WHY AND HOW DO FIRMS DIVEST? DISSERTATION Presented in Partial Fulfillment of the Requirements for
the Degree Doctor of Philosophy in the Graduate
School of The Ohio State University
By Naga Lakshmi Damaraju B.Sc. M.B.A. P.G.D.E.M. L.L.B. M.B.A. M.S. * * * * * The Ohio State University 2008 Dissertation Committee: Professor Jay B. Barney, Adviser Approved by Professor Michael Leiblein Professor Anil Makhija ______ Professor Sharon James Adviser Business Administration Graduate Program
Copyright by Naga Lakshmi Damaraju 2008
ABSTRACT This dissertation extends and examines the implications of real options theory—a theory that developed primarily in the context of investment decisions—in the context of divestment of business units. In doing so, it challenges some of the existing ideas about divestments—i.e., that firms’ unwillingness to divest may be due to agency reasons or that divestments are a reactive response to poor performance at parent and/or business unit levels. Further, alternative explanations for divestments and divestment mode choices are considered along with explanations from the real options theory and the relative explanatory power of rival theories is established in this series of three closely- related essays. All hypotheses are tested on a sample of divesting (298) and non- divesting firms (596) that have been obtained from various sources following previous studies.
The first essay shifts the focus of real options theory from investment decisions to divestment decisions. The key implications of the theory are developed and tested for business unit divestments as ‘put’ options on real assets. The results show that uncertainty in a business unit’s environment is negatively related with the decision to divest. Further, this negative relationship remains when environmental uncertainty increases and disappears when environmental uncertainty falls. Together, these results
ii
suggest that firms’ decisions to divest may be driven by option considerations when the environment of the business unit is uncertain.
While the first essay is about the decision to divest, the second essay follows closely and is about divestment mode choices. In the second essay, the relative explanatory power of real options theory and information asymmetry is explored in the context of divestment mode choices. In particular, it is tested whether staged divestments are more likely as compared to sell-offs due to 1) uncertainty in a business unit’s environment and/or 2) information asymmetry about the true value of the business unit.
The results show that information asymmetry explains staged divestments whereas uncertainty is irrelevant to divestment mode choice decisions. This shows that while explanations exist for staged investments from a real options perspective, staged divestments do not seem to follow a similar explanation. Therefore, a separate ‘real options theory of divestments’ may be needed, rather than simply expecting the predictions from ‘real options theory of investments’ to apply in the context of divestments.
The third essay is also related to the first two essays and focuses on the relative impacts of dynamic changes in uncertainty and information asymmetry on the decision to divest, per se. While prior studies have examined explanations from real options theory and information asymmetry separately, there could be an interesting linkage between these two theories in the context of divestments. While dynamic changes in uncertainty have been dealt with in the first essay, in the third essay the dynamic changes in information asymmetry have also been considered to delineate the conditions where
iii
options thinking may have a dominating effect as compared to information asymmetry concerns. Results support the theoretical prediction that options considerations, under certain conditions, may dominate information asymmetry concerns.
Together, these three essays represent an important extension of real options theory and improve our understanding of the divestment phenomenon.
iv
Dedicated to the Union Public Service Commission of India for rejecting me in 2001, the failure that made me change direction and embark on this yet another journey of excellence. Further, this dissertation is also dedicated to the taxpayers in India and the United States whose moneys have been utilized for my education. v
ACKNOWLEDGMENTS
I thank my mother—my first teacher and angel investor, and my father—an uncompromising perfectionist disciplinarian who wants me to win against all odds at all times, who have laid foundation for my personality and aspirations. I learned compassion and commitment from my mom, a very delicate person at heart. I learned language, reasoning, debating, thinking, determination, outspokenness and selfless service from my dad who is very true at heart, incredibly articulate and powerful in expression. I also thank my brothers Ramu and Anil, sister-in-law Latha, and my little nephew Ganesh Srinivas, for taking care of things back home when I was away, without which this expedition to the United States would not have been a success. These people are the locus of my existence.
This dissertation also is the culmination of years of efforts, direct and indirect, of several teachers who educated me. The first, and as important as my mom and dad, is my primary school teacher, Lalitha. She gave me a lot of personal attention and sincerely believed that I was destined to achieve something great. She had deep impact on my thinking at that early age. My privilege to hoist the national flag in the school every year on the Independence and Republic days sowed seeds of commitment to a career in the service of my nation, an ambition that I pursued over the next 25 years of my life. I thank her for her dedication to the profession of shaping young minds and wish my country the good fortune of having such wonderful individuals in the years to come.
vi
I also thank my teachers at Pragathi High School, Hyderabad, India—Sarada,
Rama, Satya, Shobha, Vidya, Mandakini, Saroja, Rajya Lakshmi, Rajeswari, Surya
Prabha, to name a few— who have provided a very nurturing yet a very challenging learning environment during my adolescent years. At this school, teaching was not a business. It was a passion. A sense of commitment to students’ holistic development pervaded its environs. This shaped my philosophy of training, which eventually became my profession. I specially thank my mathematics teacher and a very good friend, Gopal, because of who I later majored in Mathematics, a subject that I would otherwise have not dared to venture into. While mathematics never has been, nor will be, a subject of deep interest to me, studying it taught me a core principle of learning—i.e., nothing is so difficult to learn, if one really wants to learn it.
My teachers—Rajya Lakshmi, Shyamala, Nirmala and Lakshmi—from undergraduate college, Vanitha Maha Vidyalaya, have provided very valuable guidance when I was at cross-roads in terms of further education and career choices. They supported me when I wanted to embark on unfamiliar roads and advised that I should pursue what I truly desired. Rajya Lakshmi and Shyamala remained my personal advisors ever since. I sincerely acknowledge their contribution. Dr. Lakshmikanth
Reddy, my marketing teacher at Badruka College, where I got my first MBA in 1994, is another person who deserves an honorable mention.
Any number of thanks will be insufficient to acknowledge another great gentleman, Dr. S. Pratap Reddy, Principal-Director, Dhruva College of Management.
Dhruva is a unique institution that provides a nurturing environment not only to its
vii
students, but also to its faculty members. I learned the power of listening to an opposite point-of-view with patience and disagreeing with respect from Dr. Reddy (our Tiger). He and his wife, Pushpa Latha Reddy, supported all my endeavors—whether it was pursuing my dream of joining the Indian Administrative Services or pursuing education abroad later on. They were truly a phone call away to act on my requests to handle things back home when I was in the Untied States. I am not sure if I can ever repay their kindness. I had the good fortune of serving this school for a decade as a faculty member and am now extremely honored with an invitation to be on its governing council.
It was at Dhruva that I met the person who changed the trajectory of my life—Dr.
Rao H. Unnava, Professor of Marketing at the Fisher College of Business, The Ohio State
University. Dr. Rao is an incredible gentleman, a very fine teacher, and a true mentor. I thank him for his immense faith in my abilities when he advised me to join a PhD program in the United States way back in 1999. Dr. Rao’s entry into my life seems so predetermined. I picked up on his suggestion to come over to the U.S. two years later when I finally failed at getting my dream job and being a part of the administrative machinery in my country. I chose to get into the PhD program in Strategy at The Ohio
State almost five years after I first met him. He was never impatient with me about which school I chose to go to or which subject I eventually wanted to pursue my PhD in.
Dr. Rao taught me, by example, a very important thing about advising—i.e., good advisors give good advice and also leave enough latitude in implementation. He opened a new window of thought in my mind…it became the window to the world. He truly worked in my best interest and is a mentor for life.
viii
Two other incredible mentors who paved my way to Dr. Jay Barney at the Ohio
State University are Dr. Gregory Dess and Dr. Robert Kieschnick during my masters’ program at the University of Texas at Dallas. Dr. Dess had a vision for me and has been very very affectionate. He truly brought me back into academia, helped me make the decision to pursue doctoral studies and encouraged me to go to an advisor who could contribute most to my interests. As his TA, I got good training in academic writing and enjoyed great flexibility and independence. He created my first copyrights and royalties.
He has been, and continues to be, a great source of encouragement and support.
Dr. Kieschnick, my finance professor and mentor, was always willing to discuss, gave time and attention, been affectionate and shaped my preliminary ideas that eventually led to choosing the dissertation topic. According to Dr. Kieschnick, an assistant professor’s tenure clock starts right at the time he/she joins the PhD program .
He therefore thought I should be very well-prepared even before I got into it. Both Dr.
Dess and Dr. Kieschnick wanted me to be prepared for a conversation with my advisor the very first day I meet him. That preparation helped me hit-the-ground-running when I eventually joined the doctoral program at Fisher College. I am deeply indebted them. I am also grateful to Jyothi Mallick, Director—Cohort MBA program and Professors,
Joseph Picken, Constantine Konstans, Michael Oliff for their valuable recommendations,
Mary and David Lawson and friends, Charul, Pranav, Arturo, Sherif, Bryan, Daniel (my
Cohort), Vinetta, Margie and Taylor Dess for keeping life going while at UTD.
I also thank several Professors at the Ohio State University for their support throughout the PhD program. Michael Leiblein, Mona Makhija, Jay Anand, Anil
ix
Makhija and Sharon James served on my committees and have been very supportive.
Mona was the first professor I worked with at the Ohio State as a teaching assistant and she has been a source of valuable advice at important times. Anil’s corporate finance seminar helped me understand the nuances of empirical research in finance, an exercise that proved very valuable for this dissertation. He has been very encouraging and supportive throughout the dissertation process. Sharon Alvarez has been very warm, encouraging and supportive. Konstantina Kiousis has been very friendly. Dr.
Greenberger, the department chair, has been very considerate and accommodated several requests. Professors Miyazaki (Economics) and Omer Ozturk (Statistics) have been very empathetic in tough times. Professors Stephen Cosslett (Economics), Doug Schroeder,
David Williams, John Fellingham (Accounting), Mike Fligner (Statistics), Dr. Rao
Unnava (Marketing) also need to be specially acknowledged for their support during my dissertation. Special thanks go to Jay Dial—an extraordinary teacher—who has taken time to educate me on fine-tuning my teaching style. I also thank Roy Lewicki for letting me attend his EMBA classes. Professors Karolyi, Werner, Hirshleifer, and others from
Finance, Oded Shenkar (International Business), Steffanie Wilk and others from MHR have also been very considerate of my questions and requests.
Kathy, Heidi and Joan, our administrative support, have been immensely helpful and are the best people I have seen in these roles. They truly enjoy their jobs and have been very kind and nice. Kathy had to have extraordinary patience in dealing with my several appointments, cancellations and rescheduling throughout this time and she has always been nice, warm, welcoming and helpful. Shirley Gaddis (Marketing), Nancy
x
Ray (Entrepreneurship center), Mr. Popovich (Librarian), Amy Richards and Naveen
Baddam (Advising) have also been very kind on several occasions. I also thank Council for Graduate Education in Business Administration and Hayes Forum for travel funding.
Throughout this journey, I enjoyed incredible support, trust and camaraderie of several wonderful and very capable friends, who were also my very tough competitors.
My high school buddies—Ravi, Padma and Raji, my friend from undergrad studies—
Roopa and her family, my friends from MBA—Glory, Manohar, Chetan and Senthil, my civil service buddies—Raghunath, Swarna and later Aparna (GMAT), have been with me regardless of success or failure and been my sponsors at several important times. Padma and her husband Raj have been of great support, both back in India and in the United
States. Roopa and Glory can be rightfully described as my ‘alter egos’. I truly cherish their friendship and owe them a lot for their faith in my decisions. My very close friend and cousin NVK Rao, his wife Raji and their family are inextricably twined in this journey. Padma and Srinath, who consider me their daughter, are a very important part of the support network. All these people know my aspirations and boosted my courage when it was hitting rock bottom. I cannot thank them enough. My colleagues at Dhruva
College, Venu, Loknath, Aparna, Sailaja, Usha, Chakri, Suresh, Dashrath, Gopal and
Veera have taken care of several issues back home when needed and truly deserve to be mentioned. My students at Badruka and Dhruva have been sources of fresh ideas and enthusiasm always. Ramanathan, NIS-Sparta, my friend and guide from the corporate world, has been very encouraging, helpful and supportive of my choices. Ganesh Sharma,
HDPI-HSBC, has also been a valuable source of advice at important junctures.
xi
Added to these friends are Angie, Roger, Taylor, Rose and Jerome (Finance),
Hammad and Ayesha (Economics), Vikas, Somnath and Sheenu (Statistics), Jenny and
Sandeep (Marketing), Nilesh, Anup, Doug, Alison, Ty and Asli (Strategy), Janice, Chad,
Marie and Aden (MHR) at the Ohio State. They entertained numerous questions and have been patient with me throughout the PhD program. Susan, Charlie, Joe, Erin, Suresh,
Sungho, Song, Yuping and Chris are valuable additions. Sridhar (SAS consultant) has been extremely generous in investing time and effort for training me in SAS and continues to extend his professional expertise to my projects. Hammad has been extremely helpful with econometric modeling and been very willing to entertain my numerous questions. Angie, Anup, Sridhar and Hammad have truly been the econometric and programming backbone for this dissertation.
My friend Mary (OSU alumnus) and her family, Nilesh, Sujith, Jay Dial,
Sandhya, Sridevi (University village) and their families, Gopesh and Sowmya (UIUC), the Bendapudis, the Makhijas, the Unnavas, and my little roomies at 251 West 9 th
Avenue —Ashwini, Shikha, Rimky and Vidya—have all been an important part of my social life in Columbus, OH. Mary and her family helped with numerous moves, in rain and snow, have been very dependable and kept life under control. I cannot imagine how life would have been without Mary in Columbus. She has been my ‘one-stop shop solution’ for all questions and problems and is now a friend for life. Professors Tom
Lumpkin (TexasTech), Jay Janney (University of Dayton), Alan Eisner (Pace), Marilyn
Taylor (UMKC), Saikat Chowdhury (Wharton), Nandini Lahiri (UNC), Seung-Hyun Lee
(UT Dallas) and my colleagues from other universities Manisha Singal (Virginia Tech),
xii
John Upsom (Florida State), Ekin Alakent, Irem Demirikan, Ted Khoury and Kiran
Ismail Mirza (UT Dallas) have given valuable advice and been supportive at very important times.
Language has its limitations in capturing the range of human emotions. I experience this practically when I have to describe my gratitude to my advisor, philosopher and mentor, Dr. Jay Barney. I discovered Jay Barney in my second semester at UT Dallas, through his writings, and am very thankful to him for admitting me as his student. He exceeded my expectation for an advisor every time on every issue. He is not just an individual. He is a philosophy…of life and research. He is an interesting
Valuable, Rare, Inimitable combination of extreme levels of intelligence, great insight, genuine curiosity for exploring new ideas, deep compassion and affection for his students….to say the least. For him, research is an art and I can say he is a great artist and a very skilled trainer. A friend advised me early on that if I do not see a twinkle in
Jay’s eyes when I describe an idea, then it is not worth pursuing him on it. I had the good fortune of seeing that twinkle quite a few times in the last couple of years. It is exhilarating.
If Jay gets interested, it is like being with a play-mate. He plays along and will be unbelievably flexible about the approach one takes to problems. I am truly amazed at his flexibility. He is a great developer of his students’ abilities. Without his flexibility, I could not have overcome my aversion to programming and become a reasonably good programmer. It is also possible that I will venture into mathematical modeling sometime.
xiii
Jay’s focus is extraordinary. He X-rays through a 60-page paper to get to the core problem in a matter of seconds. When he is at work, it is MEDITATION…nothing distracts him. When people get his time, they completely get it. There are no phone calls, no other conversations in between unless it is an overriding priority. It may be quite a wait to get that time, but it is truly worth waiting for. Speaking with him for a few minutes always gave me direction for next 2-6 months. Our conversation usually opened with the incisive question “So, what have you got?” If I could not answer that question in the quick 1-2 minutes, I would realize myself that I have not done my homework well. I quickly rescheduled in such instances and walked back to my office to prepare better.
Talking to Jay for 2 minutes meant a couple of months of preparation. This helped me get precise in expression and keep focus on the most important issues. He is not a
‘hands-on’ supervisor. That is his power. I feel obliged to finish my work knowing very well that he will not ask for it. Time and distance do not matter when one works with
Jay…work gets done. He is an academic Chief Executive….and it was my duty to optimally utilize his time.
On the other hand, when there was an emergency back home, in school, in course work or anything, Jay was most willing to speak and resolve problems. I never had any difficulty getting his time. He is not bureaucratic. At several times when I was on the low of the “PhD roller-coaster” (Jay’s description), he had the patience to hear me out and instilled confidence that I CAN DO IT. I never, at such times, felt he was busy.
When it was time for my proposal defense, dissertation, or job interview, he spent hours with me on several days fine-tuning the presentations, papers and even my vita until it
xiv
was up to our satisfaction. He really cares and is very nurturing. I am extremely privileged to have Jay as my GURU, a great role model, and truly cherish these four years of being his student. Finishing at The Ohio State University only means the beginning of life-long studentship with him and an intrinsically challenging career to pursue ….there is no end to learning from Jay…there is no limit….
Finally, I thank God Almighty who continues to manifest himself in several of these human forms and enriches my life…Tamasoma Jyotirgamaya…in this journey from darkness to light…
xv
VITA Master of Science August, 2004 The University of Texas at Dallas Concentration: Strategy
Master of Business Administration December, 2003 The University of Texas at Dallas Concentration: Strategy Bachelor of Laws November, 1999 Osmania University, India Post-graduate Diploma in Environment Management August, 1995 Hyderabad Central University, India Master of Business Administration April, 1994 Osmania University, India Specializations: Marketing(1994) & Finance (1995 ) Bachelor of Science April, 1992 Osmania University, India Group: Mathematics, Physics & Chemistry Qualified in the University Grants Commission’s National Eligibility Test for Lecturership at University Level in India, December 1995. Fisher College of Business, The Ohio State University from Sep 2004 Graduate Teaching Assistant Dhruva College of Management, Hyderabad, India Oct 1998 to Aug 2002 Associate Professor (on sabbatical from Aug 2002 to Jun 2008) Badruka College PG Center, Hyderabad, India Jul 1994 to Mar 1997 Lecturer xvi
PUBLISHED WORK Damaraju, N., Byrne, J. C., and Eisner, A. B. 2004. Ford Motor Company in 2004: Entering second century of existence. In Dess, Lumpkin and Eisner, Strategic Management, Text & Cases, second edition: 641-649. New York: McGrawHill-Irwin. Damaraju, N., Gaetz, J. R., Taylor, M. L, and Dess, G. G. 2004. JetBlue Airways: Is “High Touch Service” the key driver of JetBlue’s future success?. In Dess, Lumpkin and Eisner, Strategic Management, Text & Cases, second edition: 700-709. New York: McGrawHill-Irwin. (presented at the Society for Case Research Conference in Kansas City, 2004) Damaraju, N., Gaetz, J. R., Taylor, M. L, and Dess, G. G. 2004. Southwest Airlines: How much can ‘LUV’ do?. In Dess, Lumpkin and Eisner, Strategic Management, Text & Cases, second edition: 837-845. New York: McGrawHill-Irwin. Damaraju, N. 1999. The role of business schools in small & medium enterprise development . The Economic Times, Learning Management section , 05/02/1999 . Damaraju, N. 1999. Role of leadership in franchisee management. The Economic Times, Learning Management section, 04/29/1999. Mallikharjuna, C. and Damaraju, N. 1996. Management education--A closer look. In Gautam V. (Ed.), Learning Management: 202-205. New Delhi: Allied Publishers. (Presented at a seminar sponsored by All India Council for Technical Education on 'Planning and Management of Management Education in India' in 1995). FIELDS OF STUDY Major Field: Business Administration Minor Field: Statistics xvii
TABLE OF CONTENTS
Page
Abstract……………………………………………………………………………… ii
Dedication…………………………………………………………………………….v
Acknowledgments...………………………………………………………………….vi
Vita……………………………………………………………………………………xvi
List of Tables………………………………………………………………………….xx
List of Figures…………………………………………………………………………xxii
Chapters:
1. Introduction……………………………………………………………………1
2. Divestment as a real option: Firm choices under conditions of uncertainty…..6
2.1 Theoretical background…………………………………………………..8 2.2 Theory development and hypotheses…………………………………….12 2.2.1 Uncertainty and divestments as ‘put’ options……………………12 2.2.2 Changes in the levels of uncertainty and the decision to divest…14 2.3 Method…………………………………………………………………...16 2.3.1 Data………………………………………………………………17 2.3.2 Other data sources………………………………………………..20 2.3.3 Variables and measurement……………………………………...20 2.4 Analysis and results……………………………………………………...29 2.5 Discussion and Conclusion………………………………………………36 3. Uncertainty, information asymmetry and divestment strategies…………………54 3.1 Prior literature…………………………………………………………....56 3.1.1 Real options……………………………………………………...56 3.1.2 Information asymmetry………………………………………….59
3.2 Theory and hypotheses…………………………………………………..61 3.2.1 Staged divestments: The options perspective…………………....61 3.2.2 Staged divestments: The information asymmetry perspective…..62 3.3 Method…………………………………………………………………...64 3.3.1 Rationale for the choice of methodology………………………...64 3.3.2 Data……………………………………………………………....66 3.3.3 Other data sources………………………………………………..70 3.3.4 Variables and measures for the selection equation……………....70 3.3.5 Variables and measures for the outcome equation………………78 3.4 Analysis and results……………………………………………………...80 3.4.1 Summary statistics……………………………………………….80 3.4.2 Selection equation results………………………………………..82 3.4.3 Outcome equation results………………………………………..84 3.5 Discussion and conclusion………………………………………………86 4. When does information asymmetry matter to the decision to divest?...... 100 4.1 Prior literature………………………………………………………….102 4.2 Theory development and hypotheses…………………………………..105 4.2.1 Information asymmetry and uncertainty: The constructs……...105 4.2.2 The case of increasing environmental uncertainty and the impact of information asymmetry……………………...108 4.2.3 The case of decreasing environmental uncertainty and the impact of information asymmetry……………………...110 4.3 Method………………………………………………………………….111 4.3.1 Data……………………………………………………………..115 4.3.2 Other data sources………………………………………………119 4.3.3 Variables and measurement…………………………………….119 4.4 Analysis and results…………………………………………………….127 4.5 Discussion and conclusion……………………………………………...131 References………………………………………………………………………………146 Appendix………………………………………………………………………………..155
LIST OF TABLES Table Page 2.1 Summary statistics………………………………………………………………41 2.2 Correlation matrix……………………………………………………………….42 2.3 Results from binary probit for the relationship between uncertainty and the decision to divest…………………………………………………………….45 2.4 Marginal effects from binary probit for the relationship between uncertainty and the decision to divest………………………………………………………..46 2.5 Probit results for sub-sample A (firms experiencing increase in uncertainty from ‘t-1’ to ‘t’)………………………………………………………………………..47 2.6 Marginal effects for sub-sample A (firms experiencing increase in uncertainty from‘t-1’ to ‘t’)…………………………………………………………………..48 2.7 Probit results for sub-sample B (firms experiencing decrease in uncertainty from ‘t-1’ to ‘t’)………………………………………………………………………..49 2.8 Marginal effects for sub-sample B (firms experiencing decrease in uncertainty ‘t-1’ to ‘t’)………………………………………………………………………..50 2.9 Results from combined estimation using dummy variables……………………..51 3.1 Summary statistics for divesting and non-divesting firms……………………….90 3.2 Summary statistics for firms engaging staged and complete divestments……….91 3.3 Correlation matrix………………………………………………………………..92 3.4 Main effects for selection and outcome equations from probit with selection…..94 3.5 Marginal effects for the selection equation………………………………………98 3.6 Marginal effects for the outcome equation………………………………………99
4.1 Summary statistics for divesting and non-divesting firms……………………..134 4.2 Correlation matrix………………………………………………………………135 4.3 Main effects for the case of increasing uncertainty in a business unit’s environment…………………………………………………………………….137 4.4 Marginal effects for the case of increasing uncertainty in a business unit’s environment…………………………………………………………………….139 4.5 Main effects for the case of decreasing uncertainty in a business unit’s environment…………………………………………………………………….141 4.6 Marginal effects for the case of decreasing uncertainty in a business unit’s environment…………………………………………………………………….143
LIST OF FIGURES Figure Page 2.1 Relationship between changes in levels of uncertainty and the decision to divest………………………………………………………………..53 4.1 The case of increasing uncertainty and the impact of information asymmetry..145
4.2 The case of decreasing uncertainty and the impact of information asymmetry..145
CHAPTER 1
INTRODUCTION
Real options theory has primarily been a theory of investments (Reuer and Tong,
2007; Li et al., 2007). Two key predictions from real options theory that have been tested and supported time and again are: 1. firms would be deterred from making costly-to- reverse investments under conditions of high uncertainty (Campa, 1993; Guiso and
Parigi, 1999; Folta and O’Brien, 2004) and, 2. if investments were to be made under conditions of high uncertainty, staged/sequential investments will be preferred (eg., joint ventures as opposed a complete investment) because they provide flexibility to deal with uncertainty (Kogut, 1991; Cuypers and Martin, 2007).
