from over 20,000 news sources in addition to regulatory filings, transcripts, investor presentations and company websites. The Key Developments feature tracks over 100 company event types including SEC inquiries, expansions, reorganizations, client announcements, corporate governance, dividends, earnings, transaction milestones and rumors, executive/board changes, litigations, labor relations, product announcements, investor relations, strategic alliances, etc.

Factiva, a division of Dow Jones & Company, provides access to more than 14,000 sources (such as newspapers, journals, magazines, news and radio transcripts, etc.) from 152 countries in 22 languages, including more than 120 continuously updated newswires.

LexisNexis® Academic, a division of Reed Elsevier Inc., provides access to full-text news, business, and legal publications providing access to over 6,000 news, business, and legal sources.

SDC Platinum is the industry standard for information on mergers and acquisitions which includes joint ventures and alliances. The database and has been widely used in academic research. The SDC database comprehensively covers all worldwide mergers and acquisitions, joint ventures, and alliances from 1985.

A4. Data Collection We use multiple sources to ensure that we are exhaustive about collecting announcements and also

remove ambiguous announcements. Our unit of analysis is a specific research project. McWilliams and Siegel (1997) focus on three important assumptions about running event studies – 1) Market efficiency 2) Unanticipated events 3) Confounding events. They argue that the assumption of market efficiency is difficult to reconcile with a long event window. The use of very long windows implies that researchers do not believe that the effects of events are quickly incorporated into stock prices. We believe that the market incorporates the information within a short period and hence use the -1 to +1 event window to calculate the returns. For fulfilling the second assumption, we sort the results of our search based on the oldest report on a particular make, buy, or ally. This helps us identify the first release of information to the market and excludes redundant

45

announcements. As for fulfilling the third assumption, we control for seven different types of confounding events around the -1 to +1 window. The following events could positively or negatively affect returns - declaration of dividends, signing of a contract (government and private), announcement of a new product, filing of a large damage suit, announcement of earnings, end of lock-up period, and change in a key executive.

For each make, buy, or ally if any of the seven different confounding events are announced on the day before the event day, on the event day, or on the day after the event day, we exclude the make, buy, or ally from our payoff analyses. We thus isolate the make, buy, and ally from any noise. We know of no other study in marketing which has carried out such an extensive check of seven different confounding events. To determine if an announcement is a make, we carefully read the text of the announcement. If the announcement clearly states that a firm is carrying out any of the four different types of make explained in the definitions section, we code the announcement as a make. For buy, we follow a coding process similar to Ahuja and Katila (2001). We examine the text of the acquisition announcement to establish if the buying firm reported an innovative technology, process, product, service, procedure, or system as a motivating factor for the acquisition or if an innovative technology, process, product, service, procedure, or system was a part of the transferred assets. We classify the acquisition as a buy if either of these conditions was met. For ally, we follow the coding process same as buy above. We drop alliances such as supply, manufacturing, and marketing alliances, which don’t have any innovation component in them. Our sample of alliances is either joint ventures or strategic alliances. Some of the alliances are licensing agreements. We select only those licensing agreements where the firm in our sample is the licensee rather than the licensor. The scheme for coding the make, buy, and ally is available in Online Appendices A5, A6, and A7 below.

A5.1. Terms Used in Factiva and Lexis-Nexis to Search for Make Announcements New product, new, innovation, develop, developing, novel, start, research, development, laboratory, R&D, laboratory, lab, new process, opening, research and development, new division, new service, establishing, new system, new project, technical center, innovation center, new method

A5.2. Description of Types of Make

Type Description

1 Starting a research and development center or research laboratory or innovation center or technical center or operational center

2 Starting a new process or system - A process innovation is one that involves the use of a new approach to creating products or services. Example: Henry Ford’s use of the assembly line in the manufacture of automobiles (Chandy & Prabhu 2011)

3 Starting a new entity (new subsidiary, business unit, division) for developing a new product or service

4 Starting a new project to develop a new product - A product innovation involves the development of a product that is new to customers such as the Kodak Brownie in 1900. It is an example of a product innovation. It offered cheap and easy use of photography to adults and children alike Starting a new project to develop a new service -A service innovation involves the development of a service that is new to customers such as FedEx and UPS. They are examples of service innovations. They enabled customers to deliver documents and packages overnight (Chandy & Prabhu 2011).

A5.3. Examples of Types of Make from Data

Make Type 1: Siemens AG Opens Siemens Corporate Technology China Research Center in Beijing: Siemens AG announced the opening of its Siemens Corporate Technology China research center in Beijing. By the year 2010, Siemens plans to invest around EUR 80 million (USD 102 million) in locations in the Chinese municipalities of Beijing and Shanghai. It also plans to conduct research on new environmental, energy, healthcare, and automation technologies. At present, the enterprise has 200 scientists, and the number will grow to 300 by the year 2008.

46

Make Type 2: IBM to Build $86 Million Energy Efficient Data Center in Boulder as Part of Project Big Green: IBM (NYSE: IBM) announced today an $86 million data center expansion that will add an additional 80,000 square feet of data center space to its Boulder facility. IBM will use state-of-the-art technology to build a "green data center" to help IBM and its clients reduce energy costs…..IBM plans to install high density computing systems utilizing virtualization technology, along with its Cool Blue portfolio of energy efficient power and cooling technologies. These technologies, in conjunction with the energy efficient design and construction, will allow IBM to reduce its overall carbon footprint compared to standard data centers, and lessen the impact to the environment. Make Type 3: Polo Ralph Lauren Corporation announced that the company is starting a new group named Global Brand Concepts (GBC), which will develop new lifestyle brands for specialty and department stores. Through experience, customer insight and knowledge, the Company can provide the design supervision and the expertise needed to navigate and succeed in this complicated market. The group will be available to design, market and advertise full lifestyle product including men's, women's, children's, home and accessories. Make Type 4: News Corp. will introduce a mobile entertainment store as it seeks to tap the growing global cellular phone market. The store, Mobizzo, and a production studio will develop content for mobile phones in a way that 20th Century Fox creates movies and television programming. Unlike predecessors and peers, Mobizzo will sell its offerings directly to consumers and won't partner with a cellular service provider.

A6.1 Terms Used in Factiva and Lexis-Nexis Search to Find Buy Announcements

Acquire or buy or merger or M&A or acquisition or purchase or bought or target or bidder or bid

A6.2. Examples of Types of Buy from Data

Buy Target Patents: 3M announced today that it has entered into a definitive agreement to acquire HighJump Software Inc., a provider of supply chain execution software and solutions, including related trademarks and patents. Buy Target Research Personnel: Motorola, Inc. acquired the Research and Development Team and a research center in western France from Melco Mobile Communication Europe, a subsidiary of Mitsubishi Electric Corporation. Buy Target Software/Technology: Illinois Tool Works Inc., announces the purchase of Quasar International, Inc., the inventor of PCRT (Process Compensated Resonant Testing), a revolutionary new NDT (Nondestructive Test) method. Buy Target Product: Allergan has acquired EndoArt SA, a Swiss provider of remote-controlled implants used in the treatment of morbid obesity and other conditions.

A7.1. Terms Used in Factiva and Lexis-Nexis to Search for Ally Announcements

Strategic alliance, joint venture, co-development, co-develop, co-developing, licensing, cross-licensing, alliance, research alliance, licensing of technology, research and development alliance, R&D alliance, consortium, collaboration+ university, collaboration+ research+institute, co-creation, technology transfer, joint exploration, cross technology transfer, co-invent

A7.2. Description of Types of Ally

Strategic Alliance: if the alliance is a cooperative business activity, formed by two or more separate organizations for strategic purpose(s), which does not create an independent business entity, but allocates ownership, operational responsibilities, and financial risks and rewards to each member, while preserving each member's separate identity/autonomy (SDC database 2013). Joint Venture: if the alliance is a cooperative business activity, formed by two or more separate organizations for strategic purpose(s), which creates an independent business entity, and allocates ownership, operational responsibilities, and financial risks and rewards to each member, while preserving each member's separate identity/autonomy. The new entity can either be newly formed or the combination of pre-existing units

47

and/or divisions of the members. Even if the members' stake in the new entity varies, the members are all considered owners/parents of the new entity (SDC database 2013). Licensing Agreement: Agreement under which the owner (licensor) of an innovation (know how, patent, or other intellectual property), allows a licensee to use the owner’s innovation. We select only those licensing agreements where the firm in our sample is the licensee rather than the licensor.