Real options theory is still nascent in the area of divestments (Li et al., 2007).
There seems to be an implicit assumption that the predictions of real options theory, as developed for investment decisions, will apply in the context of divestment decisions.
This dissertation examines whether and to what extent do the predictions from real options theory hold in the context of decisions to divest business units. Further, the conditions under which this theory may have better or less explanatory power as compared to rival theories are also examined.
A dataset that captures information at both parent and segment levels is used to test the relevant hypotheses. The dataset includes firms that have engaged in segment
1 divestitures and a control set of firms that have not engaged in divestments. The sample period is from1980 to 2004. All kinds of divestments i.e., spin-offs (where equity in a divested unit is issued to existing shareholders of a firm pro-rata), equity carve-outs
(where equity in a divested unit is issued to new shareholders) and sell-offs (where the business unit is completely sold and becomes a part of a different firm) are included.
The first essay conceptualizes divestment of business units as real ‘put’ options
(Dixit and Pindyck, 1995). The first key prediction of real options theory i.e., high uncertainty in a business unit’s environment deters its divestment, is empirically tested.
Further, whether this deterring effect changes with changes in environmental uncertainty is also examined.
Probit models are used to tests these predictions. Results show that, indeed, high uncertainty in a business unit’s environment has a significant negative association with the decision to divest the business unit. Also, an increase in uncertainty accentuates this negative association of uncertainty with the decision to divest whereas a decrease renders uncertainty irrelevant to the decision to divest. Together, these results show that firms’ decisions’ to divest business units may be driven by option considerations, particularly when the conditions surrounding the divestment are uncertain. This result holds after controlling for all major theoretical explanations in the divestment literature.
It will be expected that the second prediction from the real options theory about flexible modes of governance under conditions of uncertainty will also hold for divestments as it does for investments. However, an equally powerful argument for
2 staged divestments comes from the information asymmetry perspective. According to this perspective, firms engage in staged divestments to reduce information asymmetry and maximize proceeds from sales (Zingales, 1995). The second essay builds arguments from real options theory and information asymmetry for staged divestments and tests their relative explanatory power.
These hypotheses are tested using a probit model with selection correction. This method takes the two decisions—the decision to divest and the decision to divest in a staged manner—together and jointly maximizes the likelihood. Possible selection effects between non-divesting firms that could potentially impact the decision to divest in staged manner are accounted for. The methodology also shows the close connection between the first and second essays. The first essay forms the theoretical basis for the selection equation in the second essay.
Results show that uncertainty is not significant to the decision to engage in staged divestments. This is a very interesting result because it highlights that real options theory may not apply the same way to divestments, as it does in the case of investments.
Therefore, it may be important to think of a ‘real options theory of divestments’. Further, information asymmetry has a significant positive association with the decision to engage in staged divestments. This shows that staged divestments may be means to mitigate adverse selection costs and maximize proceeds from sales under conditions of information asymmetry rather than options to deal with high uncertainty.
A third important question is about the relative importance of information asymmetry and uncertainty to the decision to divest per se. It has been argued that 3 information asymmetry has an adverse impact on the decision to divest (Myers and
Majluf, 1984). Also as shown in essay 1, from a real options perspective, the decision to divest may be driven by options considerations under conditions of high uncertainty.
However, information asymmetry may be less relevant to the decision to divest when there is a great deal of uncertainty about the information itself. Therefore, uncertainty and options considerations may dominate information asymmetry concerns, under certain conditions. Current literature does not consider information asymmetry and uncertainty together to explore these relationships.
The third essay brings these two competing theoretical perspectives together and explores the dynamic relationships between information asymmetry and uncertainty. In particular, under conditions of increasing uncertainty, uncertainty is expected to have a dominating effect over information asymmetry. Therefore, changes in the levels of information asymmetry are not expected to change the lack of significance of information asymmetry to the decision to divest. On the other hand, under conditions of decreasing uncertainty, information asymmetry concerns are expected to surface. Therefore, changes in the levels of information asymmetry should matter. More specifically, an increase in information asymmetry is expected to strengthen the negative relationship of information asymmetry with the decision to divest as compared to a decrease in information asymmetry. Further, even under conditions of increasing information asymmetry, uncertainty is expected to be negatively related to the decision to divest, on an average.
4 These hypotheses are tested using probit analyses. Results support the idea that under certain conditions, options considerations may be more important for the decision to divest than concerns arising due to information asymmetry.
Together these three essays thus make important contributions to the real options theory in general, and to divestment literature, in particular.
5 CHAPTER 2
DIVESTMENT AS A REAL OPTION: FIRM CHOICES UNDER CONDITIONS OF UNCERTAINTY
Real options theory focuses on the need for flexibility under conditions of uncertainty (Kogut, 1991). And, since strategic decisions are often made under conditions of uncertainty, real options theory has found broad application in the strategic management literature. Mergers and acquisitions (Warner, Fairbank and Steensma,
2006), joint ventures (Kumar, 2005; Reuer and Tong, 2005), entry into foreign markets
(Gilroy and Lukas, 2006), entrepreneurial ventures (O’Brien, Folta and Johnson, 2003;
McGrath, 1999) and other investment decisions have all been studied using a real option analog to a financial ‘call option’ i.e., the right to buy or invest in an asset at a future date without an obligation to do so.
The key implications of options theory in the context of real assets—that firms are be deterred from engaging in costly-to-reverse investments under conditions of high uncertainty and that investments are structured to deal with such uncertainty—have been largely corroborated (Chi and McGuire, 1996; Folta, 1998; Folta and Leiblien, 1994;
Reuer and Leiblein, 2000; Kogut, 1991). Several other aspects of strategic investment under conditions of uncertainty— such as different options embedded in investment decisions (Kulatilaka, 1995), competitive considerations in options logic (Smit and 6 Ankum, 1993), and interactions among various options (MacMillan and McGrath, 2002;
Vassolo, Anand and Folta, 2004)—have also been studied.
Divestments, on the other hand, are analogous to financial ‘put’ options i.e., the right to sell or divest assets at a future date without an obligation to do so. Real options theory has not been applied as extensively to the study of such ‘put’ options. Indeed, divesting real assets under conditions of high environmental uncertainty could lead to costly-to-reverse losses if uncertainty resolved in ways inconsistent with profit maximizing interests (Dixit and Pindyck, 1995). In this sense, the option to defer the decision to divest—a ‘put’ option—can be valuable. While some analytical work has focused on the option value associated with divestments under conditions of uncertainty
(Dixit, 1992; Myers and Majd, 1990), empirical work has been more limited. Much less is known about the actual value created by such ‘put’ real options and how such options are exercised.
In this paper, the key implications of these real ‘put’ options are developed and tested in the context of the divestment of business units. In particular, the relationship between the levels of uncertainty in a business unit’s environment and the likely decision to divest and, how the value of the ‘put’ option changes dynamically with changes in the levels of uncertainty, are examined. Consistent with the findings of real options theory in the context of investments, high environmental uncertainty was negatively associated with the decision to divest the business unit. This supports the idea that divestments are
‘put’ options on real assets.
7 With regard to the effect of changes in the levels of environmental uncertainty, when environmental uncertainty increased, uncertainty was still significantly negatively associated with the decision to divest. On the other hand, when environmental uncertainty decreased, uncertainty did not necessarily induce divestments. That is, uncertainty, in this setting, was irrelevant to the decision to divest—indicating a loss of
‘put’ option value. Therefore, the value of retaining the option to divest differs based on whether uncertainty increases or decreases over time. Together, these results support the notion that a firm’s reluctance to divest may be driven by option considerations when the conditions surrounding divestments are uncertain.
2.1 Theoretical background
Real options theory originated in finance. Myers’s (1977) seminal idea that firms’ discretionary investment opportunities can be viewed as ‘call’ options on real assets, analogous to financial call options, laid the foundation for this theory. The strategic management literature followed, with Kogut’s seminal work on joint ventures, focusing on the design of governance mechanisms for investments under conditions of high uncertainty (Kogut, 1983; 1985; 1991).
Decisions to invest in real assets under conditions of high uncertainty are analogous to financial ‘call’ options. As with financial ‘call’ options, real options on investments are claims or rights to buy an underlying asset at a pre-determined price, on or before a future date, without an obligation to do so. For real options under conditions of high uncertainty, call options are more valuable than an outright purchase. By
‘waiting-to-invest’ or by investing ‘stage-wise’, rather than making costly-to-reverse
8 investments, firms gain flexibility—acquire ‘call’ options—on investment opportunities under conditions of high uncertainty (Trigeorgis, 1988; Kogut, 1991). Such flexibility allows firms to benefit from the upside potential ––exercising the call options and making full commitment— if situations turn favorable and be protected from downside losses if situations turn unfavorable. Based on this logic, the key predictions of real options theory for investment decisions are: 1. firms are deterred from engaging in costly-to- reverse investments under conditions of high uncertainty, and, 2. flexible governance mechanisms will be chosen under conditions of high uncertainty.
These key predictions from real options theory in the context of investment decisions have been supported by a number of studies. For example, Guiso and Parigi
(1999) found in a sample of Italian manufacturing firms, that increasing uncertainty lowered firm-level planned investment scale. Campa (1993) reported a negative relationship between exchange rate volatility and the number of foreign entries into the
U.S. market. These results support the first prediction. Work on joint ventures and strategic alliances as desirable governance mechanisms under conditions of uncertainty are examples in support of the second prediction (Kogut, 1991; Chi, 2000;) 1.
1 Other research on real options theory of investments includes work on: ‘growth options’ (Kogut, 1983; McGrath, 1997), ‘switching options’ (Kogut, 1983, 1985, 1989), option interactions within the context of same investment (eg., growth versus deferral options (Folta and O’Brien, 2004; Kulatilaka and Perotti, 1998; Leiblein and Zeidonis, 2007)) or in the context of different investments (eg., pharmaceutical alliances as exploratory investments in real options (Vassolo et al., 2004)), the role of endogenous uncertainty and learning (Sanchez, 1993; Sanchez and Mahoney, 1996). Valuation implications of real options, however, are still a nascent area of study with mixed empirical evidence (Reuer and Tong, 2007). Interested readers may refer to Li et al, 2007 and Reuer and Tong, 2007 for extensive review of literature on real options and also guidance for future research. 9 Despite this growing literature on real ‘call’ options, there has been much less research on real ‘put’ options. Divestments are real ‘put’ options in the following sense.
Financial ‘put’ options are rights to sell assets at a pre-determined price (exercise price), on or before a future date, without an obligation to do so. Put options, therefore, are protections against falling asset prices. If a situation turns favorable and prices rise, there will be no need to exercise the put option. One can benefit from the upside by retaining or selling the assets at market prices. On the other hand, if a situation turns unfavorable and prices fall, the ‘put’ option can be exercised and the asset can be sold at the pre- determined price, which is presumably higher than the market price. This ensures protection from downside losses.
Similarly, ‘put options’ on real assets will also be valuable under conditions of high uncertainty. The key idea is that it might be valuable to ‘wait’ and retain the option, rather than divest, when the conditions surrounding divestments are highly uncertain
(Merton, 1998; Myers and Majd, 1990). This allows flexibility to gain from the upside potential and simultaneously prevent costly-to-reverse losses if uncertainty were to resolve in ways inconsistent with the decision to divest.
Of course, analyzing divestments as ‘put’ options has attracted some research attention. Myers and Majd (1990) modeled project divestments as ‘put’ options, with the value of the ‘put’ increasing with increasing uncertainty. According to Dixit (1992), divestments may not occur even though the underlying causes for investments no longer exist, and that such effects would be accentuated under conditions of high uncertainty and irreversibility. Also, Dixit and Pindyck (1995) argued that irreversible divestments will
10 be deterred under conditions of high uncertainty. Further, Chi and Nystrom (1995) suggested that greater endogenous uncertainty could lead to exit delays. Kogut and
Kulatilaka (2001) argued that managers in highly volatile environments hesitate to radically change their tightly-coupled organizations in the hope of better states of world.
There has also been some empirical testing of these ideas. A prominent empirical test infers the value of a firm’s abandonment options from its balance sheet and shows that such options have a positive effect on firm value (Berger, Ofek and Swary, 1996).
Other evidence comes from experimental studies by Bragger et al. (1998) who find that in a simulated economy, participants who received feedback with higher variability delayed exit decisions longer and invested more often than participants who received feedback with lower variability. Vassolo, Anand and Folta (2004) suggested that, as a base case, one can expect a negative relationship between uncertainty and the exercise of on-going options. Therefore, exit will be deterred under conditions of high uncertainty.
They test and support this hypothesis in the context of bio-tech alliances.
On the other hand, despite the large and growing literature on the divestment of business units by corporations, the idea that such divestments are like ‘put’ options and their exercise may be deterred under conditions of high uncertainty, is yet to be examined empirically. This approach has several possibly important implications for the study of divestments. For example, prior divestment research argues that poor divisional performance leads to business unit divestment. This may be because poorly performing units can put financial pressure on the parent corporation and also distract managerial attention from more profitable businesses (Ravenscraft and Scherer, 1991; Chang, 1996;
11 Duhaime and Grant, 1984). However, under conditions of uncertainty, current divisional performance may not be a good predictor of future performance. A real options approach in this setting suggests that divestment may not be forthcoming, despite low levels of current performance.
Also, much of the current literature on divesting business units attributes delays in the divestment decision to vested interests within a firm (Cho and Cohen, 1997) or other agency problems (Bethel and Liebeskind, 1993; Shimizu and Hitt, 2005). Organizational inertia--the inability of organizations to adapt quickly to changes in the external environment--is seen to stem, at least partly, from managers desire for status-quo and therefore can be considered an agency issue. A real options perspective suggests that, under conditions of uncertainty, it may make sense for a firm to delay making a divestment decision and in so doing keep its ‘put’ real options open, and thus that such delays have nothing to do with agency problems.
2.2 Theory development and hypotheses
2.2.1 Uncertainty and divestments as ‘put’ options
As outlined earlier, business unit divestments can be conceptualized as ‘put’ options on real assets (Myers and Majd, 1990). This is because when the conditions surrounding divestments are uncertain, the true value of business units may not be known, and their current value may not be a good indicator of their future value.
Divesting business units under such conditions could lead to several costly-to-reverse losses, and also prevent firms from benefiting from emerging opportunities.
12 If temporary market conditions were to dictate divestments, it could lead to loss of highly valuable tangible and intangible assets (Dixit and Pindyck, 1995). Such capital might have been acquired or built over long periods of time and may be very difficult or impossible to regain once lost (Carruth, et al. 2000; Dierickx and Cool, 1989). For example, specialized human skills and key customer relationships could be hard to reacquire; highly intangible assets such as brand image could be very difficult to re-build once lost. Further, in case of closely related business units, divestment of one or more business units could disturb valuable relationships and linkages across them. Such relationships may not be easy to re-establish if the previously divested business units were to be re-started. All these costly-to-reverse losses make profitable resumption of operations difficult, even if market conditions were to improve.
Even when there are no costly-to-reverse losses, divestments may still not be desirable under conditions of high uncertainty. This is because of the possible inability to gain from upside potential. For example, some technological breakthroughs could generate unexpected synergies. Firms may be unable to spot and capitalize valuable growth options—follow-on investment opportunities created by such synergies—once they are out of a particular business line. Even more importantly, firms may lose access to critical information when they get out of a particular business, even if they operate in other businesses in an industry. Such loss of access can prevent them from reentering and capitalizing on follow-on opportunities. Also, competitors could gain access to such opportunities and information disadvantaging firms’ in other related businesses as well.
13 These competitive considerations could be as important, if not more important, for divestment decisions as compared to investment decisions (Kester 1984; Smit and
Ankum 1993).
For these reasons, it might be worthwhile to wait rather than exit a business under conditions of high uncertainty i.e., treat divestments as ‘put’ options and deal with them flexibly as uncertainty evolves (Kogut, 1991). The value of such ‘put’ options will be higher when the uncertainty about the value of the underlying businesses is higher.
If firms, consider divestment of businesses as ‘put’ options on real assets, then
Hypothesis 1: Firms are less likely to divest a business unit when there is high
uncertainty in the business unit’s environment.
2.2.2 Changes in levels of uncertainty and the changes in ‘put’ option value
The negative relationship of uncertainty with the decision to divest, at a particular point in time ‘t’, also varies depending on whether that uncertainty is lower or higher as compared to a previous point in time ‘t-1—i.e., the option value changes with an increase or decrease in uncertainty over time. It would be easier to understand these dynamic relationships with the help of Figure 1.
From Figure 1, it can be seen that there are four possible situations for understanding the relationship between uncertainty and the decision to divest, when uncertainty changes over time. Situations in cell 1 and cell 4 are not of interest in this real options context because they do not involve changes in uncertainty. However, situations in cell 2 and cell 3, where there are changes in the levels of uncertainty from
‘t-1’ to ‘t’, are interesting for this analysis.
14 Cell 2 is a case of increasing uncertainty from‘t-1’ to ‘t’. In this case, all arguments made in the previous section about the need for flexibility and the option value in delaying divestments under conditions of high uncertainty hold. If anything, the option value should be greater for firms in this situation, since they experience an increase in uncertainty, as compared to those in Cell 3 that experience a decrease in uncertainty. In Cell 3, uncertainty decreases. In this setting, the value of a put option will decrease, or may even completely disappear. Taken together, these arguments suggest,
Hypothesis 2: Firms are less likely to divest a business unit when uncertainty in
the business unit’s environment increases than when it decreases.
Of course, there are other factors—even within the real options framework—that impact the decision of a firm to divest a business unit under conditions of uncertainty.
For example, both the “exercise price”—i.e., the price at which a divestment can occur— and the “spot price”—i.e., the value that a firm places on a business unit to the rest of its organization— impact the decision about whether or not to divest. However, since these factors are difficult to measure and observe—especially in large sample archival studies—they are not included in this discussion. In looking at the relationship between changes in uncertainty and the value of real ‘put’ options, the impact of these changes on the strength of the relationship between uncertainty and the decision to divest is central, over and above the potential impact of the exercise and spot price on the decision to divest.
15 2.3 Method
The probability that a firm engages in a divestment decision is estimated by a probit. In this method, the decision of the ith firm to divest or not divest depends on an unobservable utility index I i that is determined by explanatory variables X ij in such a way that the larger the value of the index I i, the greater the likelihood that the firm engages in a divestiture. This index I i can be expressed as
I i = β0+ βj X ij where Xij refers to the different independent variables in the model, with subscript ‘i’ referring to the particular firm in question and subscript j=1…n refer to the independent or explanatory variables in the model.
This index is related to the actual decision to divest as follows. If Y=1 for a firm engaging in divestiture of a business unit and Y=0 for a firm that does not divest at all, it can be assumed that there is a critical threshold for each firm. I i * such that if I i > I i* then the firm divests and otherwise it will not. This threshold I i* is not observable. However, it can be assumed to be normally distributed with the same mean and variance. With this assumption of normality, the probability that I i* is less than or equal to Ii can be computed using the standardized normal CDF as
-t2/2 Pi=Pr(Y=1)=Pr(I i*<= I i)=F(I i)= 1/ √(2 Π) ∫ e dt where the limits on integration run from (-∞, I i] or alternatively (-∞, β0+ βj X ij ] and ‘t’ is
- a standardized normal variable i.e, t~N(0,1). Ii is obtained by taking the inverse I i =F
1 -1 (I i)= F (P i)= β0+ βj X ij .