A7.3. Examples of Types of Ally from Data

Joint Venture: Intel Corp and Legend Group formed a joint venture named Intel-Lenovo Technology Advancement Center (ILT) to own and operate technology advancement center in Beijing, China. ILT was to focus on trusted computing and next generation Internet technologies. Strategic Alliance: Apple Computer Inc and Softbank Corp planned to form a strategic alliance to provide research and development services of mobile phones that have iPod music players function in Japan. Licensing Agreement: Research In Motion and Kodiak Networks formed a strategic alliance wherein Kodiak Networks licensed Research in Motion to utilize its series of integrated mobile applications including Push-to-Talk (PTT) on BlackBerry handsets. The alliance allowed Reliance in Motion to integrate a Java version of Kodiak Networks' handset client with BlackBerry handsets.

A8.1. Relatedness Schema

Scale based on the existing capabilities of firms. The scale is graded on a scale from 1-10 and anchored by 1 for High unrelatedness and 10 for High relatedness. Example: 1 - Highly unrelated: Toyota Motors acquires Kellogg's cornflakes (Toyota has no capabilities in cereals) 4 - Weakly related: Toyota Motors has acquired Sony (Toyota has weakly related capability due to its manufacturing processes) 7 - Moderately related: Toyota Motors is set to enter the motorbike market and thus pose competition to Yamaha (Toyota has moderately related capabilities in transportation vehicles for consumers) 10 - Highly related: Toyota Motors acquires Nissan a firm that makes similar cars (Toyota has the requisite capabilities)

A8.2. Customer Benefit Schema

Scale - The innovation is graded on a scale from 1-10 and anchored by 1 for minor benefit to customers and 10 for major benefit to customers. Minor benefit: Innovation that offers minor improvements in benefits to consumers relative to existing products, services, or processes. Major benefit: Innovation that offers substantially higher customer or user benefits relative to existing products, services, or processes. Example: 1 - Toyota offering a different color for one of its car models 5 - Toyota redesigning one of its car models for extra storage space in the trunk 10 - Toyota offering hybrid cars

Online Appendix B - Estimation Procedure for the Carhart Four Factor Model

The Fama-French (Fama and French 1993) 3 Factor Model posits that not only the market but also a firm’s market capitalization and value affect stock returns. Carhart (1997) added a fourth factor, price-momentum, to explain the persistence effect in returns reported by Jegadeesh and Titman (1993). So, we estimate expected returns on any given day using the Four Factor model (Fama and French 1993; Carhart 1997) that controls for all these four known effects, as follows:

48

2ˆ

2

1

2

2

P

AARAARD

Dt

t

AAR

2

1

21

tt

tt

ittt ARCAR ,

(B.1)

Here i stands for firm, t stands for time, denotes the returns for firm i on day t, stands for

returns on corresponding daily equally-weighted market index on day t, denotes the returns on a portfolio of small stocks minus returns on large stocks, HML stands for returns on a portfolio of stocks with high book-to-market ratio minus the returns on a portfolio of stocks with low book-to-market ratio, UMD

stands for Carhart’s Price-Momentum Factor that captures one-year momentum in returns and stands for

the disturbance term. We estimate the Carhart four factor model over a period -270 to -6 trading days prior to event day. Thus we use 265 days of data for estimating the parameters. We remove announcements, which have an estimation period smaller than 30 days. The abnormal return to an event is calculated as the difference between the normal return, which would have occurred on that day given no event and the actual return that did occur because of the event, thus:

(B.2)

where and are the estimated parameters. We calculate the cumulative average abnormal

returns (CAR) in an event window =-1 between =+1 as

(B.3)

Thus, “returns” represents the cumulative average abnormal returns as defined by Equation B.3.

We use the time series standard deviation test for calculating the t-value (Brown and Warner 1980). This test avoids the potential problem of cross-sectional correlation of security returns. The estimated variance of the average abnormal return AARt for the portfolio of N make or buy announcements is

(B.4) where the model parameters are estimated over the estimation period of P=D2 - D1+1 days (D1 =0 and D2 =265)

(B.5)

(B.6)

Assuming time-series independence, the t-statistic for is

(B.7)

itR mtR

SMB

1

it

iiii 321ˆ,ˆ,ˆ,ˆ i4̂

1t 2t

tAAR

AAR

tAAR

P

AAR

AAR

D

Dt

t

2

1

N

AR

AAR

N

i

it

t

1

1

4321 ittititimtiiit UMDHMLSMBRR

tititimtiiitit UMDHMLSMBRRAR 4321ˆˆˆˆˆ

49

Online Appendix C - Details of Fixed and Random Effects Estimator

An Ordinary Least Squares (OLS) regression analysis of payoffs without controlling for the firm’s strategic choice (make, buy, ally) would lead to biased estimates. The reason is that the firm’s decision of whether to make, buy, or ally is self-selected, depending on the market situation, its abilities, or its expectation of payoffs. For instance, skilled acquisitive firms are more likely to buy because they obtained higher payoffs to buy in the past. Failure to control for this selection bias will yield an estimated buy effect on payoffs that is biased upwards. Generally, to correct for this selection bias, one uses the Heckman (1979) two-stage model (Shaver 1998). Here, one computes the Inverse Mills Ratio from a first stage binary probit model and incorporates the Inverse Mills ratio from the first stage as an explanatory variable in the OLS regression.

Our situation calls for a different strategy to control for selection bias because of two issues. First, instead of a Binary choice model, we have a Multinomial choice model. Second, instead of cross-section data, we have panel data. There exist computationally straightforward approaches to control for selection in a Multinomial choice model for cross-section data. Lee (1982) shows how one can calculate the inverse mills ratio from a Multinomial Logit Choice model in the first stage and incorporate the inverse mills in the second stage regression. However, because we have panel data, we need a different procedure to control for selection (Vella 1998). We follow the procedure suggested by Verbeek and Nijman (1992). If we don’t find evidence of selection, we estimate the model without controlling for selection bias. If we find evidence for such bias, we resort to procedure suggested by Wooldridge (1995).

We describe the approach to test for selection bias below. Consider the case for buy,

(C.1) itti

'

it

*

it eξμβxy +++= (Primary equation)

(C.2) itti

'

it

*

it vψαγzd +++= (Selection equation to buy)

(C.3) 0dif1d *

itit >=

(C.4) it

*

itit d*yy =

where i, (i=1,…….., N) denotes firms and t, (t=1, 2,….T) denotes time. The dependent variable, i.e., payoff to buy, in equation C.1 (primary equation) is only observed for the observations

satisfying the selection rule (namely 0d*

it > i.e. a firm choses to buy only if it crosses a threshold).

Note that the exact formulation of the selection equation does not matter. To introduce selection

bias, we assume the errors for each equation can be decomposed into a firm effect ( iμ and iα ), a

time effect ( tξ and tψ ) and an idiosyncratic effect ( ite and itv ). Each of these error components is

assumed to be normally distributed and correlated with the component of the same dimension in the other equation. As the treatment of time effects increases the difficulty of estimation, it is simpler to

treat them as fixed time effects and absorbed in ix and iz . We can estimate equation C.1 with either

a fixed effect or random effect estimator.