16 For the purposes of estimation, this normit is converted into a probit and I i estimated using
Ii= β0+ βj X ij + u i
2 where u i is the stochastic disturbance term . The model for estimating the divestment decision for the ith firm can therefore be written as
Ii = β0+ β1*segment industry uncertainty+ β2-12 * controls+ u
2.3.1 Data
All forms of divestments—sell-offs, spinoffs, equity carve-outs were considered for analysis. Preliminary data were obtained from Security Data Corporation’s (SDC)
Mergers and Acquisitions and New Issues databases. The sample was obtained after cleaning the initial list. The data filters used (See Appendix) are those that are standard in the finance literature on divestitures (Chen and Guo, 2005; Powers, 2003). Sell-offs were identified by dropped segments in a company’s portfolio matched against announcements in the SDC database (Schlingemann, Stulz and Walkling, 2002). All deals were verified by searching through SEC filings, Factiva news articles/newswires, and on Lexis-Nexis. The sample period was from January 1980 to December 2003.
Only completed deals were included in the sample because intentions that do not materialize later on cannot be considered as decisions. Deals with no match on
Compustat were also deleted from the list. This was to ensure availability of adequate
2 The probit estimation procedure is adapted from Gujarati, D. N. 1998. Basic Econometrics, Chapter 15, pg. 491-493. 17 accounting information for analysis. Also, since this analysis required data at the segment level, only those companies with corresponding matches in the Compustat segment database were included.
Only the first decision on a particular segment was taken into account (i.e., if a segment was found to be both in the spin-off/carve-out and sell-offs samples, the first of the deals was taken into account to avoid any confounding factors affecting the follow-up decision). Also, only parent firms which are in the manufacturing and related categories
(as indicated in the Robins & Wiersema (1995) grouping), other than utilities (SIC codes
4000-4999) were included 3. Utilities are in the regulated industry category, and therefore
3 A primary SIC code is assigned by Standard & Poor’s to each company in the COMPUSTAT databases according to its primary business activity (as determined by revenues). Therefore, a company’s primary SIC code may change depending on the change in the proportion of sales contributed by particular segments. When this change is affected retrospectively, it is possible that there could be differences between what has been the primary SIC code of the firm in a particular year and the one that has been assigned retrospectively. Segments, on the other hand, are identified based on the primary and secondary products and retain their SIC codes. Segments SIC codes could change when they are merged with other segments and reorganized. However, there is no evidence that these changes in SIC codes are retrospective. Therefore, segment SIC codes are more reliable. Impact on the parent’s portfolio relatedness measure: There is NO impact of a possibly different historic SIC code for the parent firm on the calculation of the parent’s portfolio relatedness measure and the segment’s relatedness measure. This because parent SIC code does not enter the calculation of these measures at all. There are two measures of parent’s portfolio relatedness that have been calculated. Their calculations are detailed below: 1. The Robins and Wiersema (1995) measure of firm’s portfolio resource based relatedness: According to Robins and Wiersema, for each combination of two different industry categories ‘i’ and ‘j’ in a firm’s portfolio, the sales-weighted measure of interrelationship Rij is given by R ij =P i r ij + P j r ij , where Pi= percentage of sales in industry category i and Pj= percentage of sales in industry category j. These weighted measures of similarity between paired industries (Rij) are summed over all combinations of two industries that could be formed in a business portfolio of the firm resulting in an aggregate index of interrelationship of the businesses of the firm Mk= Σ Rij= Σ rij(Pki+ Pkj), where i and j represent any two different industries in which the firm k is active. The authors then introduce a correction factor to avoid double counting. This is the parent’s total portfolio interrelatedness. Segment’s interrelatedness is measured as the sum of the segment’s relatedness with other segments in the parent’s portfolio. 2. The measure of related entropy in a parent’s portfolio that reflects the level of diversification. The basic entropy index was computed by Jacquemin and Berry (1979) as E= ΣPi* ln(1/Pi). In this equation, E signifies the entropy measure and Pi is the proportion of a firm’s sale in SIC industry i. This is typically treated as a measure of total diversification when calculated at the 2-digit SIC level. The unrelated diversification is computed using the proportions at the 2-digit SIC level and subtracted from the 4-digit level to obtain a measure of ‘related’ diversification in the portfolio. 18 it would not be appropriate to treat them along with other industrial groupings. Only domestic parent firms were included to eliminate influences of differing institutional regimes. American Depository Receipts (ADRs), were excluded for a similar reason.
Limited partnerships that operate under different legal rules were also excluded.
A sample of matched firms that have not engaged in a divestiture was also obtained. Matched parents were found using the parent’s total assets as a matching criterion within the same relationship category matched at the same four-digit SIC code
(Chen and Guo, 2005; Krishnaswami and Subrahmaniam, 1999).4 The time period used for selecting the matched sample is the same as that for the divesting firms 5. There were some cases where there was no match found but still the original divesting firm was retained in the sample. There were 1115 matched parent-segments. Together, the final
As can be seen, these calculations involve the segment level SIC code information and not the parent SIC code. Therefore, changes in the historic SIC codes of a parent firm has no impact on the computation of these measures. Other measures such as uncertainty and information asymmetry are calculated at the segment industry relatedness category level. The industry relatedness category that a firm belongs to is determined by its SIC code. However, the uncertainty and information asymmetry measures are averages of all firms in the industry categories. Therefore, if firms do not get included in a category they belong to, it is also possible that some firms get included where they do not belong to and, on average, the differences may get compensated. The effect will be expected to be similar with regard to identifying the control firms that have been chosen based on the parent’s SIC code. Therefore, the potential source of error due to differences in the historic SIC codes is not expected to be of major concern. In any case, historic SIC codes have been verified for about 15 per cent of the sample. In the sub- sample, the industry relatedness category was different only for about 13 per cent of the firms. The movements from one relatedness category to another were not systematic. Therefore, it is not a major concern. 4Parent firm size and industry grouping at four digit level have been used as the matching criterion following common practice (Chen and Guo, 2005; Krishnaswami and Subramaniam,1999). Since parents belong to manufacturing and related industries, size measured by parent assets is more appropriate. Both the first best match and the next best matched parent firms that have not engaged in a divestiture were chosen in order to ensure a close to 2:1 ratio of matched versus divesting firms. 5 Matched firms were not necessarily chosen from the same year as the divesting firm, though matching firms in the same year were not eliminated by design. 19 sample had 182 sell-offs, 102 spin-offs and carve-outs, and 598 non-divested firm segments with no missing values for any of the required variables 6.
The data collection procedure clearly restricts the analysis to large publicly traded firms with listed segments. This leaves out several small firms not listed in the databases used. However, inclusion of small firms would only strengthen the results because small firms are susceptible to the influences of uncertainty more than larger firms. This study is therefore a conservative test of the hypotheses. Also, large private firms are not in this list, but such firms are relatively uncommon, at least in the U.S.
2.3.2 Other data sources
The data to calculate the uncertainty measure were obtained from the Center for
Research in Security Prices (CRSP) monthly stock price database. Data for calculating the information asymmetry measure were obtained from the summary statistics on the
Institutional Brokers Estimates System (IBES) database. Block-holder data was obtained from Thomson Financial’s Spectrum Institutional Stockholding database.
2.3.3 Variables and measurement
Dependent variable
The dependent variable is dichotomous where the divesting firm has been represented by ‘1’ and a non-divesting firm or the matching firm has been represented by
‘0’.
6 Reasonable efforts have been made to recover missing values from 10-ks of firms. 20 Control variables
The control variables for this study come from a wide variety of explanations and theories have been used to understand divestments. They are detailed below, briefly, along with the measures used.
Parent and segment performance
By far, the most dominant explanations for divestments in the literature are based on parent and segment/business unit level performance. Divestments have been seen as means to restore corporate efficiency. Therefore, poor parent performance has been argued to positively influence the decision to divest (Cho and Cohen, 1997; Harrigan,
1981, 1982; Jain, 1985; Montgomery and Thomas, 1988). Further, poor performance at the business unit level has also been found to be a key determinant of divestments
(Vignola, 1974; Patton and Duhaime, 1978; Ravenscraft and Scherer, 1991; Chang, 1996;
Duhaime and Grant, 1984; Hamilton and Chow, 1993; Hitt et al., 1996).
In this study, parent and business unit level performance were controlled for using return-on-assets. This measure was considered more appropriate for this study since firms operating in manufacturing and related industries are asset intensive. Return on assets was computed as operating income before depreciation over total assets
(Powers, 2001) and both the business unit and corporate levels. 7
7 Return on sales measure has been computed as operating income before depreciation over net sales and was used to check the robustness of results to alternative performance measures. Accounting measures of performance have been chosen as opposed to market measures since market measures are not available at the segment level. 21 Agency theory
Another major explanation for divestment decisions was based on agency theory.
Managerial self-interest and poor governance mechanisms were argued to adversely influence decisions to divest (Bethel and Leibeskind, 1993; Finkelstein and Hambrick,
1989; Jensen and Murphy, 1990). Strong governance mechanisms such as large block- holder ownership, higher number of outsiders on boards etc., were shown to reduce such tendencies and favorably influence divestments (Bethel and Liebeskind, 1993;
Hoskisson, et al., 1994; Sanders, 2001).
While there have been arguments for and against the monitoring efficacy of outside board members (Baysinger and Hoskisson, 1990), block-holders seem to have an incentive to monitor firms more closely. Therefore, agency explanations of divestments were controlled in this study by using the level of stock held by block-holders as a proxy 8. The normally accepted definition of block-holders as those who control more than 5 per cent of the firm shares has been used (Bethel and Liebeskind, 1993).
Transaction costs economics and behavioral uncertainty
Transaction costs economics is yet another theory applied to understand divestments. The core of the TCE argument is that, at any point in time, uncertainty about the behavior of partners to a transaction will have an impact on whether or not a transaction is internalized. Transactions with higher behavioral uncertainty that can be efficiently managed by ‘fiat’ will be internalized, and those that involve less behavioral
8 Measuring block holder ownership only by considering institutional owners could cause a slight bias because individual block-holders are not included. 22 uncertainty for which ‘fiat’ would be relatively costly, will not be integrated or will be divested (Hill and Hoskisson, 1987; Jones and Hill, 1988; Markides, 1992).
A measure of behavioral uncertainty should capture the magnitude of the coordination problems that arise due to transaction specific investments in recurrent transactions, particularly under conditions of uncertainty (Williamson, 1979; 1985). It thus can be measured as an interaction between transaction specific investments, frequency of transactions, and the levels of uncertainty. Unfortunately, several TCE studies have captured only one of these three dimensions of behavioral uncertainty and, the divestment literature is no exception (David and Han, 2004; Carter and Hodgson,
2006).
In this study, it was important to capture the behavioral uncertainty between the business unit and the parent firm in order to assess the transaction costs considerations in the decision to divest. If a business unit is very closely related to the other business units in the parent firm, it would be expected to have greater need for coordination with the rest of the firm. For example, if the business unit is a part of an integrated production process, there are likely to be more transaction specific investments between the parent and the business unit. Further, the frequency of interactions with other business units that are a part of the same process are also likely to be higher. On the other hand, if the business unit is a stand-alone and is less connected with other units, there will be lesser need for coordination since transaction specific investments will be fewer and also interactions with other units will be relatively infrequent.
23 Therefore, a business unit’s relatedness to the parent company can be a proxy for the level of transaction-specific investments and frequency of interactions between the business unit and the parent firm. The interaction of the business unit’s relatedness with parent firm and, the level of uncertainty in the segment’s industry can thus proxy the behavioral uncertainty. Controlling for this variable became important in order to separate the real options explanation (based on environmental uncertainty) from that of
TCE (behavioral uncertainty).
Information asymmetry
Information asymmetry is another factor, primarily in the finance literature, that has been shown to influence divestment decisions. There are arguments for both positive and negative effects of information asymmetry, depending on over-valuation or undervaluation of the parent’s stock, managerial intentions, and the type of divestment being studied (Myers and Majluf, 1984; Nanda, 1991; Vijh, 2002; Powers, 2003).
Generally, information asymmetry about a parent firm is taken as a measure. However, if a business unit is being divested, whether or not the unit will be properly valued will be a function of whether or not there is an appropriate comparison available in this business unit’s industry.
Therefore, the relevant measure would be the level of information asymmetry in the business unit’s industry. The level of a business unit’s industry information asymmetry was measured as the mean value of the standard deviation in analysts’ forecasts from IBES for the business unit’s industry grouping in a particular year
(Krishnaswami and Subramaniam, 1999). Parent’s valuation was measured as market
24 value to net sales. Parent’s market value was computed as the price at the end of the fiscal year multiplied by the total outstanding common stock. Another measure of valuation was the parent’s market-to-book ratio. 9
Parent and business unit relatedness
Further, the level of parent’s portfolio interrelatedness and the level of relatedness of the business unit to be divested with the parent firm were other important determinants of divestment decisions. Greater levels of relatedness, in general, have been found to be negatively related to divestments (Hoskisson, Johnson and Moesel, 1994; Chang and
Singh, 1999).
The level of diversification was measured using related entropy (Jacquemin and
Berry, 1979) 10 since parent’s resource-based relatedness was highly correlated with several other variables of the study. In the present case, the relationship between the particular division in question (the division divested via spin-off, carve-out or sell-off) and the other divisions within the parent company was important. Therefore, a sales-weighted measure of interrelationship of the focal division with the other divisions in the parent firm was calculated following Robins and Wiersema (1995) 11 and then
9 Market-to-book ratio was obtained by summing market value of shares, preferred stock, parent’s long term debt, total current liabilities net of cash and divided by the parent’s operating income before depreciation. Parent’s market value was computed as the price as the end of the fiscal year multiplied by the total outstanding common stock. Another measure used was the parent’s market-to-sales ratio. 10 The basic entropy index was computed by Jacquemin and Berry (1979) as E= ΣPi* ln(1/Pi). In this equation, E signifies the entropy measure and Pi is the proportion of a firm’s sale in SIC industry i. This is typically treated as a measure of total diversification when calculated at the 2-digit SIC level. The unrelated diversification is computed using the proportions at the 2-digit SIC level and subtracted from the 4-digit level to obtain a measure of ‘related’ diversification in the portfolio. 11 Robins and Wiersema (1995) developed a resource-based relationship index to measure a firm’s overall portfolio interrelatedness. According to Robins and Wiersema, for each combination of two different industry categories ‘i’ and ‘j’ in a firm’s portfolio, the sales-weighted measure of interrelationship Rij is given by R ij =P i r ij + P j r ij , where Pi= percentage of sales in industry category i and Pj= percentage of sales in industry category j. These weighted measures of similarity between paired industries (Rij) are summed 25 summed to get the extent of resource-based relatedness of the segment with the parent firm.
Other controls
Other important variables that needed to be controlled, following previous literature, were parent debt position (leverage), firm size, and business unit size. Parent’s debt position (leverage) was measured as total long term debt over total common equity
(lev1) and long term debt over market value of equity (lev4), again to check the robustness of results to alternative measures (Chang and Singh, 1999). Firm size was measured as log of total assets and business unit size was measured as the proportion of the parent’s total assets that are in the segment i.e., segment assets/parent’s total assets
(Duhaime and Baird, 1987; Bergh, 1995). 12 Again, an asset based measure was considered more appropriate due to the nature of the firms in this study.
Also, a measure of growth options in other business units/segments of the firm was calculated using the market-to-book value of the median firm in the industry to which the business units belonged to. This median value was weighted with the segment’s proportion of the parent firm’s sales. The weighted sum was taken as a
over all combinations of two industries that could be formed in a business portfolio of the firm resulting in an aggregate index of interrelationship of the businesses of the firm Mk= Σ Rij= Σ rij(Pki+ Pkj), where i and j represent any two different industries in which the firm k is active. The authors then introduce a correction factor to avoid double counting. This is the parent’s total portfolio interrelatedness. Segment’s interrelatedness is measured as the sum of the segment’s relatedness with other segments in the parent’s portfolio. The parent’s resource-based portfolio interrelatedness measure was very highly correlated with segment relatedness calculated using the resource-based measure. Therefore, the entropy measure was used for parent’s portfolio relatedness. Results on key variables are robust when the segment relatedness is dropped and the parent’s resource based portfolio measure was used. Parent’s resource-based relatedness measure was significant itself in negative direction as expected. 12 Controlling for the total number of segments was considered. However, just that number would not provide any more information than what is obtained from firm size and the measures of relatedness. 26 measure of the total growth options in other businesses of the firm. All variables were taken as of the financial year preceding the event and the level of information asymmetry was considered with a one-period lag.
Independent variables
Uncertainty
Uncertainty is a key variable in this model and a time-varying estimate of business unit’s/segment’s environmental uncertainty was needed. It is a common practice to quantify the construct of uncertainty by calculating the variance of indicators such as stock price, GDP or sales over time. Such approaches fail to account for trends in data that can increase the measured variance while actually not being an element of uncertainty if they were predictable. Also, such approaches do not accommodate the possibility of variances being heteroskedastic that is very characteristic of time series.
Following Folta and O’Brien (2004), Carruth et al., (2000), generalized autoregressive conditional heteroskedasticity models (GARCH) were used. The conditional variances generated were used as a measure of uncertainty (Bollerslev, 1986;
Engle, 2001). In particular, the GARCH-M (1,1) models were run on value-weighted industry portfolio 13 returns that were developed from monthly stock returns (adjusted for dividends and splits) for all firms in the CRSP database from 1950-2004. 14
13 All industry groupings are based on Robins and Wiersema (1995) classification. 14 The data has been checked for white noise using the Portmonteau’s Q-test, the correlograms and Bartlett’s periodogram based white noise test. Where there was evidence of white noise, an ARIMA (0,0,1) term was introduced to mitigate the situation and the tests were repeated to confirm white noise. 27 The GARCH-M model can be written as follows:
rt = α+γht-1+ρrt-1+δε t-1+εt 2 ht = κ+ρ1 h t-1+ δε t-1 εt = sqrt( h t zt) and zt ~ N(0,1) This model represents the generalized autoregressive conditional heteroskedasticity in-mean specification, GARCH-M, with ARMA (1, 1) in the mean equation. ‘ εt’ representing the error term, is conditionally normally distributed and serially uncorrelated. ‘h t’, the conditional variance, is a linear function of the past period’s
2 15 squared errors, εt-1 , and the last period’s conditional variance, h t-1, i.e., GARCH(1,1) .
The ARMA (1,1) in the mean equation implies that the conditional returns in this model are a linear function of the last period’s conditional variance, past conditional returns and past disturbance.
Under this rich specification, volatility can change over time and expected returns are a function of volatility as well as past returns. The monthly conditional variances were averaged to obtain annual figures. The variable business unit’s environmental uncertainty is computed as the square root of the average yearly conditional variance. 16 The lagged variable on uncertainty was used by considering the uncertainty in the year preceding the closest financial year to the event date.
15 In fitting time series models, the simplest models are fitted first. Higher models with more number of autoregressive terms call for greater number of parameters. They are not used unless there is a good reason to believe that the model specified is not capturing the process sufficiently. Here, the GARCH-M(1,1) captures the underlying process well, as in case of several other studies cited, and therefore there was no need to use an alternative specification. 16 Results do not change with the use of alternative measures such as the variance in monthly returns in the year prior to divestment and the average variance in returns in three years prior to divestment. 28 Changes in uncertainty
The change in uncertainty is calculated as a first difference of the GARCH measures for period ‘t-1’ and ‘t’. A positive difference between the uncertainty at ‘t-1’ and at ‘t’, indicates a decrease in uncertainty and a negative difference indicates an increase.
2.4 Analysis and results
The summary statistics for divesting and non-divesting parent-segment combinations are in Table 2.1. On average, parents that have divested a business unit seem to have higher percentage of shares held by institutional block-holders, greater levels of diversification in their portfolios, higher leverage ratios and larger firm sizes as compared to non-divesting firms. Also, divested segments are larger in size than non- divested segments.
Further, parents that have divested a business unit seem to have lower uncertainty in the business unit’s industry, poorer performance, lower behavioral uncertainty between the parent and business unit, lower information asymmetry in the business unit’s industry, and lower parent firm valuations as compared to non-divesting firms. Also, business units that have been divested seem to be less related to their parent firms’ than those business units that have not been divested. Some of these differences are significant, others are not. The t-tests for significance of differences between means are also reported in table 1.
The correlation matrix in Table 2.2 shows no major problems of multi-collinearity among the variables except for behavioral uncertainty and business unit relatedness
29 (0.9646) indicated by * in the table. 17 The result on the key variable, business unit’s environmental uncertainty, holds with and without the measure of behavioral uncertainty.
Therefore the results and their interpretation in this study do not change with the presence or absence of this variable.
Table 2.3 and Table 2.4 show the probit results and marginal effects,18 respectively, for testing hypothesis 1.19 Results for hypothesis 2 are in Tables 2.5-2.9.
Tables 2.5 and 2.6 show the probit results and marginal effects, respectively, for sub- sample A; Tables 2.7 and 2.8 show these results for sub-sample B and Table 2.9 shows the probit results from the combined model using dummy variables. For non-linear models such as probit, the coefficients from probit would not reveal the actual impact of a variable (Train, 1986; Hoetker, 2007a). The economic significance of variables, or lack thereof, is revealed by the marginal effects. Therefore, only the marginal effects have been interpreted. 17 Parent’s resource-based portfolio relatedness measure is significantly correlated with several other variables, including business unit’s relatedness. Therefore, this variable was replaced by the parent’s related entropy measure of portfolio relatedness to avoid the problem. Even parent performance and segment performance are correlated upto 0.66. However, results are robust to alternative measures of parent and segment performance. 18 Marginal effects are reported at the median level of other variables in all tables 3-6. Some authors (Train, 1986) recommend other ways such as calculating the effects for each observation and summing up. Since the comparisons here are about the significance and not so much about the size, marginal effects at median would suffice. Even when comparisons are made about coefficients for hypothesis 2, such comparisons are only about significance, not size of the effect. 19 While there were multiple divestments for few firms, such observations were very few to treat this as a panel dataset. Only the first decisions related to a particular segment divestiture have been included. Also, there were several restrictions imposed on the characteristics of both segment and parent firms that could have reduced the number of observations per firm, apart from the general fact that divestments are relatively few compared to the overall population of firms. There were 59 out of 233 non-divesting firms and 13 out of 258 divesting firms that had multiple observations for a parent firm. Any estimation would thus actually not be meaningful to detect firm-effects. Also, since the estimation technique is a probit and not an OLS, these methods have their own issues (Greene, 2006). All results, however, are based on robust standard errors and so are corrected for firm effects. The results were robust to clustering on parent industry category as well as the segment’s industry category (Petersen, 2007). 30 In all models, a divesting firm was represented by ‘1’ and a non-divesting firm was represented by ‘0’. In each table, model 2 is a variant of model 1 with a different leverage measure. Models 3 and 4 are variants of models 1 and 2 respectively, and control for transaction costs economics explanation using ‘behavioral uncertainty’ variable.20 Models 5-8 are counterparts of models 1-4 with controls for growth options in other businesses.
From Table 2.4, it can be seen that some predictions from past theories were confirmed in this study whereas others were not. These marginal effects show that while parent performance had a significant negative association with the decision to divest, in most models, business unit performance was not significant at all. This result calls into question the hitherto accepted idea that poor performance at the business unit level is a trigger to divestment decisions. It seems that overall corporate concerns are more relevant to the decision to divest.
Also, block-holder equity, a proxy for effective governance, was not significant in any of the models. This is an important result since agency theory is a dominant one in the divestment arena. Furthermore, the behavioral uncertainty variable that captures TCE concerns was not significant. However, since this variable is an interaction term, its interpretation in the probit model is not straightforward.21
20 The model here has an interaction term, but the results hold regardless of the presence of the interaction term. Also, the standard errors are not inflated much with the introduction of new variables. Together, these suggest there are no serious multi-collinearity problems in this data. 21 The effect of the interaction is a function of not only the coefficient for the interaction term, but also the coefficients of each of the interacted variable and the values of all the variables. The implications of this are as follows: 1. the sign of the interaction coefficient may not indicate the direction of the interaction effect. The entire interaction effect must be calculated at a given value. 2. the significance of the interaction effect is not determined just by the significance of the interaction coefficient. The interaction effect could be significant for some observations and not for others (Hoetker, 2007a). Since, TCE is not the 31 Information asymmetry also did not seem to matter for the decision to divest in any of the models in Table 2.4. This result changes with changes in the levels of uncertainty as compared to a past period (more on this later). With regard to parent’s valuation, the sign was sensitive to the measure used. A sales based measure of parent valuation had a significant negative association with the decision to divest whereas the market-to-book measure showed a positive association.
Further, in all models, greater levels of entropy or less relatedness in a parent’s portfolio of businesses were positively associated with the decision to divest. Business unit relatedness was negatively associated in models 1-6 but was not significant when both behavioral uncertainty and growth options in other businesses were controlled for.
Parent’s debt position was negative and significant in most models and was sensitive to the measure used. Also, in all models, parent firm size had a significant positive association with the decision to divest. Business unit size was positively associated in models 1-4.
From Table 2.4,22 it can be seen that the marginal effects of business unit’s environmental uncertainty were negative and significant at 5 per cent level in all models.23 This supports the core argument using real options logic that under conditions of high uncertainty in a business unit’s environment, firms will be less willing to divest the business unit (hypothesis 1). This result was obtained after controlling for the effect
focus of this study, the effect or lack thereof, of this variable has not been explored further. The results of the key variable of interest i.e., the uncertainty variable are significant with or without the presence of the interaction term. 22 It is possible to obtain marginal values greater than 1, when the slope of a curve changes rapidly. 23 The economic significance of uncertainty is that a one unit increase in uncertainty decreases the likelihood of divestment by about 2.6 to 3 per cent. 32 of most of the important parent and business unit level variables. To test hypotheses 2 24 -- whether an increase or decrease in uncertainty from a previous period changes the effect of uncertainty on the decision to divest— the sample was split into two sub-samples A and B.25 Sub-sample A included firms that experienced increase in uncertainty in the
24 Testing hypothesis 2 involves a comparison of the effect of a particular covariate between two groups i.e., groups with increasing and decreasing uncertainty. Comparing coefficients between groups is quite straight-forward in the case of linear regressions. The groups can be modeled together in a single regression, using dummy variables to distinguish between groups, and coefficients can be compared. However, in case of non-linear models such as probit, this procedure is not appropriate at least for two reasons: 1. Dummy interactions by themselves make interpretation difficult and 2. Comparing coefficients needs the assumption that each group has the same residual variation. Comparing two groups even when there are minute differences in residual variances can lead to incorrect inferences. Therefore, it is important to detect and correct for residual variation. Allison (1999) suggested methods to deal with residual variation. Hoetker (2007b) has shown that these methods may also not be very good in detecting residual variation. Regardless of such problems, the most important problem with Allison’s method is with regard to identifying the difference in the impact of a specific covariate. His method requires the very stringent assumption that other coefficients much be equal in both groups. So, in order to test the effect of any particular covariate, all others coefficients need to be assumed to be equal and this is an untestable assumption. That gets into a circular logic and limits the applicability of Allison’s method (Hoetker, 2007b). A way to get around these problems of residual variation, that make comparison difficult, is to find methods that do not depend on equality of residual variance. One such way is to model the two groups separately. This will give consistent coefficients and standard errors within each group (It is not hard to see that estimating two equations for the groups using dummy interactions, within a combined model will also produce similar coefficients, however comparison cannot be as straightforward, as noted above). These coefficients can then be compared for their significance in the respective groups, and inferences can be drawn based on the patterns of significance. This method will be particularly useful when the differences in significance levels are vastly different (not a comparison between p<.11 and p<.09). However, using this method, the magnitudes of the coefficients cannot be compared i.e., if the coefficients on a particular covariate are significant in both groups in negative (positive) direction, then there will be no meaningful information beyond the fact that the coefficients are significant in both groups in negative (positive) direction. Modeling two groups separately, to compare significance of coefficients across the groups, will be more appropriate when the group membership is based on an exogenous variable and the sample sizes are comparable. Another alternative that does not need the assumption of equal residual variation is comparing the ratios of coefficients of any two coefficient pairs in both groups. However, in this method 1) even very large differences in ratios may not be statistically significant particularly if one or more terms are estimated with poor precision, and 2) much larger sample sizes may be needed to compare ratios than comparing individual coefficients. Also, theoretically one needs to draw hypotheses that involve comparing ratios of coefficients. Comparing hypotheses that are in terms of ratios of coefficients on covariates is not the purpose of this study. Therefore, comparing the coefficients between two samples by modeling them separately was more appealing in the context of this study, particularly because it would amount to looking for extreme results to support the hypotheses. Interested readers are referred to Hoetker’s working paper (2007b) on these matters. 25 In this study, the membership of the sub-samples A or B depends on an exogenous variable i.e., whether or not industry uncertainty has changed from a previous period. Hence, there are no selection concerns 33 business unit’s environment at time ‘t’ as compared to uncertainty at ‘t-1’. And, sub- sample B included firms that experienced decrease in uncertainty in the business unit’s environment at time ‘t’ as compared to uncertainty at ‘t-1’. In sub-sample A, uncertainty in a business unit’s environment was expected to have a stronger negative correlation with the decision to divest as compared to sub-sample B.
Marginal effects for sub-sample A and sub-sample B are in tables 2.6 and 2.8, respectively.26 Results for sub-sample A (Table 2.6), show that uncertainty in a business unit’s environment was negative and significant (at p<0.05) to the decision to divest.
Further, results for sub-sample B (Table 2.8) show that uncertainty was not significant in any of the models for this sub-sample. This allowed comparison of relative effects of uncertainty between sub-samples A and B, because the difference in significance, one being significantly negative and another being insignificant, permits such comparison.27
The results suggest that firms under conditions of increasing uncertainty are less likely to divest as compared to those firms under conditions of decreasing uncertainty— i.e., the
‘put’ option value increases. Thus hypothesis 2 is supported. The results also show that decrease in uncertainty can decrease the ‘put’ option value and hence uncertainty may become irrelevant to the decision to divest. The key insight from these results was that
here. Also, both sub-samples are large. This made it worthwhile to model the two groups separately and compare the significance of variables. 26 A comparison of probit results will suffice in this context because the concern is not about the size of the coefficients but only about the significance. 27 It is also meaningful to compare the significance of uncertainty between sub-sample A and sub-sample B, because the comparison is between vastly different p-values ( not a comparison between p<.09 and p<.11) . Also in this study, the models include variables from all important theories and explanations and therefore the models are fairly well-specified. 34 the relationship between uncertainty and the decision to divest differs based on dynamic changes in uncertainty.
Further, in sub-sample B where firms experienced decrease in uncertainty, the level of information asymmetry in a business unit’s industry was significantly negatively related with the divestment decision. This was not the case in sub-sample A. The size and significance of other variables were also different in sub-samples A and B (More about this in discussion section.)
Table 2.9 shows the results from probit for two groups, sub-samples A and B, within a combined model (with the use of dummy variables). It can be seen that the coefficients and significance of variables in such models are similar to those obtained from estimating the sub-samples separately. However, as noted, comparing coefficients for understanding the differing effects of uncertainty in the two groups is fraught with problems related to residual variation between them. Therefore, these results were not used for comparing the coefficients.
In terms of overall model fit, for all full-sample and sub-sample analyses, the log- likelihoods were comparable for models 1-4, and this statistic was smaller for models 5-8 that controlled for growth options in other segments. All models had a discriminatory power of about 68-70 per cent (lroc). The Hosmer-Lemeshow statistic rejected the null that the model was not a good fit for all models and, therefore, the models are a good fit for the data.
35 2.5 Discussion and conclusion
This study extended and applied real options theory to the phenomenon of divestments. Divestments have been considered as ‘put options’. This study showed that firms are less likely to divest businesses when there is significant environmental uncertainty, compared to when there is less uncertainty. These results are consistent with the key prediction of the real options theory that under conditions of high uncertainty, firms are apparently reluctant to engage in divestments but, rather, prefer to keep their divestment “options open”. This result holds even when controlling for other explanations of divestment, including poor performance in a divested business, information asymmetry between a firm and potential acquirers, and so forth.
Further, the results of this study show that the option value changes with changes in the levels of uncertainty. In situations of increasing uncertainty in a business unit’s environment (sub-sample A), uncertainty was still negatively related with the decision to divest (indicating the continued existence of option value). On the other hand, when there was a decrease in uncertainty as compared to a previous period (sub-sample B), uncertainty did not have any significant relationship with the decision to divest. This means that when uncertainty decreases, uncertainty will become irrelevant to the decision to divest (indicating that the ‘put’ option value no longer exists).
Also, previously it has been shown that poor performance at parent and business unit levels had a positive impact on the decision to divest (Duhaime and Grant, 1984;
Hamilton and Chow, 1993; Cho and Cohen, 1997). Results from this study support past findings that poor performance at corporate parent level favors the decision to divest. On
36 the other hand, business unit performance was not a significant determinant of the decision to divest in any of the models in both full sample and sub-sample analyses.
In a way, this result supports the real options reasoning that if business units are, indeed, considered ‘options’, even poor-performing units may be retained if their future prospects are brighter. On the other hand, even well-performing units may be divested if future growth prospects are poor. Further, in very complexly related business portfolios, it may even be difficult to assess the true contribution of a business unit adequately.
Even though a particular business unit may be performing poorly, its contribution to the profitability of other business units, in terms of shared customer base etc., might be so valuable that such business units get retained despite their poor performance. In other instances, it could be the case that a very profitable business unit may have to be divested because it cannibalizes the business of other units.
Further, according to previous literature, failure to divest a business that was performing poorly was generally attributed to agency problems inside a firm. Large block-holders were thought to positively influence divestment activity, since such block- holders can reduce agency problems in a firm (Bethel and Liebeskind, 1993; Hoskisson et al., 1994). However, in the models presented here, block-holder equity was not a significant determinant of divestment. This suggests that lack of divestment by firms need not entirely be motivated by self-interest seeking behavior on part of managers, but, in fact, may be consistent with the interests of equity holders. On the other hand, it could also mean that block-holders may not be able to guide the decision making in firms under conditions of high uncertainty.
37 Also, previous empirical studies using transaction cost analyses of divestment suggest that high uncertainty leads to less divestment (Bergh and Lawless, 1998). This argument is theoretically incomplete since uncertainty alone is never the central focus in transaction costs analysis. However, transactions costs economics does suggest that high uncertainty about the behavior of exchange partners leads to more hierarchical forms of governance, i.e., less divestment. In this paper, uncertainty about the behavior of exchange partners and uncertainty about industry prospects of a business unit were measured separately. When both types of uncertainty were included in the analysis, uncertainty about the industry prospects was statistically significant and uncertainty about exchange partner behavior was not significant. This suggests that real options considerations are important for divestment decisions under uncertainty, and that threats of opportunism may be less important. However, the result on behavioral uncertainty is not definitive. Further analysis is needed to interpret since this construct was represented by an interaction term in the probit model.
The results of this study also throw light on the roles of environmental uncertainty and information asymmetry on the decision to divest. Information asymmetry was not significant in models that did not account for a reduction in uncertainty in the business unit’s environment as compared to a previous period of time. Previous theory suggests reasons why information asymmetry could influence divestments (Myers and Majluf,
1984). None of the prior empirical studies, however, account for uncertainty and information asymmetry in the same models.
The results presented here show that in the presence of high uncertainty,
38 information asymmetry could be of less concern. However, when uncertainty decreases, information asymmetry could be an important determinant for the decision to divest.
This relationship is investigated in the third essay of this dissertation. Also, if divestments are more likely to occur under conditions of relatively less uncertainty, information asymmetry could have an impact not only on the divestment decision, but also on the way the divestment is carried out. Thus, the relative influences of uncertainty and information asymmetry are worthy of further research and can substantially improve our understanding of the divestment process.
The other control variables—parent’s level of diversification, leverage, and firm size—were all significant in the expected direction. The consistency of these results with prior research suggests that not only do some major explanations of divestment hold in the current sample, but that even controlling for these explanations, the level of uncertainty still was negatively associated with the propensity to divest. Again, the sub- sample analyses show that the importance of these factors could vary depending on the changes in the levels of uncertainty. This could be another area for further enquiry.
Of course, this study has limitations. This study focuses on the causes of divestment rather than the process of divestment. It is important to know not only ‘why firms divest?’ but also ‘how firms divest?’ This means studying divestment modes—for example, selling-off of a unit or spinning off/carving out a unit—as well. While these modes of divestment have been studied extensively in finance literature, very little has been done to examine these decisions controlling for the determinants of divestments in the first place (Allen and McConnell, 1997; Krishnaswami and Subramaniam, 1999;
39 Lang, Poulsen and Stulz, 1995; Vijh, 2002 etc.). A selection model appears to be promising approach to understanding the divestment process more comprehensively
(Allison, 1984; Chang and Singh, 1999).
This paper also did not include any discussion on the ‘dueling’ nature of growth and deferral options, a topic that has been of substantial interest to scholars in the field of real options. In the context of investments, deferral options have an adverse affect on firms’ propensity to invest, whereas the presence of growth options actually induce firms to invest. This causes the ‘dueling’ effect (Kogut, 1991; Folta and O’Brien, 2004;
Kulatilaka and Perotti, 1998) on investment behavior. In the context of divestments, however, the presence of deferral options to divest, and the presence of growth options in the particular business to be divested, both induce a firm not to divest (provided losses are not large enough to destroy any option value). Hence, there need not be any ‘dueling’ effect between ‘growth’ and ‘deferral’ options in the case of divestments. If there would be any ‘dueling’ at all, it might arise from the presence of ‘growth’ options in other businesses within a firm’s portfolio. Such growth options should positively influence firms to divest the businesses on which deferral options to divest are held. This study finds no support for this type of ‘dueling’ options.
40 t-test for difference in Firm type Non-Divesting Firms Divesting Firms means Variable Obs Mean Std. Dev. Obs Mean Std. Dev. Uncertainty in business unit’s environment 598 0.06 0.01 284 0.06 0.01Significant Parent ROA 598 0.12 0.17 284 0.08 0.16 Significant Segment ROA 598 0.09 0.25 284 0.03 0.29 Not significant Institutional Block-holder Share 598 10.45 13.71 284 12.13 13.19Significant Behavioral Uncertainty 598 0.02 0.03 284 0.01 0.02Significant Business unit’s industry information asymmetry 598 0.33 2.28 284 0.21 0.28Not significant Parent's Valuation (sales measure) 598 1.95 6.66 284 1.07 1.44Not significant Parent’s Level of Diversification (Related Entropy) 598 0.05 0.18 284 0.09 0.2Significant
Parent's portfolio Resource-based Relatedness 598 0.51 0.47 284 0.3 0.4Significant Business unit's resource-based relatedness 598 0.37 0.55 284 0.19 0.36Significant Parent Debt (lev1) 598 0.19 0.16 284 0.23 0.16Significant Parent Debt (lev4) 598 0.43 0.68 284 0.77 1.9Significant Parent Size 598 8.19 1.98 284 8.38 1.89Not Significant Segment Size 598 0.53 0.39 284 0.57 0.39 significant
Table 2.1: Summary statistics
41 Table 2.2 (Continued)
Table 2.2: Correlation Matrix 42 Table 2.2 (Continued)
Table 2.2 (Continued) 43
Table 2.2 (Continued)
44 Divesting firm=1 Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8
Uncertainty in business unit's environment -7.112** -6.910** -9.289** -8.972** -7.291** -7.346** -8.543** -8.391** [2.123] [2.075] [2.292] [2.216] [2.136] [2.165] [2.067] [2.034] Parent performance -2.354*** -2.518*** -2.339*** -2.499*** -2.355*** -2.556*** -2.331*** -2.537*** [3.679] [3.972] [3.659] [3.947] [3.147] [3.458] [3.110] [3.427]