Fixed Effect Estimator Consider the data for the fixed effect estimator. Changing the data for the explanatory

variables into deviations from individual means via the transformation:

50

(C.5)

T

s

isisit

d

it dxxx1

/

T

s

isd1

; if

T

s

isd1

>0

and doing the same for y, the fixed effects estimators for the unbalanced panel is:

(C.6)

it

d

it

N

i

T

t

d

itit

d

it

N

i

T

t

d

itFE dyxdxxInvU1 11 1

''

)(

Random Effects Estimator

Consider the random effects estimator. Define the following vectors; ).......,( 21 itiii yyyy ,

),.....,( 21 itiii xxxX , ),.....,,( 21 itiii eeee . We define the number of units for which 1itd as Ti

and define a Ti X T matrix Ri transforming iy into the Ti dimensional identity matrix corresponding

to 0itd . After defining the T dimensional unit vector , the variance of the error term in

equation D.1 can be written as 2

' + Ie

2 . The random effect estimator is:

(C.7) )X)Ω(invX(Inv)U(β 0

ii

0

iRE∑=

′ )y)Ω(invX( 0

ii

0

i∑ ′

where '

iii RR and iii XRX 0

Consistency (unbiased estimates) of the payoff to buy regression using a Fixed effects

estimator requires 0]d,x|e[E itit

d

it = and consistency of the payoff to buy regression using a

Random effects estimator requires 0]d,x|eμ[E itititi =+ . Thus, consistency of the fixed estimator

requires .0ev That is the sample selection bias must operate purely through the firm specific

terms. Whereas the consistency of the random effects estimator requires the stronger condition that

0ev and 0 . That is, the random effects estimator cannot produce consistent estimates if

the selection is operating through firm and/or idiosyncratic effects. Verbeek and Nijman (1992) propose a strategy to test selection by using the Hausman (1978) testing framework. Under the null

hypothesis of no sample selectivity, the probability limits of FE (estimates of β in eq. C.1 using a

fixed effect estimator) and REβ (estimates of β in equation C.1 using a random effects estimator) are

identical as both procedures are consistent. Under the alternative, however, the two estimators diverge as both estimators are inconsistent, under different forms of selection bias. Note there is no reason to suspect that the inconsistency is the same.

Thus, we test for selection bias by running a Fixed Effects model and then running a Random Effects model for the buy choices and use the Hausman test to check if coefficients differ. Similar to buy, we run the test for the make choices and ally choices. We run the following model:

(C.8) 8

itexpitiit εβ*controlαpayoff ++= .

We include the relatedness of the innovation and customer benefit of the innovation as control variables. The other control variables include three industry dummies, marketing investment, R&D investment, financing capability, firm’s diversification strategy, systematic risk of the strategy, competition, location of the strategy (emerging market), and sub-category of strategy (e.g., ally: joint

51

venture, strategic alliance, licensing). We run equation C.8 for each strategy. The Hausman test results in the table below indicate that there is no sample selection bias as there is no difference in the estimates of the Fixed Effects and Random Effects models. Note that this test absolves us from not using the firms that don’t innovate in our analysis. We have 28 firms out of the 192 firms that do not innovate.

Hausman Test Results for Checking Sample Selection Bias

Model N Chi-Square Prob>Chi-Square

Ally Model 394 16.62 0.2769

Make Model 441 10.46 0.728

Buy Model 754 7.78 0.9817

Online Appendix D - Descriptive Statistics of Models for Choice and Payoff

Model for Choice (N=3260) Model for Payoff (N=1589) Variable Mean S.D. Mean S.D.