Segment performance 0.242 0.257 0.211 0.226 0.014 0.022 -0.014 -0.002 [0.619] [0.656] [0.540] [0.576] [0.030] [0.049] [0.031] [0.005] Institutional blockholder share 0.004 0.005 0.004 0.004 0.001 0.001 0.001 0.001 [1.302] [1.371] [1.223] [1.297] [0.188] [0.249] [0.151] [0.219]
Behavioral uncertainty 7.747 7.185 4.712 3.828 [1.209] [1.123] [0.714] [0.582] Business unit's industry information asymmetry -0.06 -0.053 -0.062 -0.055 -0.053* -0.049* -0.054* -0.049* [1.326] [1.597] [1.235] [1.497] [1.701] [1.951] [1.646] [1.909] Parent valuation (Sales measure) -0.075** -0.085** -0.077** -0.087** [2.014] [2.427] [2.070] [2.482] Parent valuation (market measure) 0.013*** 0.013*** 0.013*** 0.013*** [2.692] [2.701] [2.692] [2.702] Level of Diversification (entropy measure) 0.668*** 0.635** 0.686*** 0.653** 0.758** 0.716** 0.770** 0.726** [2.621] [2.495] [2.684] [2.558] [2.484] [2.336] [2.521] [2.367] Business unit's relatedness -0.545*** -0.547*** -1.003** -0.972** -0.503*** -0.499*** -0.785* -0.727* [4.652] [4.682] [2.486] [2.423] [3.795] [3.766] [1.860] [1.743]
Leverage (lev1) 0.709** 0.695** 0.734** 0.729** [2.282] [2.231] [2.101] [2.087] Leverage (lev4) 0.15 0.148 0.204** 0.205** [1.632] [1.609] [2.060] [2.075]
Firm size 0.106*** 0.103*** 0.104*** 0.101*** 0.082** 0.081** 0.081** 0.080** [3.473] [3.356] [3.423] [3.312] [2.386] [2.363] [2.346] [2.331]
Segment size 0.277** 0.273** 0.279** 0.275** 0.222 0.207 0.223 0.208 [2.051] [2.027] [2.059] [2.036] [1.446] [1.357] [1.452] [1.361] Growth options in other businesses -0.003 -0.003 -0.003 -0.003 [1.201] [1.225] [1.182] [1.208] Constant -0.749** -0.782** -0.608 -0.648* -0.588 -0.606 -0.508 -0.538 [2.063] [2.166] [1.553] [1.657] [1.538] [1.584] [1.237] [1.307] Observations 882 882 882 882 706 706 706 706 Prob>chi-sqaure 0 0 0 0 0 0 0 0 Wald Chi-square 76.72 78.33 77.75 79.22 62.97 63.03 63.23 63.15 LogLik -512.442 -511.214 -511.903 -510.739 -420.776 -420.746 -420.602 -420.628 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 2.3: Binary probit results testing the relationship between uncertainty and the decision to divest
45 Divesting firm=1 Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8
Uncertainty in segment’s environment -2.61** -2.56** -3.42** -3.33** -2.74** -2.79** -3.21** -3.19** [2.12] [2.06] [2.27] [2.22] [2.12] [2.15] [2.05] [2.02]
Parent performance -0.86*** -0.94*** -0.86*** -0.93*** -0.89*** -0.97*** -0.88*** -0.96*** [3.62] [3.94] [3.60] [3.92] [3.13] [3.46] [3.10] [3.42] Segment performance 0.09 0.09 0.07 0.08 0.01 0.01 -0.01 0 [0.62] [0.66] [0.54] [0.58] [0.03] [0.05] [0.03] [0.01]
Institutional block- holder share 0 0 0 0 0 0 0 0 [1.30] [1.37] [1.23] [1.30] [0.19] [0.25] [0.15] [0.22] Behavioral uncertainty 2.85 2.66 1.77 1.45 [1.23] [1.30] [0.71] [0.58]
Business unit’s industry information asymmetry -0.02 -0.02 -0.02 -0.02 -0.02* -0.02* -0.02* -0.02* [1.33] [1.60] [1.24] [1.50] [1.69] [1.94] [1.65] [1.91] Parent valuation (Sales measure) -0.03** -0.03** -0.03** -0.03** [1.97] [2.38] [2.02] [2.44] Parent valuation (market measure) 0.01*** 0.01*** 0.01*** 0.01*** [2.69] [2.70] [2.69] [2.70] Parent's Level of Diversification (Related entropy) 0.25*** 0.24** 0.25*** 0.24** 0.29** 0.27** 0.29** 0.28** [2.67] [2.54] [2.74] [2.60] [2.53] [2.37] [2.56] [2.40] Business unit's relatedness -0.20*** -0.20*** -0.37** -0.36** -0.19*** -0.19*** -0.30* -0.28* [4.33] [4.39] [2.44] [2.38] [3.63] [3.62] [1.840] [1.73] Parent Debt (lev1) 0.26** 0.26** 0.28** 0.28** [2.30] [2.25] [2.10] [2.09] Parent Debt (lev4) 0.06* 0.05 0.08** 0.08** [1.66] [1.64] [2.08] [2.01] Firm size 0.04*** 0.04*** 0.04*** 0.04*** 0.03** 0.03** 0.03** 0.03** [3.47] [3.36] [3.42] [3.31] [2.38] [2.36] [2.34] [2.33] Segment size 0.10** 0.10** 0.10** 0.10** 0.08 0.08 0.08 0.08 [2.09] [2.03] [2.1] [2.07] [1.47] [1.37] [1.47] [1.38] Growth options in other segments 0 0 0 0 [1.19] [1.22] [1.17] [1.21] Observations 882 882 882 882 706 706 706 706 Log Likelihood -512.44 -511.21 -511.9 -510.74 -420.78 -420.75 -420.6 -420.63 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 2.4: Marginal effects for testing the relationship between uncertainty and the decision to divest 46 Divesting firm=1 Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8
Uncertainty in business unit's environment -9.811** -9.461** -13.147** -12.644** -11.944** -11.775** -14.433** -14.025** [2.023] [1.98] [2.303] [2.225] [2.416] [2.401] [2.458] [2.388] Parent performance -1.034 -1.384 -0.966 -1.225 -1.412 -1.722* -1.347 -1.665* [1.114] [1.50] [1.049] [1.341] [1.405] [1.729] [1.346] [1.678]
Segment performance -0.159 -0.102 -0.235 -0.184 -0.584 -0.506 -0.662 -0.578 [0.275] [0.183] [0.412] [0.321] [0.919] [0.802] [1.046] [0.916] Institutional blockholder share 0.001 0.002 0.001 0.001 -0.001 0 -0.001 0 [0.232] [0.43] [0.110] [0.252] [0.125] [0.023] [0.200] [0.086]
Behavioral uncertainty 12.404 11.762 9.597 8.228 [1.443] [1.378] [1.084] [0.942] Business unit's industry information asymmetry -0.02 0.032 -0.033 0.014 -0.102 -0.028 -0.111 -0.037 [0.096] [0.16] [0.160] [0.068] [0.499] [0.144] [0.549] [0.188] Parent valuation (Sales measure) -0.079 -0.083** -0.082 -0.104* [1.375] [2.21] [1.426] [1.913] Parent valuation (market measure) 0.021*** 0.020*** 0.021*** 0.021*** [3.036] [2.991] [3.062] [3.019] Level of Diversification (entropy measure) 0.961** 0.893** 1.005*** 0.940** 0.869** 0.818** 0.900** 0.845** [2.516] [2.333] [2.632] [2.455] [2.202] [2.060] [2.290] [2.140] Business unit's relatedness -0.505*** -0.506*** -1.290** -1.251** -0.459** -0.450** -1.059* -0.965 [3.214] [3.25] [2.161] [2.129] [2.467] [2.443] [1.711] [1.606]
Leverage (lev1) 0.772* 0.732 0.799* 0.787* [1.74] [1.633] [1.702] [1.680]
Leverage (lev4) 0.222* 0.219* 0.314** 0.318** [1.699] [1.690] [2.392] [2.421]
Firm size 0.04 0.04 0.038 0.039 0.075 0.074 0.074 0.073 [0.874] [0.85] [0.816] [0.839] [1.514] [1.492] [1.487] [1.472]
Segment size 0.223 0.222 0.232 0.229 -0.031 -0.047 -0.026 -0.043 [1.130] [1.13] [1.170] [1.158] [0.143] [0.218] [0.117] [0.198] Growth options in other businesses -0.030*** -0.029*** -0.030*** -0.029*** [3.685] [3.501] [3.725] [3.539] Constant -0.053 -0.09 0.173 0.123 -0.102 -0.108 0.055 0.037 [0.098] [0.168] [0.294] [0.209] [0.183] [0.194] [0.092] [0.061] Observations 405 405 405 405 393 393 393 393 Prob>chi-sqaure 0.0004 0.0005 0.0005 0.0008 0 0 0 0 Wald Chi-square 33.55 33.33 34.54 33.49 52.66 47.73 53.65 48.66 LogLik -247.651 -247.393 -246.883 -246.958 -224.699 -226.061 -224.306 -225.76 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 2.5: Probit results for sub-sample A (firms that experienced an increase in uncertainty from ‘t-1’ to ‘t’)
47 Divesting firm=1 Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8 Uncertainty in business unit’s environment -3.74** -3.65** -5.03** -4.89** -4.69** -4.65** -5.68** -5.55** [2.01] [1.96] [2.28] [2.20] [2.40] [2.39] [2.44] [2.37] Parent performance -0.39 -0.53 -0.37 -0.47 -0.55 -0.68* -0.53 -0.66* [1.11] [1.49] [1.04] [1.34] [1.40] [1.72] [1.34] [1.67]
Segment performance -0.61 -0.04 -0.09 -0.07 -0.23 -0.2 -0.26 -0.23 [0.28] [0.18] [0.41] [0.32] [0.92] [0.80] [1.05] [0.92] Institutional block- holder share 0 0 0 0 0 0 0 0 [0.23] [0.43] [0.110] [0.25] [0.12] [0.02] [0.20] [0.09] Behavioral uncertainty 4.75 4.55 3.77 3.25 [1.44] [1.37] [1.08] [0.94]
Business unit’s industry information asymmetry -0.01 0.01 -0.01 0.01 -0.04 -0.01 -0.04 -0.01 [0.10] [0.16] [0.160] [0.07] [0.50] [0.14] [0.55] [0.19] Parent valuation (sales measure) -0.03 -0.03** -0.03 -0.04* [1.35] [2.17] [1.40] [1.88] Parent valuation (market measure) 0.01*** 0.01*** 0.01*** 0.01*** [3.02] [2.98] [3.05] [3.02]
Parent's Level of Diversification (Related entropy) 0.37** 0.34** 0.38*** 0.36** 0.34** 0.32** 0.35** 0.33** [2.57] [2.37] [2.69] [2.50] [2.22] [2.08] [2.31] [2.16] Business unit's relatedness -0.19*** -0.20*** -0.49** -0.48** -0.18** -0.18** -0.42* -0.38 [3.07] [3.12] [2.13] [2.10] [2.41] [2.40] [1.70] [1.60]
Parent Debt (lev1) 0.30* 0.28 0.32* 0.31* [1.74] [1.64] [1.70] [1.47] Parent Debt (lev4) 0.08* 0.08* 0.12** 0.13** [1.73] [1.720] [2.41] [2.44] Firm size 0.02 0.02 0.01 0.01 0.03 0.03 0.03 0.03 [0.87] [0.85] [0.82] [0.84] [1.51] [1.49] [1.48] [1.47] Segment size 0.09 0.09 0.09 0.09 -0.01 -0.02 -0.01 -0.02 [1.14] [1.14] [1.180] [1.16] [0.14] [0.22] [0.12] [0.20] Growth options in other businesses -0.01*** -0.01*** -0.01*** -0.01*** [3.60] [3.44] [3.64] [3.48] Observations 405 405 405 405 393 393 393 393 Log Likelihood -247.65 -247.39 -246.88 -246.96 -224.7 -226.06 -224.31 -225.76 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 2.6: Marginal effects for sub-sample A (firms that experienced an increase in uncertainty from ‘t-1’ to ‘t’) 48 Divesting firm=1 Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8
Uncertainty in business unit's environment -3.497 -2.739 -3.984 -2.953 -1.852 -1.598 -2.069 -1.397 [0.655] [0.512] [0.607] [0.447] [0.348] [0.301] [0.312] [0.211] Parent performance -3.500*** -3.554*** -3.499*** -3.554*** -3.560*** -3.708*** -3.558*** -3.709*** [3.133] [3.245] [3.130] [3.242] [2.959] [3.106] [2.951] [3.100] Segment performance 0.497 0.523 0.49 0.52 0.482 0.475 0.479 0.478 [0.792] [0.825] [0.771] [0.807] [0.717] [0.702] [0.702] [0.694] Institutional blockholder share -0.002 -0.002 -0.002 -0.002 -0.002 -0.002 -0.002 -0.002 [0.326] [0.372] [0.329] [0.372] [0.359] [0.399] [0.360] [0.396] Behavioral uncertainty 1.655 0.71 0.738 -0.672 [0.143] [0.061] [0.062] [0.056] Business unit's industry information -0.042*** -0.041*** -0.042*** -0.041*** -0.044*** -0.043*** -0.044*** -0.043*** [3.619] [3.629] [3.622] [3.632] [3.689] [3.727] [3.693] [3.731] Parent valuation (Sales measure) -0.101* -0.092* -0.101* -0.093* [1.751] [1.743] [1.756] [1.748] Parent valuation (market measure) 0.005 0.004 0.005 0.004 [0.631] [0.585] [0.630] [0.585] Level of Diversification 0.287 0.259 0.287 0.259 0.254 0.198 0.255 0.197 [0.524] [0.472] [0.525] [0.472] [0.457] [0.352] [0.458] [0.351] Business unit's relatedness -0.498** -0.500** -0.59 -0.539 -0.494** -0.493** -0.535 -0.456 [2.542] [2.551] [0.915] [0.836] [2.434] [2.408] [0.801] [0.677] Leverage (lev1) 0.707 0.706 0.904 0.905 [1.261] [1.255] [1.644] [1.643] Leverage (lev4) 0.043 0.043 0.166 0.166 [0.250] [0.251] [1.038] [1.042] Firm size 0.107** 0.104** 0.107** 0.104** 0.120** 0.116** 0.120** 0.116** [2.141] [2.073] [2.136] [2.072] [2.350] [2.273] [2.348] [2.278] Segment size 0.323 0.329 0.321 0.328 0.349 0.339 0.348 0.34 Growth options in [1.440] [1.478] [1.437] [1.481] [1.488] [1.462] [1.489] [1.470] other businesses 0.003 0.003 0.003 0.003 [0.874] [0.816] [0.875] [0.813] Constant -0.716 -0.859 -0.685 -0.845 -1.117* -1.176** -1.104* -1.189* [1.167] [1.415] [1.061] [1.319] [1.893] [1.998] [1.758] [1.894] Observations 322 322 322 322 308 308 308 308 Prob>chi-sqaure 0 0 0 0 0 0 0.0001 0.0001 Wald Chi-square 40.83 42.53 41.12 42.75 40.29 41.61 40.37 41.58 LogLik -186.245 -185.479 -186.238 -185.477 -180.243 -179.46 -180.241 -179.459 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 2.7: Probit results for sub-sample B (firms that experienced a decrease in uncertainty from ‘t-1’ to ‘t’) 49 Divesting firm=1 Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8
Uncertainty in business unit’s environment -1.3 -1.01 -1.48 -1.09 -0.67 -0.58 -0.75 -0.51 [0.65] [0.51] [0.60] [0.45] [0.35] [0.30] [0.312] [0.21]
Parent performance -1.30*** -1.32*** -1.30*** -1.32*** -1.29*** -1.35*** -1.29*** -1.35*** [3.06] [3.22] [3.06] [3.22] [2.94] [3.11] [2.93] [3.11] Segment performance 0.18 0.19 0.18 0.19 0.17 0.17 0.17 0.17 [0.79] [0.82] [0.77] [0.80] [0.72] [0.70] [0.70] [0.69] Institutional block- holder share -0.01 0 0 0 0 0 0 0 [0.33] [0.37] [0.33] [0.37] [0.36] [0.40] [0.36] [0.40] Behavioral uncertainty 0.61 0.26 0.27 -0.24 [0.14] [0.06] [0.06] [0.06]
Business unit’s industry information asymmetry -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** [3.54] [3.55] [3.55] [3.55] [3.60] [3.63] [3.60] [3.63] Parent valuation(sales measure) -0.04* -0.03* -0.04* -0.03* (sales measure) [1.69] [1.70] [1.70] [1.71] Parent valuation (market measure) 0 0 0 0 [0.59] [0.59] [0.63] [0.59] Parent's Level of Diversification (Related entropy) 0.11 0.1 0.11 0.1 0.07 0.07 0.09 0.07 [0.53] [0.48] [0.53] [0.48] [0.35] [0.352] [0.46] [0.35] Business unit's relatedness -0.19** -0.19** -0.22 -0.2 -0.18** -0.18** -0.19 -0.17 [2.42] [2.43] [0.91] [0.83] [2.29] [2.41] [0.80] [0.69] Parent Debt (lev1) 0.26 0.26 0.33* 0.33* [1.27] [1.26] [1.65] [1.65] Parent Debt (lev4) 0.02 0.02 0.06 0.06 [0.250] [0.25] [1.05] [1.06] Firm size 0.04** 0.04** 0.04** 0.04** 0.04** 0.04** 0.04** 0.04** [2.16] [2.09] [2.15] [2.09] [2.37] [2.29] [2.37] [2.30] Segment size 0.12 0.12 0.12 0.12 0.13 0.12 0.13 0.12 [1.47] [1.51] [1.47] [1.51] [1.54] [1.50] [1.54] [1.51] Growth options in other businesses 0 0 0 0 [0.88] [0.82] [0.88] [0.82] Observations 322 322 322 322 308 308 308 308 Log Likelihood -186.25 -185.48 -186.24 -185.48 -180.24 -179.46 -180.24 -179.46 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 2.8: Marginal effects for sub-sample B (firms that experienced a decrease in uncertainty from ‘t-1’ to ‘t’)
50 Variable Model 1 Model 2 Model 3 Model 4 G1- Uncertainty in business unit’s environment -10.17 -9.69 -13.77 -13.29 [2.08] [2.01] [2.41] [2.34] G1-Parent performance -0.67 -0.92 -0.63 -0.88 [0.86] [1.19] [0.81] [1.14] G-1Segment performance -0.37 -0.33 -0.41 -0.38 [1.40] [1.25] [1.56] [1.41] G1-Institutional block- holder share 0 0 0 0 [0.19] [0.33] [0.06] [0.19] G1-Behavioral uncertainty 13.5 12.92 [1.58] [1.52] G1-Business unit’s industry information asymmetry -0.01 0.04 -0.02 0.02 [0.04] [0.18] [0.12] [0.10] G1-Parent valuation -0.08 -0.1 -0.08 -0.11 [1.40] [1.89] [1.45] [1.94] G1-Parent's level of diversification (related entropy) 1.00*** 0.94** 1.05*** 0.99*** [2.63] [2.45] [2.75] [2.58] G1-Business unit’s relatedness -0.48 -0.48 -1.34 -1.3 [3.15] [3.18] [2.25] [2.22]
G1-Parent debt (lev1) 0.7 0.67 [1.55] [1.49] G1-Parent debt (lev4) 0.22* 0.21 [1.65] [1.64] G1-Firmsize 0.04 0.04 0.04 0.04 [0.82] [0.84] [0.77] [0.79] G1-Segment size 0.23 0.23 0.24 0.24 [1.16] [1.15] [1.21] [1.20] G1-Increase group -0.05 -0.09 0.19 0.15 [0.1] [0.17] [0.32] [0.25] Table 2.9: Results from Combined Estimation using Dummy Variables Table 2.9 (continued) 51 Table 2.9 (continued) G2- Uncertainty in business unit’s environment -3.71 -2.95 -4.34 -3.32 [0.69] [0.55] [0.66] [0.50] G2-Parent performance -3.34 -3.4 -3.35 -3.4 [3.30] [3.46] [3.31] [3.47] G-2Segment performance 0.27 0.3 0.27 0.3 [0.75] [0.81] [0.74] [0.80] G2-Institutional block- holder share 0 0 0 0 [0.34] [0.39] [0.34] [0.39] G2-Behavioral uncertainty 2.14 1.22 [0.19] [0.11] G2-Business unit’s industry information asymmetry -0.04 -0.04 -0.04 -0.04 [3.62] [3.63] [3.62] [3.63] G2-Parent valuation -0.1 -0.09 -0.1 -0.09 [1.72] [1.69] [1.72] [1.70] G2-Level of Diversification 0.31 0.28 0.31 0.28 [0.57] [0.52] [0.57] [0.52] G2-Business unit’s relatedness 0 0 0.74 0.75 [0.00] [0.00] [0.84] [0.87] G2-Parent debt (lev1) 0.71 0.71 [1.27] [1.26] G2-Parent debt (lev4) 0.04 0.04 [0.24] [0.24] G2-Firmsize 0.11** 0.11** 0.11** 0.10** [2.14] [2.08] [2.14] [2.07] G2-Segment size 0.34 0.34 0.33 0.34 [1.50] [1.54] [1.49] [1.54] G2-Decrease group -0.72 -0.87 -0.68 -0.84 [1.18] [1.43] [1.06] [1.32] Observations 727 727 727 727 LogLik -433.19 -432.57 -432.28 -431.7 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1%
52
Uncertainty level at time ‘t’
Low High
Low 1 2 Uncertainty level at Uncertainty at ‘t’ more time ‘t-1’ relevant 3 4 High Uncertainty at ‘t’ less relevant/irrelevant
Figure 2.1: Relationship between changes in the levels of uncertainty and the impact of uncertainty on the decision to divest due to changes in ‘put’ value
53 CHAPTER 3 UNCERTAINTY, INFORMATION ASYMMETRY AND DIVESTMENT STRATEGIES The first essay shows that firms’ decisions’ to divest business units may be driven by flexibility considerations under conditions of high uncertainty. More specifically, the study shows that high uncertainty in a business unit’s environment reduces the probability of its divestment. Further, the study also shows that firms are less likely to divest when uncertainty increases than when uncertainty decreases. In effect, these results suggest that firms may retain their real ‘put’ options—i.e., keep flexibility— rather than divest under conditions of high uncertainty and that the ‘put’ option value changes with increases and decreases in uncertainty (Dixit and Pindyck, 1995; Myers and
Majd, 1990).
Flexibility considerations may be important for firms not only in decisions regarding ‘whether to divest?’ but also for ‘how to divest?’ i.e., for divestment mode choices. In highly uncertain environments where the true value of a business unit is not known, firms may prefer divestment modes that allow flexibility over future courses of action i.e., keep their ‘options’ open (Perotti and Rossetto, 2007). Some divestment modes are inherently more flexible than others. For example, sell-offs, in which a
54 business unit is sold completely to another company, are costly-to-reverse ‘complete divestments’. On the other hand, spin-offs and equity carve-outs (in which a part of the equity in the unit to be divested is distributed pro-rata to existing shareholders or issued to new shareholders, respectively) are ‘staged/partial divestments’. These forms of divestment give firms’ flexibility over future courses of action—whether to further divest, stop mid-way, or regain control over the business unit—depending on how situations evolve (Zingales, 1995). Thus, staged divestments can be thought of having at least some attributes of real options (Perotti and Rossetto, 2007). Firms may choose them over costly-to-reverse complete divestments, particularly under conditions of high uncertainty.
However, staged divestments may also be preferred because of information asymmetry (Bergh, Johnson, Dewitt, 2008). When there is high information asymmetry—a condition in which insiders of a firm know the true value of the assets/business units better than outsiders (Akerlof, 1970)—communicating the true value of business units’ to outsiders/potential buyers may be very difficult. Staged divestments can help in such situations because they increase transparency about the value of the business unit ((Krishnaswami and Subrahmaniam, 1999). Further, they also help in timing the release of information to potential buyers which itself can be valuable in maximizing the proceeds from sale (Mittendorf, 2004).
These two theoretical perspectives—real options and information asymmetry— have yet to be simultaneously examined in the literature. Partly this is because the second of these two perspectives—information asymmetry—has dominated previous 55 research (Nanda, 1991; Nanda and Narayanan, 1998; Powers, 2003), while the first—real options—has only recently begun to be applied to divestment and related decisions
(Perotti and Rossetto, 2007). The purpose of this paper is to bring these theoretical perspectives together and test whether either one or both of these explain divestment mode choices.
The empirical implications of these two perspectives are examined in a data set that includes complete divestments (sell-offs) and staged divestments (spin-offs and equity carve-outs) using a probit model with selection correction for non-divesting firms.
The results show that information asymmetry increases the probability of staged divestments (or alternatively decreases the probability of complete divestments) whereas uncertainty is not a significant determinant of divestment mode choices.
3.1 Prior literature
3.1.1 Real options
Real options theory originated in finance. Myers’s (1977) seminal idea that firms’ discretionary investment opportunities can be viewed as ‘call’ options on real assets, analogous to financial call options, laid the foundation for this theory.
Options are rights, without obligations, to undertake specific actions at a future date. They can be ‘call options’ which are rights to buy/invest, or ‘put options’ which are rights to sell/divest at a future date. When there are options, one can wait (generally) for uncertainty to resolve rather than commit to costly-to-reverse actions by making investments or divestments under conditions of high uncertainty (Dixit and Pindyck,
56 1995, Merton, 1998). Therefore, options help in benefiting from the upside potential if situations become favorable while preventing losses if situations become unfavorable.
The strategic management literature built on the work in finance. This began with
Kogut’s seminal work on the real option properties of joint ventures, focusing on the design of governance mechanisms for investments under conditions of high uncertainty
(Kogut, 1983; 1985; 1991). Ever since, a growing body of research in real options has focused on the impact of uncertainty on investment decisions (Campa, 1993) and on flexible governance mechanisms for investments under conditions of high uncertainty
(Cuypers and Martin, 2007; Folta and Miller, 2002). Further, research attention was also focused on the different types of options that arise in the investment contexts such as growth options (Kogut, 1983; McGrath, 1997), deferral options (Guiso and Parigi, 1999), switching options (Kogut, 1983, 1985) etc., and the nature of interactions between these various options (eg., the growth versus deferral options debate (Folta and O’Brien, 2004;
Kulatilaka and Perotti, 1998; Leiblein and Zeidonis, 2007)).
Analyzing divestments as ‘put’ options also attracted some research attention.
Several arguments have been made for delays in costly-to-reverse exit decisions under conditions of high uncertainty, because of the ‘put’ option value (Dixit, 1992; Dixit and
Pindyck, 1995). There has also been some empirical testing of these ideas. A prominent empirical test infers the value of a firm’s abandonment options from its balance sheet and shows that such options have a positive effect on firm value (Berger, Ofek and Swary,
1996). Other evidence comes from experimental studies by Bragger et al. (1998) who find that in a simulated economy, participants who received feedback with higher 57 variability delayed exit decisions longer and invested more often than participants who received feedback with lower variability. Vassolo, Anand and Folta (2004) test and support exit delays under conditions of uncertainty in the context of bio-tech alliances.