Prior Make Payoff 0.00 0.01

Prior Buy Payoff 0.00 0.02

Prior Ally Payoff 0.00 0.02

Number of Patents 4.35 6.11

Number of Commercializations 0.03 0.03

Managerial Capability 2.41 2.04

Financing Capability 0.09 0.07 0.08 0.07

Prior Number of Make 0.22 0.34

Prior Number of Buy 0.18 0.17

Prior Number of Ally 0.22 0.32

Marketing Investments -0.24 0.46 -0.24 0.51

R&D Investments -0.06 0.72 -0.02 0.8

Related-Diversified Firms 0.21 0.40 0.23 0.42

Unrelated-Diversified firms 0.29 0.45 0.28 0.45

High Diversified Firms 0.28 0.45 0.22 0.41

B2B goods 0.39 0.49 0.43 0.49

B2C goods 0.33 0.47 0.3 0.46

B2C services 0.25 0.43 0.21 0.41

Return (Dependent Variable in Payoff Model)

0.00 0.02

Ally Indicator 0.25 0.43

Make Indicator 0.28 0.45

Innovation Relatedness 7.15 1.95

Customer Benefit 6.18 1.7

Online Appendix E - Results of Multinomial Model of Choice – Estimates

Relative to Make Choice

Buy Choice Ally Choice

Independent Variables Coefficient Coefficient

52

Prior Make Payoff -15.48** -8.67*

Prior Buy Payoff -2.48 -5.46

Prior Ally Payoff -3.48 -3.72

Number of Patents -0.01 0.04

Number of Commercializations -17.99** 2.62

Managerial Capability 0.03 0.01

Financing Capability 1.94 -0.19

Prior Number of Make -1.00** -1.40

Prior Number of Buy 0.42 -0.28

Prior Number of Ally 0.76 0.59

Marketing Investment -0.16 -0.24

R&D Investment -0.02 -0.19

Related-Diversified Firms 0.04 -0.03

Unrelated-Diversified firms -0.41 -0.52*

High Diversified Firms 0.17 -0.52*

B2B goods -1.14* -0.69

B2C goods -2.04** -0.68

B2C services -2.10** -0.69

Intercept 1.96** 0.79

Variance of random effect of ally 0.24**

Variance of random effect of buy 0.13

Covariance of random effects of ally and buy 0.38 The symbols * and ** denote statistical significance at the 0.05 and 0.01 levels; Log Likelihood Value = - 3090.75; Estimates are relative to Make choice coded as 1;

Online Appendix F - Results of Model of Choice: Marginal Effects with Longer

Windows

Using 2 year window for prior payoff variables

Using 3 year window for prior payoff variables

Make Buy Ally Make Buy Ally

Independent Variables Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient

Prior Make Payoff 2.23* -2.74** -0.37 3.18* -3.47** -0.04

Prior Buy Payoff 0.26 -0.24 0.05 0.34 -0.21 -0.09

Prior Ally Payoff 0.15 -0.32 0.05 0.30 -0.59 0.09

Number of Patents 0.00 -0.01 0.01 0.00 -0.01 0.01

Number of Commercializations 3.07* -4.81** 1.70 3.11* -4.84** 1.69

Managerial Capability 0.00 0.01 0.00 0.00 0.01 0.00

Financing Capability -0.31 0.55* -0.23 -0.28 0.52* -0.25

Marketing Investment 0.04 -0.02 -0.02 0.03 -0.02 -0.02

R&D Investment 0.01 0.01 -0.02 0.01 0.01 -0.02

Related-Diversified Firms -0.01 0.01 -0.01 -0.01 0.01 -0.01

53

Unrelated-Diversified firms 0.09 -0.07 -0.03 0.10 -0.19 -0.03

High Diversified Firms 0.00 0.08 -0.07* 0.00 0.09 -0.07*

B2B goods 0.70 -0.05 0.01 0.71 -0.05 0.01

B2C goods 0.28** -0.31** 0.13 0.29** -0.31** 0.14

B2C services 0.18* -0.31** 0.16* 0.18* -0.31** 0.16*

Prior Number of Make 0.24* -0.14 -0.11 0.24* -0.14 -0.10

Prior Number of Buy -0.05 0.12 -0.06 -0.05 0.12 -0.07

Prior Number of Ally -0.17 0.14 0.01 -0.20 0.13 0.01

Intercept -0.37* 0.43** -0.05 -0.37* 0.43** -0.05 * and ** denote statistical significance at the 0.05 and 0.01 levels;