The first essay extended the real ‘put’ option arguments to the study of the decision to divest a business unit. Prior work on divestments attributed delays in the decision to divest to managerial vested interests or other agency problems (Boot, 1992;
Cho and Cohen, 1997; Shimizu and Hitt, 2005). In the first essay, it has been found that business unit divestments may, indeed, be driven by option considerations when the conditions surrounding divestments are uncertain. Therefore, managerial inaction with regard to divesting a business, under conditions of high uncertainty, may be consistent with shareholder interests.
Real options theory may also be useful in understanding firms’ divestment mode choices. Very recently, Perotti and Rossetto (2007) developed a model that suggests that equity carve-outs, unlike sell-offs, are like real options because they give flexibility to decide on future course of action as uncertainty unfolds.28 However, the empirical implications of real options theory are yet to be explored fully in the context of divestment mode choices.
28 Perotti and Rosetto (2007) do not distinguish between uncertainty about the future value of a business unit, which is in the domain of real option theory, and the impact of noise about the value of synergies between a parent and its subsidiaries on valuing the business unit, which is in the domain of information asymmetry. Also many of their arguments would apply as much to spin-offs, which are also staged divestments. 58 3.1.2 Information Asymmetry
Myers and Majluf (1984) were the first to apply the concept of information asymmetry to corporate financing decisions. According to Myers and Majluf, in the presence of information asymmetry, managers of firms acting in the interests of existing shareholders will be willing to forego even positive NPV projects, than issue equity, when firms have undervalued assets. Therefore, new equity issues are more likely when investments are not very promising. The market anticipates this and discounts firms’ new issues. Managers anticipate this market discounting and may be even more unwilling to issue new equity. Thus, information asymmetry effectively discourages new equity issues. Issuance of equity is akin to divestment. By issuing equity, entrepreneurs or promoters of the company divest themselves of a portion of ownership to outsiders.
Therefore these arguments made in the context of equity issuances also apply to decisions to divest.
Implications of information asymmetry have been explored quite extensively and tested in the context of divestment mode choices— i.e., between complete divestments and staged divestments such as equity carve-outs and spin-offs—than divestments per se.
Nanda (1991) extended Myers and Majluf’s framework and argued that the seemingly contradictory positive stock market reactions associated with equity carve-outs are an indication that parent company assets are undervalued. Nanda and Narayanan (1998) suggested that spin-offs are cost effective means of divesting business units under conditions of high information asymmetry and that spin-offs facilitate more accurate valuation of the business units by markets. Krishnaswami and Subramaniam (1999) 59 showed empirically that, indeed, spin-offs are chosen by high information asymmetry firms and that information asymmetry decreases post-spin-off. Powers (2003) argued that equity carve-outs may be preferred by firms with overvalued assets.
More recently, Bergh et al., (2008) used the asymmetric information argument to explain the choice between sell-offs and spin-offs. They argue that firms divesting assets in their primary business lines or firms with low levels of diversification are likely to have higher information asymmetry problems between insiders and outsiders about the true value of the business unit. Therefore, such firms may prefer spin-offs to mitigate information asymmetry. On the other hand, firms divesting assets in secondary businesses or firms with higher levels of diversification are likely to choose sell-offs since the value of the businesses is relatively transparent to market. They show empirically that these divestment mode choices mediate the relationship between corporate restructuring activities and financial performance.
None of these previous studies—either from a real options or information asymmetry perspective—examine the choice between ‘staged’ versus ‘complete’ divestment using these two points of view simultaneously. This is the case even though these two arguments are very different in flavor. The real options argument becomes relevant when there is uncertainty about the future value of a business unit. The information asymmetry argument becomes relevant when there is a ‘true’ value for the business unit that insiders know better than outsiders—which implies that there is a reasonable degree of certainty about the value of the business unit.
60 3.2 Theory and hypotheses
3.2.1 Staged divestments: The options perspective
Firms may structure their divestments in order to proactively deal with uncertainty much as they structure their investments under conditions of uncertainty (Kogut, 1991).
As mentioned earlier, a key difference between sell-offs and spin-offs/equity carve-outs is that unlike in a sell-off where a business unit/subsidiary is completely lost after divestment to another firm, by resorting to a spin-off/carve-out, parent firms gain flexibility over future events— i.e., whether to further spin-off, carve-out, sell-off or roll- back business units into parents by repurchasing the shares issued (Zingales, 1995). Such flexibility can be valuable when the conditions surrounding divestments are uncertain.
Uncertainty about the prospects of the business unit’s industry and the likely future synergies between the parent and the business unit, make costly-to-reverse actions
(such as sell-offs) unattractive (Perotti and Rossetto, 2007). This is because the full value of the business unit may not be realized under conditions of uncertainty. Also, new technological breakthroughs may bring in unanticipated synergies, legal or regulatory regime changes may lead to more favorable environment for certain businesses etc. By choosing staged divestments (spin-offs/carve-outs), as opposed to complete divestments
(sell-offs), parent firms can maintain their access to critical information in the subsidiary industries’ and can adjust their divestment strategy as new information arrives. If situations turn better than expected, the business unit can be further divested at a better price, or brought back into the parent’s fold more easily in case of a partial divestment
61 than in a complete sale. On the other hand, if situations turn worse, again the parent firm will have the flexibility to choose a course of action that best fits its requirements.
Viewed this way, performing a spin-off/carve-out, as opposed to a complete sale, is like creating future real options. The value of such options increases with increasing uncertainty in a business unit’s environment. Therefore,
Hypothesis 1: Higher uncertainty in a business unit’s environment will be
positively associated with staged divestment.
3.2.2 Staged divestments: The information asymmetry perspective
As outlined earlier, information asymmetry has been a dominant explanation for staged divestments (Krishnaswami and Subrahmaniam, 1999). There are at least three arguments. The first one is about taking advantage of imperfections caused because of information asymmetry. As firms grow larger and more diverse, the value of the parent firms and the business units within them becomes less transparent to outsiders. This lack of transparency could lead to undervaluation or overvaluation of business units that are part of large parent companies. Managers may take advantage of such imperfections by engaging in equity carve-outs 29 when the subsidiary shares are overvalued (Myers and
Majluf, 1984; Nanda, 1991; Powers, 2003) or spin-offs when undervalued
(Krishnaswami and Subrahmaniam, 1999). Thus, staged divestments are likely under
29 It is unclear why arguments have not been made for 100 per cent issuance of equity to take full advantage of the overvaluation!!! 62 conditions of information asymmetry, whether it is due to overvaluation or undervaluation. In this explanation, the issue is managers taking advantage of a window of opportunity, not about flexibility.
A second argument about information asymmetry concerns maximizing the proceeds from sale. As mentioned in the previous paragraph, when business units are part of larger parent conglomerates, it is very possible that they are not valued appropriately (Berger and Ofek, 1995; Lang and Stulz, 1994). This is because external markets may not be able to understand the linkages between various business units within a parent firm. Sometimes, this lack of transparency could be purposeful to protect competitive advantages (James and Shaver, 2008; Ozbilgin and Penno, 2005; Yosha,
1995). In either case, this could consequently lead to over or undervaluation of the business units.30 Insiders of firms may know the true value of the business units.
However, this value may be difficult to be communicated to the external market.
Directly selling business units to third parties may not be ideal in such circumstances because adverse selection conditions hold (Akerlof, 1970). Therefore, firms may choose to spin-off/carve-out business units to mitigate the adverse selection costs, let the market gauge the true value of business units, and then depending on the market reaction choose to proceed with the divestment.31
30 While it is tempting to think that overvaluation may not be a problem when the goal is to sell, the potential legal issues and subsequent loss of trust cannot be underestimated. 31 In some sense, carve-outs and sell-offs can be considered more similar than different because they involve selling to external shareholders or third parties. However, carve-outs ensure flexibility that is missing in sell-offs. 63 Further, a third argument is about judicious timing of information release in the divestment process in the presence of information asymmetry. Sellers may find releasing the information costly because it puts them and buyers on level informational footing.
Therefore, there is a cost-benefit trade-off involved in releasing information via staged divestments to ensure that bargaining power is not lost vis-à-vis buyers. This trade-off is likely to be stronger when the level of information asymmetry about the true value of the business unit is greater (Mittendorf, 2004 32 ). Therefore,
Hypothesis 2: Higher information asymmetry about the value of a business unit
will be positively associated with staged divestment.
3.3 Method
3.3.1 Rationale for the choice of methodology
Divestment mode choices may be better understood in conjunction with the decision to divest. One way to model this would be to include divestment mode choices and the alternative of non-divestment in a multinomial model (Chen and Guo, 2005).
However, non-divestment may be an alternative to divestment, rather than an alternative to divestment mode choices. Further, the decision to divest and the decision to divest through a particular mode may be driven by a set of overlapping, yet different, factors
(Chang and Singh, 1999). The multinomial model would not work because it treats all alternatives as being driven by the same set of factors.
32 Again, Mittendorf (2004) uses an options reasoning. However, options reasoning would be applicable in case of uncertainty about the future of the business. Problems that do not involve such uncertainty are better approached from other theoretical perspectives. 64 Modeling using two separate probits/logits allows the decision to divest and the divestment mode choices to be driven by different sets of factors. However, such modeling does not correct for possible self-selection mechanisms operating between divesting and non-divesting firms. This is important because divestment mode choices are observed only for firms that engage in divestments. The selection effects between divesting and non-divesting firms could potentially influence the divestment mode choice decisions. Therefore, a model that accommodates the possible dependence between the decisions to divest and divestment mode choices and makes necessary corrections is more appropriate.
Accordingly, a probit model with selection correction has been chosen (Dubin and
Rivers, 1989). This method is similar to the Heckman method but is different in that both the selection and the outcome decisions involve dichotomous variables. There is a probit model in the selection equation (the decision to divest versus not divest) and another probit model in the outcome equation (the decision to engage in staged divestment versus complete divestment). The ‘heckprob’ command in STATA jointly maximizes the likelihood of the decision to divest or not divest and the divestment mode choice decision. This method accommodates the possibility that these two decisions are not independent.
The observations in a sample with these selection issues would have the following probabilities. If y1 refers to the decision to divest (y1 =1) or not divest (y1 =0) and y2 refers to the staged divestment i.e., spin-off/carve-out (y2=1) or complete divestment i.e., sell-off (y2=0), then 65 Pr(y1 = 0) = Φ(-x 1β1) ------ (1) for non-divesting firms Pr(y1 = 1, y2 = 0)= Φ(-x 1β1)- Φ2(x 1β1, x 2β2, ρ) ------ (2) for divesting firms that choose complete divestment via sell-off Pr(y1 = 1, y2 = 1)= Φ2(x 1β1, x 2β2, ρ) ------ (3) for divesting firms that choose staged divestments via spin-off/carve-out The joint likelihood function can be written as
For i=1 to N,
lnL = Σ { yi1 yi2 ln Φ2(x 1β1, x 2β2, ρ) + yi1 (1- yi2) ln [ Φ(x 1β1) - Φ2(x 1β1, x 2β2,
ρ)]+ (1- yi1) ln ( Φ(-x 1β1)} where ρ=0 indicates that the two decisions are independent. x 1 , x 2 are the vectors of independent variables with some overlap, and β1 and β2 are the vectors of coefficients for the selection and outcome equations respectively.
3.3.2 Data
All forms of divestments—sell-offs, spin-offs, equity carve-outs were required for this analysis. Preliminary data were obtained from Security Data Corporation’s (SDC)
Mergers and Acquisitions and New Issues databases. The sample was obtained after cleaning the initial list. The data filters used (See Appendix) are those that are standard in the finance literature on divestitures (Chen and Guo, 2005; Powers, 2003). Sell-offs were identified from dropped segments in a company’s portfolio matched against announcements in the SDC database (Schlingemann, Stulz and Walkling, 2002). All deals were verified by searching through SEC filings, Factiva news articles/newswires, and on Lexis-Nexis. The sample period was from January 1980 to December 2003.
66 Only completed deals were included in the sample because intentions that do not materialize later on cannot be considered as decisions. Deals with no match on
Compustat were also deleted from the list. This was to ensure availability of adequate accounting information for analysis. Also, since this analysis required data at the segment level, only those companies with corresponding matches in the Compustat segment database were included.
Only the first decision on a particular segment was taken into account (i.e., if a segment was found to be both in the spin-off/carve-out and sell-offs samples, the first of the deals was taken into account to avoid any confounding factors affecting the follow-up decision). Also, only parent firms which are in the manufacturing and related categories
(as indicated in the Robins and Wiersema (1995) grouping), other than utilities (SIC codes 4000-4999) were included.33 Utilities are in the regulated industry category, and
33 A primary SIC code is assigned by Standard & Poor’s to each company in the COMPUSTAT databases according to its primary business activity (as determined by revenues). Therefore, a company’s primary SIC code may change depending on the change in the proportion of sales contributed by particular segments. When this change is affected retrospectively, it is possible that there could be differences between what has been the primary SIC code of the firm in a particular year and the one that has been assigned retrospectively. Segments, on the other hand, are identified based on the primary and secondary products and retain their SIC codes. Segments SIC codes could change when they are merged with other segments and reorganized. However, there is no evidence that these changes in SIC codes are retrospective. Therefore, segment SIC codes are more reliable. Impact on the parent’s portfolio relatedness measure: There is NO impact of a possibly different historic SIC code for the parent firm on the calculation of the parent’s portfolio relatedness measure and the segment’s relatedness measure. This because parent SIC code does not enter the calculation of these measures at all. There are two measures of parent’s portfolio relatedness that have been calculated. Their calculations are detailed below: 1. The Robins and Wiersema (1995) measure of firm’s portfolio resource based relatedness: According to Robins and Wiersema, for each combination of two different industry categories ‘i’ and ‘j’ in a firm’s portfolio, the sales-weighted measure of interrelationship Rij is given by R ij =P i r ij + P j r ij , where Pi= percentage of sales in industry category i and Pj= percentage of sales in industry category j. These weighted measures of similarity between paired industries (Rij) are summed over all combinations of two industries that could be formed in a business portfolio of the firm resulting in an aggregate index of interrelationship of the businesses of the firm Mk= Σ Rij= Σ rij(Pki+ Pkj), where i and j represent any two 67 therefore it would not be appropriate to treat them along with other industrial groupings.
Only domestic parent firms were included to eliminate influences of differing institutional regimes. American Depository Receipts (ADRs), were excluded for a similar reason. Limited partnerships that operate under different legal rules were also excluded.
Spin-off and carve-out divisions fulfilled following additional criteria: 1. spin-off or carved out company was not organized as a limited partnership (to ensure that mere reorganization to be under different legal rules was not the objective), 2. listed on the
CRSP (were indeed being traded as independent entities) 3. were not tracking stock deals
(since tracking stocks do not involve any separation of ownership of the unit from the parent and are issued mainly to ensure that the accounting performance of a particular different industries in which the firm k is active. The authors then introduce a correction factor to avoid double counting. This is the parent’s total portfolio interrelatedness. Segment’s interrelatedness is measured as the sum of the segment’s relatedness with other segments in the parent’s portfolio. 2. The measure of related entropy in a parent’s portfolio that reflects the level of diversification. The basic entropy index was computed by Jacquemin and Berry (1979) as E= ΣPi* ln(1/Pi). In this equation, E signifies the entropy measure and Pi is the proportion of a firm’s sale in SIC industry i. This is typically treated as a measure of total diversification when calculated at the 2-digit SIC level. The unrelated diversification is computed using the proportions at the 2-digit SIC level and subtracted from the 4-digit level to obtain a measure of ‘related’ diversification in the portfolio. As can be seen, these calculations involve the segment level SIC code information and not the parent SIC code. Therefore, changes in the historic SIC codes of a parent firm has no impact on the computation of these measures. Other measures such as uncertainty and information asymmetry are calculated at the segment industry relatedness category level. The industry relatedness category that a firm belongs to is determined by its SIC code. However, the uncertainty and information asymmetry measures are averages of all firms in the industry categories. Therefore, if firms do not get included in a category they belong to, it is also possible that some firms get included where they do not belong to and, on average, the differences may get compensated. The effect will be expected to be similar with regard to identifying the control firms that have been chosen based on the parent’s SIC code. Therefore, the potential source of error due to differences in the historic SIC codes is not expected to be of major concern. In any case, historic SIC codes have been verified for about 15 per cent of the sample. In the sub- sample, the industry relatedness category was different only for about 13 per cent of the firms. The movements from one relatedness category to another were not systematic. Therefore, it is not a major concern. 68 unit can be monitored more closely while it is still within a parent conglomerate) 4. were incorporated in the United States (to eliminate any institutional advantages from separation), 6. were not owned by multiple parents, 7. were not a direct result of mergers
(to prevent other confounding factors).
A sample of matched firms that have not engaged in a divestiture was also obtained. Matched parents were found using the parent’s total assets as a matching criterion within the same relationship category matched at the same four-digit SIC code
(Chen and Guo, 2005; Krishnaswami and Subrahmaniam, 1999).34 The time period used for selecting the matched sample is the same as that for the divesting firms.35 There were some cases where there was no match found but still the original divesting firm was retained in the sample. Together, the final sample had 182 sell-offs, 102 spin-offs and carve-outs, and 598 non-divested firm segments with no missing values for any of the required variables.36
The data collection procedure clearly restricts the analysis to large publicly traded firms with listed segments. This leaves out several small firms not listed in the databases used. However, inclusion of small firms would only strengthen the results because small firms are susceptible to the influences of uncertainty more than larger firms. This study
34 Parent firm size and industry grouping at four digit level have been used as the matching criterion following common practice (Chen and Guo, 2005; Krishnaswami and Subramaniam,1999). Since parents belong to manufacturing and related industries, size measured by parent assets is more appropriate. Both the first best match and the next best matched parent firms that have not engaged in a divestiture were chosen in order to ensure a close to 2:1 ratio of matched versus divesting firms. 35 Matched firms were not necessarily chosen from the same year as the divesting firm, though matching firms in the same year were not eliminated by design. 36 Reasonable efforts have been made to recover missing values from 10-ks. 69 is therefore a conservative test of the hypotheses. Also, large private firms are not in this list, but such firms are relatively uncommon, at least in the U.S.
3.3.3 Other data sources
The data to calculate the uncertainty measure were obtained from the Center for
Research in Security Prices (CRSP) monthly stock price database. Data for calculating the information asymmetry measure were obtained from the summary statistics on the
Institutional Brokers Estimates System (IBES) database. Block-holder data was obtained from Thomson Financial’s Spectrum Institutional Stockholding database.
3.3.4 Variables and measures for the selection equation
Dependent variable
The dependent variable in the selection equation is a dichotomous variable where the divesting firm has been represented by ‘1’ and a non-divesting firm or the matching firm represented by ‘0’.
Explanatory variables
The explanatory variables (whether they are called independent or control variables) in the selection equation come from a wide variety of explanations and theories that have been used to understand divestments. They are detailed below, briefly, along with the measures used.
Real options theory
As outlined in the literature review, uncertainty in a business unit’s environment is a key explanatory variable for testing real options theory in the context of divestments.
A time-varying estimate of business unit’s/segment’s environmental uncertainty was 70 needed. It is a common practice to quantify the construct of uncertainty by calculating the variance of indicators such as stock price, GDP or sales over time. Such approaches fail to account for trends in data that can increase the measured variance while actually not being an element of uncertainty if they were predictable. Also, such approaches do not accommodate the possibility of variances being heteroskedastic that is very characteristic of time series.
Following Folta and O’Brien (2004), Carruth et al., (2000), generalized autoregressive conditional heteroskedasticity models (GARCH) were used. The conditional variances generated were used as a measure of uncertainty (Bollerslev, 1986;
Engle, 2001). In particular, the GARCH-M (1,1) models were run on value-weighted industry portfolio 37 returns that were developed from monthly stock returns (adjusted for dividends and splits) for all firms in the CRSP database from 1950-2004.38
The GARCH-M model can be written as follows:
rt = α+γht-1+ρrt-1+δε t-1+εt
2 h t = κ+ρ1 h t-1+ δε t-1
εt = sqrt( ht zt) and zt ~ N(0,1)
This model represents the generalized autoregressive conditional heteroskedasticity in-mean specification, GARCH-M, with ARMA (1, 1) in the mean equation. ‘ εt’ representing the error term, is conditionally normally distributed and serially 37 All industry groupings are based on Robins and Wiersema (1995) classification. 38 The data has been checked for white noise using the Portmonteau’s Q-test, the correlograms and Bartlett’s periodogram based white noise test. Where there was evidence of white noise, an ARIMA (0,0,1) term was introduced to mitigate the situation and the tests were repeated to confirm white noise. 71 uncorrelated. ‘h t’, the conditional variance, is a linear function of the past period’s
2 39 squared errors, εt-1 , and the last period’s conditional variance, h t-1, i.e., GARCH(1,1).
The ARMA (1,1) in the mean equation implies that the conditional returns in this model are a linear function of the last period’s conditional variance, past conditional returns and past disturbance. Under this rich specification, volatility can change over time and expected returns are a function of volatility as well as past returns. The monthly conditional variances were averaged to obtain annual figures. The variable business unit’s environmental uncertainty is computed as the square root of the average yearly conditional variance.40 The lagged variable on uncertainty was used by considering the uncertainty in the year preceding the closest financial year to the event date.
Information asymmetry
Again, as mentioned in the literature review, information asymmetry can influence a firm’s decision to divest. Generally, information asymmetry about a parent firm has been used as a measure in prior literature. However, if a business unit is being divested, whether the unit will be properly valued will depend on whether or not there is an appropriate comparison available in the segment’s industry. Therefore, the relevant measure would be the level of information asymmetry in the business unit’s industry.