Online Appendix G: Results of Model of Choice - Marginal Effects using

Weighted CRSP

* and ** denote statistical significance at the 0.05 and 0.01 levels;

Make Buy Ally

Independent Variables Coefficient Coefficient Coefficient

Prior Make Payoff 2.85* -3.08** -0.02

Prior Buy Payoff 0.55 -0.09 -0.49

Prior Ally Payoff 0.39 -0.51 0.01

Number of Patents 0.00 -0.01 0.01

Number of Commercializations 3.11* -4.83** 1.69

Managerial Capability 0.00 0.01 0.00

Financing Capability -0.31 0.50* -0.19

Prior Number of Make 0.24* -0.14 -0.10

Prior Number of Buy -0.05 0.13 -0.07

Prior Number of Ally -0.17 0.14 0.01

Marketing Investment 0.04 -0.02 -0.02

R&D Investment 0.01 0.01 -0.02

Related-Diversified Firms -0.01 0.01 -0.01

Unrelated-Diversified firms 0.09 -0.06 -0.03

High Diversified Firms 0.00 0.07 -0.07*

B2B goods 0.71 -0.05 0.01

B2C goods 0.28** -0.30** 0.13

B2C services 0.18* -0.31** 0.18

Intercept -0.35* 0.39** 0.04

54

Online Appendix H - Stepwise Build-up of Regressions

Table H1: Results of Multinomial Model of Choice – Estimates Relative to Make Choice

Model 1 Model 2 Model 3 Model 4 Model 5 Model 5

55

Estimates are relative to Make choice coded as 1. The symbols * and ** denote statistical significance at the 0.05 and 0.01 levels;

Buy Choice

Ally Choice

Buy Choice

Ally Choice

Buy Choice

Ally Choice

Buy Choice

Ally Choice

Buy Choice

Ally Choice

Buy Choice

Ally Choice

Ind. Variables Estimate Estimate Estimate Estimate Estimate Estimate Estimate Estimate Estimate Estimate Estimate Estimate

Prior Make Payoff -19.07** -4.61 -18.34** -4.47 -15.5** -8.67 -15.5** -8.67 -15.5** -8.67 -15.48** -8.67*

Prior Buy Payoff -0.91 -2.8 -1.08 -2.8 -2.46 -5.46 -2.47 -5.47 -2.46 -5.46 -2.48 -5.46

Prior Ally Payoff -3.52 -1.7 -3.74 -1.48 -3.48 -3.72 -3.47 -3.71 -3.47 -3.71 -3.48 -3.72

Number of Patents -0.01 2.37 -0.00 1.78** -0.05 0. 73

-0.05 1.25 -0.01 0.04

Number of Commercializations

-17.39** 4.06 -18.01** 2.64 -18.0** 2.64 -18.0** 2.63 -17.99** 2.62

Prior Number of Make

-1.07** -1.22 -1.24** -1.28 -1.22** -1.25 -1.00** -1.40

Prior Number of Buy

0.67 -0.13 0.7 -0.19 0.06 -0.18 0.42 -0.28

Prior Number of Ally

0.48 0.77 0.48 0.71 0.44 0.73 0.76 0.59

Managerial Capability

-0.01 1.32 -0.02 1.6** 0.03 0.01

Financing Capability 1.84 -0.14 1.86 -0.13 1.94 -0.19

Marketing Investment

0.15 -0.4 -0.16 -0.24

R&D Investment -0.06 -0.33 -0.02 -0.19

Related-Diversified Firms

0.04 -0.03

Unrelated-Diversified firms

-0.41 -0.52*

High Diversified Firms

0.17 -0.52*

B2B goods -1.14* -0.69

B2C goods -2.04** -0.68

B2C services -2.10** -0.69

Intercept 1.96** 0.79

56

Table H2: Random Effects Regression Estimates (Dependent Variable: Returns) Independent Variables Coef. Coef. Coef. Coef. Coef. Coef.