39 In fitting time series models, the simplest models are fitted first. Higher models with more number of autoregressive terms call for greater number of parameters. They are not used unless there is a good reason to believe that the model specified is not capturing the process sufficiently. Here, the GARCH-M(1,1) captures the underlying process well, as in case of several other studies cited, and therefore there was no need to use an alternative specification. 40 Results do not change with the use of alternative measures such as the variance in monthly returns in the year prior to divestment and the average variance in returns in three years prior to divestment. 72 Information asymmetry in the business unit’s industry was measured as the mean value of the standard deviation in analysts’ forecasts from IBES for the segment’s industry grouping in a particular year (Krishnaswami and Subramaniam, 1999). This variable has also been lagged by one year.
Parent and segment performance
By far, the most dominant explanations for divestments in the literature are based on parent and segment/business unit level performance. Divestments have been seen as means to restore corporate efficiency. Therefore, poor parent performance has been argued to positively influence the decision to divest (Cho and Cohen, 1997; Harrigan,
1981, 1982; Jain, 1985; Montgomery and Thomas, 1988). Further, poor performance at the business unit level has also been found to be a key determinant of divestments
(Vignola, 1974; Patton and Duhaime, 1978; Ravenscraft and Scherer, 1991; Chang, 1996;
Duhaime and Grant, 1984; Hamilton and Chow, 1993; Hitt et al., 1996).
In this study, parent and business unit level performance were controlled for using return-on-assets. This measure has been considered more appropriate for this study since firms operating in manufacturing and related industries are asset intensive. Return on assets was computed as operating income before depreciation over total assets
(Powers, 2001) and both the business unit and corporate levels. 41
41 Return on sales measure has been computed as operating income before depreciation over net sales and was used to check the robustness of results to alternative performance measures. Accounting measures of performance have been chosen as opposed to market measures since market measures are not available at the segment level. 73 Agency theory
Another major explanation for divestment decisions was based on agency theory.
Managerial self-interest and poor governance mechanisms were argued to adversely influence decisions to divest (Bethel and Leibeskind, 1993; Finkelstein and Hambrick,
1989; Jensen and Murphy, 1990). Strong governance mechanisms such as large block- holder ownership were shown to reduce such tendencies and favorably influence divestments (Bethel and Liebeskind, 1993; Hoskisson, et al., 1994; Sanders, 2001).
While there have been arguments for and against the monitoring efficacy of outside board members (Baysinger and Hoskisson, 1990), block-holders seem to have an incentive to monitor firms more closely. Therefore, agency explanations of divestments were controlled in this study by using the level of stock held by block-holders as a proxy.42 The normally accepted definition of block-holders as those who control more than 5 per cent of the firm shares has been used (Bethel and Liebeskind, 1993).
Transaction costs economics and behavioral uncertainty
Transaction costs economics is yet another theory applied to understand divestments. The core of the TCE argument is that, at any point in time, uncertainty about the behavior of partners to a transaction will have an impact on whether or not a transaction is internalized. Transactions with higher behavioral uncertainty that can be efficiently managed by ‘fiat’ will be internalized, and those that involve less behavioral
42 Measuring block holder ownership only by considering institutional owners could cause a slight bias because individual block-holders are not included. 74 uncertainty for which ‘fiat’ would be relatively costly, will not be integrated or will be divested (Hill and Hoskisson, 1987; Jones and Hill, 1988; Markides, 1992).
A measure of behavioral uncertainty should capture the magnitude of the coordination problems that arise due to transaction specific investments in recurrent transactions, particularly under conditions of uncertainty (Williamson, 1979; 1985). It thus can be measured as an interaction between transaction specific investments, frequency of transactions, and the levels of uncertainty. Unfortunately, several TCE studies have captured only one of these three dimensions of behavioral uncertainty and, the divestment literature is no exception (David and Han, 2004; Carter and Hodgson,
2006).
In this study, it was important to capture the behavioral uncertainty between the business unit and the parent firm in order to assess the transaction costs considerations in the decision to divest. If a business unit is very closely related to the other business units in the parent firm, it would be expected to have greater need for coordination with the rest of the firm. For example, if the business unit is a part of an integrated production process, there are likely to be more transaction specific investments between the parent and the business unit. Further, the frequency of interactions with other business units that are a part of the same process are also likely to be higher. On the other hand, if the business unit is a stand-alone and is less connected with other units, there will be lesser need for coordination since transaction specific investments will be fewer and also interactions with other units will be relatively infrequent.
75 Therefore, a business unit’s relatedness to the parent company can be a proxy for the level of transaction-specific investments and frequency of interactions between the business unit and the parent firm. The interaction of the business unit’s relatedness with parent firm and, the level of uncertainty in the segment’s industry can thus proxy the behavioral uncertainty. Controlling for this variable became important in order to separate the real options explanation (based on environmental uncertainty) from that of
TCE (behavioral uncertainty).
Parent and business unit relatedness
Further, the level of parent’s portfolio interrelatedness and the level of relatedness of the business unit to be divested with the parent firm were other important determinants of divestment decisions. Greater levels of relatedness, in general, have been found to be negatively related to divestments (Hoskisson, Johnson and Moesel, 1994; Chang and
Singh, 1999).
The level of diversification was measured using related entropy (Jacquemin and
Berry, 1979) 43 since parent’s resource-based relatedness was highly correlated with several other variables of the study. In the present case, the relationship between the particular division in question (the division divested via spin-off, carve-out or sell-off) and the other divisions within the parent company was important. Therefore, a sales- weighted measure of interrelationship of the focal division with the other divisions in the
43 The basic entropy index was computed by Jacquemin and Berry (1979) as E= ΣPi* ln(1/Pi). In this equation, E signifies the entropy measure and Pi is the proportion of a firm’s sale in SIC industry i. This is typically treated as a measure of total diversification when calculated at the 2-digit SIC level. The unrelated diversification is computed using the proportions at the 2-digit SIC level and subtracted from the 4-digit level to obtain a measure of ‘related’ diversification in the portfolio. 76 parent firm was calculated following Robins and Wiersema (1995) 44 and then summed to get the extent of resource-based relatedness of the segment with the parent firm.
Other controls
Other important variables that needed to be controlled, following previous literature, were parent debt position (leverage), firm size, and business unit size. Parent’s debt position (leverage) was measured as total long term debt over total common equity
(lev1) and long term debt over market value of equity (lev4), again to check the robustness of results to alternative measures (Chang and Singh, 1999). Firm size was measured as log of total assets and business unit size was measured as the proportion of the parent’s total assets that are in the segment i.e., segment assets/parent’s total assets
(Duhaime and Baird, 1987; Bergh, 1995).45 Again, an asset based measure was considered more appropriate due to the nature of the firms in this study.
Also, a measure of growth options in other business units/segments of the firm was calculated using the market-to-book value of the median firm in the industry to
44 Robins and Wiersema (1995) developed a resource-based relationship index to measure a firm’s overall portfolio interrelatedness. According to Robins and Wiersema, for each combination of two different industry categories ‘i’ and ‘j’ in a firm’s portfolio, the sales-weighted measure of interrelationship Rij is given by R ij =P i r ij + P j r ij , where Pi= percentage of sales in industry category i and Pj= percentage of sales in industry category j. These weighted measures of similarity between paired industries (Rij) are summed over all combinations of two industries that could be formed in a business portfolio of the firm resulting in an aggregate index of interrelationship of the businesses of the firm Mk= Σ Rij= Σ rij(Pki+ Pkj), where i and j represent any two different industries in which the firm k is active. The authors then introduce a correction factor to avoid double counting. This is the parent’s total portfolio interrelatedness. Segment’s interrelatedness is measured as the sum of the segment’s relatedness with other segment’s in the parent’s portfolio. The parent’s resource-based portfolio interrelatedness measure was very highly correlated with segment relatedness calculated using the resource-based measure. Therefore, the entropy measure was used for parent’s portfolio relatedness. Results on key variables are robust when the segment relatedness is dropped and the parent’s resource based portfolio measure was used. Parent’s resource-based relatedness measure was significant itself in negative direction as expected. 45 Controlling for the total number of segments was considered. However, just that number would not provide any more information than what is obtained from firm size and the measures of relatedness. 77 which the business units belonged to. This median value was weighted with the segment’s proportion of the parent firm’s sales. The weighted sum was taken as a measure of the total growth options in other businesses of the firm.
3.3.5 Variables and measures for the outcome equation
Decisions related to divestment modes are not necessarily independent of the decision to divest a business unit. Since parent level concerns get accounted for in the decision to divest, they can serve as instruments to distinguish between the two decisions.
Including parent level variables as determinants of divestment mode choice may be redundant for the following reasons. While parent level variables have been used to explain one or more divestment mode choices in the current literature, these variables seem to influence the more basic decision to divest, than divestment through a particular mode. For example, from an agency perspective, the possibility of spin-offs at a future date has been offered as a mechanism to facilitate better incentive alignment for divisional managers, ex-ante (Aron, 1991). However, this purpose can be accomplished even with an equity carve-out. Even the possibility of a future sell-off can have a disciplining effect and align managerial incentives, ex-ante. Thus, there is no reason to expect that a particular mode of divestment will be preferred over others because of agency reasons.
Also, in some studies, parent’s level of diversification was offered and examined as a determinant of divestment mode choice. Less diversified parent firms were argued to choose staged divestments over complete divestments as compared to more diversified parent firms (Bergh et al., 2008). However, the way in which a business unit gets 78 divested is more likely to depend on its own relatedness with the parent firm than the overall parent diversification (Chang and Singh, 1999). Similar explanations hold for parent’s performance, debt level etc., as explanations for the decision to divest, than for the divestment mode choice decision.
Dependent variable
The dependent variable in the outcome equation is a dichotomous variable with a staged divestment (spin-off/carve-out) represented by ‘1’ and complete divestment (sell- off) represented by ‘0’.
Control variables
The control variables are business unit’s performance, the extent to which the business unit is related to the parent firm, the behavioral uncertainty related to the business unit, and business unit size, which have been commonly studied in the literature.
The relevant measures have been explained as part of the selection equation.
Independent variables
Uncertainty
Uncertainty is a key variable in this model for the divestment mode choice decision. The relevant measure of uncertainty has been explained as part of explanatory variables for the selection equation—i.e., the decision to divest.
Information asymmetry
Information asymmetry is the other key variable in this model for the divestment mode choice decision. The measure of information asymmetry has also been explained as part of explanatory variables for the selection equation. 79 3.4 Analysis and results
3.4.1 Summary statistics
The summary statistics for divesting and non-divesting parent-segment combinations are in Table 3.1. A quick glance shows that, on average, the parents that have divested a business segment seem to have higher institutional block-holdings, greater levels of diversification in their portfolios, higher leverage ratios, larger firm sizes and more growth options in other businesses as compared to non-divesting firms. Also, divested segments are larger in size than non-divested segments.
On the other hand, parents that have divested a business segment seem to have lower uncertainty in the segment’s industry, lower information asymmetry in the segment’s industry, poorer performance, lower behavioral uncertainty between the parent and segments, and lower parent firm valuations as compared to non-divesting firms.
Further, the segments that have been divested seem to be less related to their parent firms’ and had poor performance as compared to those segments that have not been divested. Some of these differences are significant whereas others are not. The t-tests for significance of differences between means are also reported in table 3.1.
Table 3.2 shows the summary statistics for firms that engage in staged divestments and complete divestments. Parent firms engaging in staged divestments seem to have higher information asymmetry in the segment’s industry, better performance at parent and segment levels, more tightly related portfolios of businesses, better valuations, and larger firm sizes as compared to the firms divesting through sell-
80 offs. Further, the segments being divested in a staged manner are more closely related to their parent firms than those being divested through sell-offs.
On the other hand, parent firms engaging in staged divestments seem to have lower uncertainty in the segment’s environment, less ownership by institutional block- holders, lower debt, lesser behavioral uncertainty between the parent and the segment, and fewer growth options in other segments as compared to firms engaging in complete divestments. Also the segments that are divested in a staged manner seem to be smaller in size as compared to those divested completely. The t-tests show that only the differences in information asymmetry in segment’s industry, parent and segment performance and, parent’s valuation are significant whereas other differences are not significant. The difference in debt positions is sensitive to the measure used.
The correlation matrix in Table 3.3 shows no major problems of multi-collinearity among the variables except for behavioral uncertainty and segment relatedness (0.961) indicated by * in the table. 46 The result on the key variable—segment’s industry uncertainty—holds with and without the measure of behavioral uncertainty. Therefore the results or interpretation of this study do not change with the presence or absence of this variable.
46 Parent’s resource-based portfolio relatedness measure is significantly correlated with several other variables, including segment’s relatedness. Therefore, this variable was replaced by the parent’s related entropy measure of portfolio relatedness to avoid the problem. Even parent performance and segment performance are correlated upto 0.66. However, results are robust to alternative measures of parent and segment performance. 81 Table 3.4 shows the main effects for the selection and outcome equations. The economic significance will be revealed by the marginal effects. Therefore, they are the most relevant for interpretation in probit analysis and only those results have been interpreted.
3.4.2 Selection equation results Table 3.5 shows marginal effects 47 for the selection equation. In each table, model 2 is a variant of model 1 with a different leverage measure. Models 3 and 4 are variants of models 1 and 2 respectively, and control for transaction costs economics explanation using ‘behavioral uncertainty’ variable. 48 Models 5-8 are counterparts of models 1-4 with controls for growth options in other businesses.
From Table 3.5, it can be seen that some results for the selection equation—i.e., for the decision to divest—were very consistent with previous studies on divestment whereas others were not. Uncertainty in the business unit’s environment was significant and negatively associated with the decision to divest. This result is the same as obtained in the first essay and supports the real options view that firms’ decisions to divest business units may be driven by option considerations under conditions of high uncertainty. Also, parent performance and valuation were negatively associated with divestments. This supports the idea that parent firms may engage in divestments when they are under performance pressures and/or suffering poor valuations.
47 All marginal effects are at the mean values of other variables. 48 The model here has an interaction term, but the results hold regardless of the presence of the interaction term. Also, the standard errors are not inflated much with the introduction of new variables. Together, these suggest there are no serious multi-collinearity problems in this data. 82 Further, the level of diversification in a firm’s portfolio of businesses was positively associated with the decision to divest. Also, segment’s relatedness to the parent firm was significant and negatively associated with the decision to divest. This result, however, disappeared in models 7 and 8 that controlled for both behavioral uncertainty between the segment and parent and also the growth options in other segments. Firm size was also positively associated with the decision to divest (again, firm size was not significant in models 5-8 that included controls for growth options in other businesses).
All these results are, in general, consistent with several previous studies on divestments which show that increased levels of diversification and larger firm sizes may trigger divestments, whereas segments that are more related to the parent firms are less likely to be divested.
On the other hand, some of the results for the selection equation were either counter-intuitive or contradictory to previous studies. For example, information asymmetry in a segment’s industry was not significant to the decision to divest. This is quite counterintuitive since information asymmetry was expected to deter divestments in view of the adverse selection costs associated with it. Further, the level of stock held by institutional block-holders, a measure of effective governance structures in organizations, was not significant in models that involved controls for growth options in other businesses (model 5-8). Again, this is not a strong support for the popular agency theory according to which stronger governance mechanisms, such as block-holder ownership,
83 induce divestments.49 Also, segment performance was not significant in any of the models. This result on segment performance contradicts previous findings that poor segment performance triggers divestments (Cho and Cohen, 1997). The debt position of the parent firm was sensitive to the measure used and was significant with positive sign only in one model.
Furthermore, the behavioral uncertainty variable that captures TCE concerns was not significant. However, since this variable is an interaction term, its interpretation in the probit model with a selection correct is not straightforward (Hoetker, 2007a).
Segment size also did not have any significant relationship with the decision to divest. On the other hand, growth options in other segments were significant and negatively associated with the decision to divest. This is also counter-intuitive since firms may divest some business units to redeploy capital into other business units with more promising options.
3.4.3 Outcome equation results
Table 3.6 shows the marginal effects 50 for the outcome equation. From this table, it can be seen that segment level performance was, in general, not significant for the decision to engage in staged divestments. This variable was significant only in model 7 at 5 percent level. In all other models, it was significant only at less than 10 per cent level. Segment’s relatedness, on the other hand, was negatively associated with staged divestments. This is very counterintuitive compared to earlier findings that closely
49 In fact, in models where only the decision to divest is considered, agency explanation was not significant at all. 50 All marginal effects are at the mean values of other variables. 84 related segments are more likely to be divested in a staged manner (Chang and Singh,
1999). This variable was, of course, not significant in models that included the behavioral uncertainty variable (which is highly correlated with segment’s relatedness). As mentioned previously, the lack of significance of the behavioral uncertainty variable needs further analysis. Also, segment size was not significant in any of the models. This contradicts earlier findings that smaller segments are more likely to be sold-off (Chen and
Guo, 2005).51 Further interpretation and implications of these results are in the discussion section.
Hypothesis 1 was about the real options explanation for divestment mode choice—i.e., high uncertainty in a business unit’s environment positively influences its divestment in a staged manner. This hypothesis was not supported in any model.
Uncertainty in a business unit’s environment did not seem to matter for the divestment mode choice. This variable was insignificant in most models. Even where this variable was significant at 10 per cent level, it had a negative sign indicating that higher uncertainty may actually lead to complete divestments. This shows that staged divestments may not be driven by option considerations.
Hypothesis 2 was the information asymmetry explanation for divestment mode choice—i.e., high information asymmetry in a business unit’s environment positively influences its divestment in a staged manner. This hypothesis was supported in all models in Table 3.6. The level of information asymmetry in a business unit’s industry
51 In a certain sense, the outcome equation results are not comparable to prior studies because the modeling is very different here. 85 was positive and significant to the decision to engage in staged divestments. This result supports the idea that staged divestments such as spin-offs/carve-outs are more likely under conditions of high information asymmetry about the true value of a business unit.52
Further, Table 3.5 shows the results of likelihood ratio test for independence of selection and outcome equations—i.e., the decision to divest and the divestment mode choice decision. This tests the null that the two decisions are independent. The null was rejected at less than 5 per cent level in all models showing that the two decisions are not independent.
3.5 Discussion and conclusion
This study tested whether firms engage in staged divestments in order to keep their ‘options open’ under conditions of high uncertainty, or due to the presence of high information asymmetry, or both. It showed that firms are more likely to engage in staged divestments (spin-offs/carve-outs) when there is significant information uncertainty in the business unit’s industry, as compared to when there is less information asymmetry. This result is consistent with the ideas that 1) when information asymmetry is high, firms try to mitigate adverse selection costs and ensure that business units are valued fairly before they sell them 2) firms are likely to keep their private information and release it at appropriate time in the divestment process in order to maximize the proceeds from sale.
This result holds even after controlling for other explanations of divestment mode choice such as the poor performance of a divested business, business unit size etc. 52 The marginal effects for the joint probability of the decision to divest and decision to engage in staged divestments also support these findings. The marginal effects for the probability of staged divestment, conditional on the decision to divest also does not alter the results with regard to the uncertainty and information asymmetry. 86 Further, this study showed that uncertainty in a business unit’s environment does not really matter to divestment mode choice decisions. This result questions the extent to which real options theory, as it developed in the context of investments, will be applicable to the context of divestments. From a real options perspective, it has been argued theoretically, and shown empirically, that firms invest in a staged manner under conditions of high uncertainty (Bowman and Hurry, 1993; Kogut, 1991). This does not seem to be the case in the divestment context.
There are at least two important reasons for staged investments under conditions of high uncertainty that would not hold for staged divestments. The first one is about staged investments being platforms for growth options and facilitating capability development (Bowman and Hurry, 1993; Kogut and Kulatilaka, 2001). This reason will not hold for staged divestments because firms are already invested in the business unit any way. Staged divestments are not likely to generate (or keep) growth options any better compared to non-divestment.
A second major reason offered for staged investments under conditions of uncertainty is competitive preemption. It has been argued that firms may prefer staged investments as compared to deferring investment completely under conditions of uncertainty. This is to obtain strategic positions in evolving markets and not being locked-out (Kulatilaka and Perotti, 1998). Such competitive preemption can also not be a reason for staged divestments. In fact, scaling down operations, by divesting even partly, can advantage competitors. Therefore, firms may actually keep their positions intact rather than exit any bit of the landscape. 87 Also, the result that uncertainty in a business unit’s environment does not really matter to divestment mode choice decisions is very interesting when understood along with the result from the selection equation that uncertainty does, indeed, have a significant negative association with the decision to divest. The picture that emerges from these results is that firms are most likely to keep ‘options open’ when the prospects of business are uncertain (Merton, 1998). However, when the decision is in favor of divestment, it is the level of information asymmetry that determines whether or not a divestment is staged.
One of the very curious results was about business unit’s relatedness to the parent and its impact on divestment mode choices. Some studies show that related business units may be divested in a staged manner, than be completely divested right away (Chang and Singh, 1999). However, when divestment mode choices were modeled jointly with the decision to divest, the results show that segments that are closely related to their parent firms are more likely to be divested completely, than in a staged manner.53
This result is still consistent with the idea that when potential buyers can siphon- off profits more easily after acquiring the control of a business unit, sell-offs are a better strategy to maximize proceeds from a divestment. For example, when subsidiaries are part of integrated production processes, it may be difficult to assess the profitability of 53 It is also interesting that Chang and Singh (1999) actually were the first to mention that divestment decisions and divestment mode choices would be driven by overlapping yet different factors, and that a two-stage model would be more appropriate for understanding these decisions. They, however, model the divestment decision and divestment mode decisions using separate logistic regressions. Here, the methodology actually allows for modeling these two decisions jointly and it is shown that, indeed, the decision to divest and the divestment mode choices are not independent (rho in all models was not equal to zero, indicating that the decisions are related, and this was highly significant result). 88 the subsidiaries independent of the profitability of the acquirer. This leaves majority shareholders more discretion in increasing the non-verifiable component of income at the expense of the verifiable one. Therefore, direct sell-off will the most profitable divestiture strategy to maximize the proceeds from the sale when businesses are highly related, and staged divestments may be preferred in other cases (Zingales, 1995).
To conclude, this study made two important contributions. The first one was theoretical, in terms of deciphering whether real options logic and/or information asymmetry explains divestment mode choices. It was important to separate these arguments because the underlying reasons for engaging in staged divestments are very different from these perspectives. The second was an important methodological contribution. A conventional multinomial model with non-divestment, staged divestments and complete divestments as the alternatives, may lead to very different conclusions than the conclusions drawn from the results of this study.54 However, the probit model with selection—which approaches managerial decision-making in a sequential manner—shows that the decision to divest and the divestment mode choice decisions are different, yet related, decisions. This changes the way divestment decisions and divestment mode choices have been studied and understood so far. At the very least, it shows that there are very counter-intuitive results that could be obtained when these decisions are modeled differently.