Buy Indicator -0.54%** -0.54%** -0.54%** -0.54%** -0.57%** -2.83%**

Ally Indicator 0.12% 0.11% 0.11% 0.12% 0.29% 0.16%

Related-Diversified 0.02% 0.02% 0.00% -0.05% 0.03%

Unrelated-Diversified -0.01% 0.00% 0.01% -0.04% 0.00%

High Diversified -0.25% -0.24% -0.28% -0.27% -0.20%

Financing Capability 0.23% 0.18% 0.25% 0.24%

Marketing Investment 0.15% 0.12% 0.12%

R&D Investment 0.02% 0.04% 0.05%

Innovation Relatedness 0.06% -0.03%

Customer Benefit 0.06% -0.02%

Prior Risk to Buy 0.15% 0.11%

Prior Risk to Ally -0.32% -0.30%

Prior Risk to Make 0.2% 0.21%

Buy – Emerging 0.1% 0.17%

Ally – Emerging -0.45% -0.31%

Make - Emerging -0.08% -0.09%

B2B goods -0.46% -0.46%

B2C goods -0.60% -0.58%

B2C services -0.68% -0.74%*

Prior Number of Buy -1.18%

Buy_Target_Product -0.50%

Buy_Target_Software -0.04%

Buy_Target_Research_Personnel -0.58%*

Buy_Target_Patent -0.77%

Ally_Joint_Venture -0.61%

Ally_Licensing -0.55%

Make_R&D_center 0.02%

Make_New_Project -0.05%

Make_New_Entity -0.35%

Buy_Target_Product -0.50%

Competition -0.06%

Buy Indicator* Relatedness 0.16%**

Buy Indicator*Customer Benefit 0.21%*

Buy Indicator*Prior Number of

Buy

1.71%*

Intercept 0.2%* 0.24% 0.22% 0.27% 0.00% 1.42%

The symbols * and ** denote statistical significance at the 0.05 and 0.01 levels;

Online Appendix I – Propensity Score Matching

The details of the Propensity Score estimation are in URL: http://bit.ly/1cIrdPE

Online Appendix J – References in Table 1

Chan, L. K., Lakonishok, J., Sougiannis, T. 2001. The stock market valuation of research

and development expenditures. The Journal of Finance, 56(6), 2431-2456.

Eberhart, A. C., Maxwell, W. F., Siddique, A. R. 2004. An Examination of Long‐Term

Abnormal Stock Returns and Operating Performance Following R&D Increases. The Journal of

Finance, 59(2), 623-650.

Park, N. K., Mezias, J. M., & Song, J. 2004. A resource-based view of strategic alliances

and firm value in the electronic marketplace. Journal of Management, 30(1), 7-27.

57

Higgins, M. J., & Rodriguez, D. (2006). The outsourcing of R&D through acquisitions in

the pharmaceutical industry. Journal of Financial Economics, 80(2), 351-383.

Kalaignanam, K., Shankar, V., Varadarajan, R. 2007. Asymmetric new product

development alliances: win-win or win-lose partnerships?. Management Science, 53(3), 357-374.

Benson, D., Ziedonis, R. H. 2009. Corporate venture capital as a window on new

technologies: implications for the performance of corporate investors when acquiring startups.

Organization Science, 20(2), 329-351.

Oxley, J. E., Sampson, R. C., Silverman, B. S. 2009. Arms race or détente? How

interfirm alliance announcements change the stock market valuation of rivals. Management

Science, 55(8), 1321-1337.

Sood, A., Tellis, G. J. 2009. Do innovations really pay off? Total stock market returns to

innovation. Marketing Science, 28(3), 442-456.

Ransbotham, S., Mitra, S. 2010. Target age and the acquisition of innovation in high-

technology industries. Management Science, 56(11), 2076-2093.

Zaheer, A., Hernandez, E., Banerjee, S. 2010. Prior alliances with targets and acquisition

performance in knowledge-intensive industries. Organization Science, 21(5), 1072-1091.

Sears, J., & Hoetker, G. 2013. Technological overlap, technological capabilities, and

resource recombination in technological acquisitions. Forthcoming, Strategic Management

Journal.