54 Results from the multinomial model show that, at least in some models, uncertainty has a significant negative relationship with the decision to engage in a complete divestment. 89 t-test for difference in Firm type Non-Divesting Firms Divesting Firms means Variable Obs Mean Std. Dev. Obs Mean Std. Dev.
Uncertainty in segment’s environment 598 0.058496 0.014475 284 0.056132 0.013338Significant
Information asymmetry in Segment industry 598 0.328437 2.284484 284 0.205025 0.280565 Not significant Institutional Blockholder Share 598 10.45375 13.71221 284 12.12977 13.19485Significant
Parent performance 598 0.116774 0.165887 284 0.081989 0.162127Significant
Segment performance 598 0.089697 0.254113 284 0.033941 0.285521 Not significant
Leverage (lev1) 598 0.191419 0.155884 284 0.22786 0.164159 Significant
Leverage (lev4) 598 0.431839 0.6753 284 0.772585 1.900307 Significant Level of diversification in parent’s portfolio (Related entropy) 598 0.053941 0.176923 284 0.08977 0.203494Significant
Parent's portfolio Relatedness (resource-based relatedness measure) 598 0.505182 0.471009 284 0.299031 0.400301Significant
Segment's resource-based relatedness 598 0.372113 0.553297 284 0.192196 0.362122Significant
Behavioral Uncertainty 598 0.022404 0.034974 284 0.011402 0.022685 Significant Parent's Valuation (sales measure) 598 1.947818 6.663471 284 1.066466 1.441725 Not significant
Parent Size 598 8.188508 1.978698 284 8.384407 1.893143 Not Significant
Segment Size 598 0.526375 0.388926 284 0.573977 0.386638 significant Growth options in other segments 463 7.713048 19.4278 248 3.316355 10.32054Not significant Table 3.1: Summary statistics for divesting and non-divesting parents 90 t-test for Staged divestment Complete divestment means Divestment Mode Obs Mean Std. Dev. Obs Mean Std. Dev. Uncertainty in segment's Not environment 102 0.054676 0.010286 182 0.056948 0.014739 significant
Information asymmetry in the segment's industry 102 0.244198 0.314952 182 0.183071 0.257634 Significant Institutional blockholders' Not share 102 10.70402 11.47279 182 12.92882 14.03503 significant Parent performance 102 0.113423 0.097519 182 0.064373 0.186875 Significant Segment performance 102 0.102332 0.149946 182 -0.00439 0.332871 Significant Parent's debt Not (lev1) 102 0.221048 0.146248 182 0.231677 0.173662 significant Parent's debt (lev4) 102 0.436083 0.765706 182 0.961174 2.284611 Significant Level of diversification in parent’s portfolio Not (Related entropy) 102 0.085237 0.210346 182 0.092311 0.200096 significant Parent's portfolio relatedness (resource-based relatedness Not measure) 102 0.326045 0.416343 182 0.283891 0.391374 significant Segment's resource-based Not relatedness 102 0.194567 0.386821 182 0.190867 0.348611 significant Behavioral Not Uncertainty 102 0.010724 0.021674 182 0.011782 0.023282 significant
Parent's valuation 102 1.406982 1.697089 182 0.875627 1.2413 Significant Not Parent firm size 102 8.536232 1.841924 182 8.299318 1.921008 significant Not Segment size 102 0.55032 0.380714 182 0.587235 0.390332 significant Growth options in Not other segments 85 2.80134 7.839004 163 4.872456 26.8707 significant Table 3.2: Summary statistics for staged and complete divestments 91 Information Uncertainty in asymmetry in Institutional segment's segment block-holder Parent Segment Variable Divesting firm industry industry share performance performance Leverage (lev1) Divesting firm 1 Uncertainty in segment's industry -0.0781 1 Information asymmetry in segment industry -0.0306 -0.0036 1 Institutional blockholder share 0.0578 0.0596 -0.0105 1 Parent performance -0.0983 -0.0406 -0.0133 0.0483 1 Segment performance -0.0981 -0.0272 -0.0094 0.0131 0.6462 1 Leverage (lev1) 0.1069 -0.0114 -0.0085 0.0404 0.0133 -0.0386 1 Leverage (lev4) 0.1303 0.0258 0.0008 0.0635 -0.0459 -0.0997 0.4673 Level of diversification in parent's portfolio (Related entropy) 0.0898 -0.0042 -0.0208 0.1236 0.0294 0.0502 0.056 Segment's resource-based relatedness to parent firm -0.166 0.1008 0.0024 -0.002 0.1119 0.1918 0.0238 Behavioral uncertainty -0.161 0.2541 0.0035 0.0225 0.104 0.193 0.0285 Parent's valuation -0.0741 -0.0189 -0.0056 0.0032 -0.2123 -0.1327 -0.1061 Firm size 0.0469 -0.0089 0.0383 0.0795 0.3475 0.1897 0.132 Segment size 0.0573 -0.0732 -0.005 -0.0219 -0.1776 -0.1342 -0.0871 Growth options in other businesses -0.0717 0.0246 -0.0104 -0.0001 0.0646 -0.0034 0.1403 Table 3.3: Correlation matrix Table 3.3 (continued)
92 Table 3.3 (continued)
Level of diversification Segment's in parent's resource- Growth portfolio based options in Leverage (Related relatedness to Behavioral Parent's Segment other Variable (lev4) entropy) parent firm uncertainty valuation Firm size size businesses Leverage (lev4) 1
Level of diversification in parent's portfolio (Related entropy) -0.026 1
Segment's resource- based relatedness to parent firm -0.006 0.0507 1
Behavioral uncertainty -0.0005 0.0364 0.961* 1
Parent's valuation -0.0959 -0.0575 -0.0911 -0.0857 1 Firm size 0.0819 0.1136 0.2073 0.2077 -0.1572 1
Segment size -0.0663 -0.254 -0.2748 -0.2736 0.1579 -0.3753 1 Growth options in other businesses 0.0311 -0.0341 0.0919 0.09 -0.0185 0.0247 -0.0645 1 93 Model 1 Model 2 Model 3 Model 4 Selection Outcome Selection Outcome Selection Outcome Selection Outcome Variable Equation Equation Equation Equation Equation Equation Equation Equation
Uncertainty in the business unit's environment -8.492** -9.344* -8.072** -9.516* -11.527** -5.53 -11.050** -5.33 [2.23] [1.79] [2.12] [1.79] [2.56] [0.90] [2.45] [0.85] Information asymmetry in the business unit's environment 0.03 0.426* 0.04 0.453** 0.01 0.449** 0.03 0.474** [0.16] [1.95] [0.27] [2.03] [0.08] [2.07] [0.20] [2.14]
Institutional blockholder share 0.008** 0.008** 0.008** 0.008** [2.42] [2.49] [2.40] [2.45]
Parent's performance -1.518*** -1.684*** -1.475*** -1.634*** [3.07] [3.40] [2.99] [3.31] Business unit's performance 0.01 0.816** 0.02 0.871** -0.04 0.819** -0.03 0.867** [0.03] [2.02] [0.06] [2.09] [0.13] [2.06] [0.11] [2.12] Parent's leverage (lev1) 0.48 0.48 [1.63] [1.64] Parent's leverage (lev4) 0.146** 0.145** [2.41] [2.40] Level of diversification in parent’s portfolio (Related entropy) 0.825*** 0.774*** 0.845*** 0.795*** [3.50] [3.28] [3.57] [3.36] Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 3.4: Main effects for selection and outcome equations Table 3.4 (continued) 94 Table 3.4 (continued) Model 1 Model 2 Model 3 Model 4 Selection Outcome Selection Outcome Selection Outcome Selection Outcome Variable Equation Equation Equation Equation Equation Equation Equation Equation Business unit's resource-based relatedness to the parent -0.539*** -0.361** -0.546*** -0.359** -1.221** 0.7 -1.199** 0.8 [4.58] [2.16] [4.64] [2.09] [2.09] [0.71] [2.07] [0.79] Behavioral Uncertainty 11.77 -18.79 11.28 -20.43 [1.20] [1.11] [1.15] [1.18] Parent's valuation -0.103*** -0.115*** -0.105*** -0.118*** [4.01] [4.46] [4.08] [4.54] Parent firm size 0.065** 0.065** 0.063** 0.063** [2.39] [2.37] [2.37] [2.32]
Business unit size 0.255* -0.07 0.242* -0.08 0.258* -0.08 0.245* -0.08 [1.90] [0.44] [1.80] [0.48] [1.93] [0.46] [1.83] [0.49] Constant -0.44 -0.645** -0.46 -0.628* -0.25 -0.856** -0.27 -0.862** [1.27] [1.99] [1.27] [1.90] [0.68] [2.31] [0.69] [2.29] Observations 882 882 882 882 882 882 882 882 LR test of independence (rho=0) 0.0003 0.0017 0.0002 0.0013 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 3.4 (continued) 95 Table 3.4 (continued) Model 5 Model 6 Model 7 Model 8 Selection Outcome Selection Outcome Selection Outcome Selection Outcome Variable Equation Equation Equation Equation Equation Equation Equation Equation
Uncertainty in the business unit's environment -9.197** -8.99 -9.130** -9.19 -11.362** -4.8 -11.170** -4.76 [2.25] [1.54] [2.24] [1.56] [2.39] [0.70] [2.35] [0.69] Information asymmetry in the business unit's environment 0.05 0.512** 0.07 0.527** 0.05 0.540** 0.06 0.554** [0.31] [2.20] [0.38] [2.24] [0.26] [2.29] [0.35] [2.32]
Institutional block- holder share 0.01 0.01 0.01 0.01 [1.32] [1.33] [1.29] [1.31] Parent's performance -1.431** -1.531*** -1.389** -1.491*** [2.56] [2.74] [2.46] [2.64] Business unit's performance -0.12 0.971** -0.12 1.007** -0.17 1.026** -0.16 1.062** [0.35] [2.09] [0.33] [2.11] [0.48] [2.18] [0.45] [2.20] Parent's leverage (lev1) 0.36 0.37 [1.07] [1.09] Parent's leverage (lev4) 0.1 0.1 [1.30] [1.34] Level of diversification in parent’s portfolio (Related entropy) 0.676** 0.644** 0.689** 0.656** [2.45] [2.33] [2.47] [2.35] Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 3.4 (continued)
96 Table 3.4 (continued) Model 5 Model 6 Model 7 Model 8 Selection Outcome Selection Outcome Selection Outcome Selection Outcome Variable Equation Equation Equation Equation Equation Equation Equation Equation
Business unit's resource-based relatedness to the parent company -0.522*** -0.469** -0.525*** -0.471** -1.067* 0.9 -1.03 0.96 [3.93] [2.29] [3.96] [2.27] [1.66] [0.77] [1.61] [0.80] Behavioral Uncertainty 9.33 -23.92 8.63 -24.99 [0.87] [1.19] [0.81] [1.22] Parent’s Valuation -0.112*** -0.118*** -0.113*** -0.120*** [3.78] [4.04] [3.80] [4.06] Parent firm size 0.055* 0.054* 0.053* 0.052* [1.78] [1.74] [1.72] [1.67]
Business unit size 0.21 -0.04 0.2 -0.04 0.21 -0.04 0.21 -0.05 [1.37] [0.20] [1.32] [0.23] [1.40] [0.22] [1.35] [0.24] Growth options in other business segments -0.006** -0.006** -0.006** -0.006** [2.02] [2.04] [1.97] [1.99] Constant -0.12 -0.671* -0.12 -0.653* 0.01 -0.898** 0 -0.894** [0.31] [1.77] [0.31] [1.70] [0.03] [2.13] [0.01] [2.10] Observations 711 711 711 711 711 711 711 711
LR test of independence of equations (rho=0) 0.0072 0.0183 0.0085 0.0203 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% 97 Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Uncertainty in segment's environment -2.964** -2.817** -4.022*** -3.853** -3.330** -3.305** -4.113** -4.042** -2.23 -2.12 -2.56 -2.45 -2.25 -2.24 -2.39 -2.35 Information asymmetry in the segment's industry 0.009 0.015 0.005 0.011 0.019 0.024 0.016 0.021 0.16 0.27 0.08 0.2 0.31 0.38 0.26 0.35 Institutional block- holdings 0.003** 0.003** 0.003** 0.003** 0.002 0.002 0.002 0.002 2.42 2.49 2.4 2.45 1.32 1.33 1.29 1.31
Parent's performance -0.530*** -0.587*** -0.515*** -0.570*** -0.518*** -0.554*** -0.503** -0.540*** -3.07 -3.4 -2.99 -3.31 -2.56 -2.74 -2.46 -2.64 Segment performance 0.003 0.006 -0.014 -0.012 -0.044 -0.042 -0.062 -0.059 0.03 0.06 -0.13 -0.11 -0.35 -0.33 -0.48 -0.45 Parent's debt position (lev1) 0.167 0.166* 0.131 0.134 1.63 1.64 1.07 1.09 Parent's debt position (lev4) 0.051* 0.051** 0.035 0.036 2.4 2.4 1.3 1.33 Level of diversification in parent’s portfolio (Related entropy) 0.288*** 0.270*** 0.295*** 0.277*** 0.245** 0.233** 0.249** 0.237** 3.5 3.28 3.58 3.36 2.45 2.33 2.47 2.36 Segment's resource- based relatedness -0.188*** -0.191*** -0.426** -0.418** -0.189*** -0.190*** -0.386* -0.373 -4.62 -4.68 -2.09 -2.07 -3.95 -3.98 -1.66 -1.61 Behavioral uncertainty 4.107 3.932 3.378 3.124 1.2 1.15 0.87 0.81 Parent's valuation -0.036*** -0.040*** -0.037*** -0.041*** -0.041*** -0.043*** -0.041*** -0.043*** -4.04 -4.5 -4.11 -4.58 -3.81 -4.07 -3.83 -4.09 Parent firm size 0.023** 0.023** 0.022** 0.022** 0.020* 0.020* 0.019* 0.019 2.39 2.37 2.37 2.31 1.78 1.73 1.72 1.67 Segment size 0.089* 0.084* 0.090* 0.085* 0.075 0.073 0.077 0.074 1.9 1.8 1.93 1.83 1.37 1.32 1.4 1.35 Growth options in other segments -0.002** -0.002** -0.002** -0.002** -2.03 -2.05 -1.97 -1.99 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 3.5: Marginal effects for selection equation 98 Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Uncertainty in segment's environment -1.843* -1.982* -1.085 -1.06 -1.81 -1.868 -0.971 -0.972 -1.78 -1.85 -0.91 -0.86 -1.51 -1.52 -0.71 -0.69 Information asymmetry in a segment's industry 0.084* 0.124** 0.088** 0.094** 0.103** 0.107** 0.109** 0.113** 1.86 2.19 1.98 2.02 2.04 2.05 2.13 2.12 Segment performance 0.161* 0.173* 0.161* 0.173* 0.195* 0.205* 0.208** 0.217* 1.87 1.91 1.93 1.95 1.89 1.87 1.96 1.93 Segment's resource-based relatedness -0.071** -0.070** 0.137 0.159 -0.094** -0.096** 0.183 0.197 -2.24 -2.15 0.7 0.78 -2.37 -2.35 0.75 0.78 Behavioral uncertainty related to the segment -3.686 -4.067 -4.84 -5.107 -1.09 -1.15 -1.15 -1.17
Segment size -0.014 -0.015 -0.015 -0.016 -0.008 -0.009 -0.008 -0.009 -0.44 -0.45 -0.46 -0.49 -0.2 -0.23 -0.22 -0.24 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 3.6: Marginal effects for outcome equation
99 CHAPTER 4
WHEN DOES INFORMATION ASYMMETRY MATTER TO THE DECISION TO DIVEST?
Information asymmetry has been a dominant theoretical perspective in the finance literature to explain managerial unwillingness to divest. Information asymmetry refers to a situation where insiders of a firm have more accurate information about the true value of a business as compared to outsiders. Myers and Majluf (1984), in their seminal paper, argued that equity issuances are fraught with information asymmetry problems.
Therefore, managers may be less willing to divest in the presence of information asymmetry, particularly when they believe assets are undervalued.
An alternative, yet equal or more powerful, argument for managerial unwillingness to divest comes from the real options theory. According to this theory, when the environment surrounding a business unit is highly uncertain, the true value of the business unit may not be known. Therefore, it might be worthwhile for managers to wait and ‘keep options open’ until at least some of the uncertainty is resolved (Merton,
1998). When uncertainty is resolved, firms may be in a better position to assess the value of the business unit than otherwise. Managers may consequently be unwilling to divest
100 when there is high uncertainty about the true value of a business unit (first essay; Dixit and Pindyck, 1995).
These two theoretical perspectives have been used rather independently of each other. However, there may be a very interesting relationship between these two theories in the context of divestments. If the environment is highly uncertain and the present value of a business unit is not likely to be an indicator of its future value, then it hard to have a ‘true’ value for the business unit. Therefore, the question whether the ‘true’ value is known to insiders better than outsiders may be less relevant. However, since the true value is not known, it could become worthwhile to wait for uncertainty to resolve in order to make a more informed decision about the divestment i.e., keep the ‘put’ option open under conditions of high uncertainty (Myers and Majd, 1990). On the other hand, if the environment is reasonably certain, then the current value of assets may indicate their future value. In such situations, if outsiders cannot assess the true current value of the business i.e., information asymmetries exist about the true value of a business unit, then outsiders may suspect the value of the assets/businesses being sold. This can lead to high adverse selection costs that can discourage divestments (Akerlof, 1970).
Considering this dynamically, when uncertainty increases, options thinking may dominate information asymmetry concerns. On the other hand, information asymmetry problems are likely to surface when uncertainty decreases. Therefore, it is important to bring these two theoretical perspectives into a common framework and understand their relative explanatory power in the context of divestments.
101 This paper brings these two theoretical perspectives together. Arguments are developed for the relative impacts of changes in uncertainty and changes in information asymmetry on the decision to divest. The hypotheses are tested using a sample of divesting and non-divesting firms. Results support the theoretical prediction that options considerations, under certain conditions, may dominate information asymmetry concerns.
4.1 Prior literature
A brief review of the literature on information asymmetry and real options as applied to divestments is reviewed in the following paragraphs.
Myers and Majluf (1984) were the first to apply the concept of information asymmetry to corporate financing decisions. In their model, managers of firms, acting in the interest of their existing shareholders’, may be willing to forego even positive NPV projects when there is asymmetric information. More specifically, when firms have undervalued assets, managers may not be willing to issue new equity. Therefore, new equity issues are more likely when investments are not very promising. The market anticipates this and discounts firms’ new issues. Knowing that they have to face market discounting, managers may be even more unwilling to issue new equity. Issuance of equity is akin to divestment. By issuing equity, entrepreneurs or promoters of the company divest themselves of a portion of ownership to outsiders. Therefore arguments made in the context of equity issuances also apply to decisions to divest in the same vein.
Implications of information asymmetry arguments have been examined or inferred primarily in terms of market reactions to equity issues. In general, markets react negatively to equity issues and this has been attributed to information asymmetry 102 between insiders and outsiders (Bayless and Chaplinsky, 1996; Mikkelson and Partch,
1986; Masulis and Korwar, 1986).
Further, the implications of information asymmetry have been explored in the context of divestment mode choices, more than for the decision to divest, per se. Nanda
(1991) extended Myers and Majluf’s framework and argued that the seemingly contradictory positive stock market reactions associated with equity carve-outs—a form of divestment in which a part of the equity in the unit to be divested is issued to new shareholders—are an indication that the assets of the parent company are undervalued.
Vijh (2002) investigated whether the gains in equity carve-outs are due to asymmetric information or due to divestiture gains and did not find support for asymmetric information hypothesis. Nanda and Narayanan (1998) put forward divisional divestment as an alternative to issuing new equity under conditions of information asymmetry. In particular, they argued that spin-offs—which involve distributions of stock to existing shareholders and not external buyers/shareholders—are less costly. Spin-offs may be a desirable way to divest since they help mitigate information asymmetry. Krishnaswami and Subramaniam (1999) show that, indeed, spin-offs are chosen by firms that have information asymmetry problems and such problems decrease post-spin-off. However, the argument that the decision to divest may itself be deterred under conditions of information asymmetry has not been subjected to empirical testing directly.
On the other hand, there are also arguments from a real options perspective for managerial unwillingness to divest. Divestments are considered ‘put’ options on real assets much as investments are considered ‘call’ options on real assets. Myers and Majd 103 (1990) modeled project divestments as ‘put’ options, with the value of the put option increasing with increasing uncertainty. Dixit (1991) argued that divestments may not occur even though the underlying causes for investments no longer exist, and that such effects would be accentuated under conditions of high uncertainty and irreversibility.
Also, Dixit and Pindyck (1995) argued that costly-to-reverse divestments will be deterred under conditions of high uncertainty. Further, focusing on reasons within the firm, Chi and Nystrom (1995) suggested that greater endogenous uncertainty could lead to exit delays. Kogut and Kulatilaka (2001) supported exit delays with the argument that managers in highly volatile environments may hesitate to radically change their tightly- coupled organizations. This is with a hope that future states of the world will provide more appealing environments.
This theoretical explanation that exit will be deterred under conditions of high uncertainty has been subject to some empirical testing. Bragger et al. (1998) showed in an experimental setting that participants who received feedback with high variability delayed their decisions to divest. Vassolo, Anand and Folta (2004) found that high environmental uncertainty was negatively associated with exits from bio-tech alliances.
The first essay tested this idea of exit deterrence under conditions of high uncertainty at the aggregate business unit level and found support that the decision to divest business units may, indeed, be driven by option considerations.
As can be seen from the previous paragraphs, arguments involving information asymmetry and the decision to divest do not explicitly take environmental uncertainty and options thinking into consideration. Similarly, arguments based on real options 104 theory do not consider the implications of information asymmetry. It therefore remains to be seen whether and when information asymmetry and environmental uncertainty have an impact on the decision to divest.
4.2 Theory development and hypotheses
4.2.1 Information Asymmetry and Uncertainty : The Constructs
It is important to distinguish between ‘information asymmetry’ and ‘uncertainty’ to be able to understand the relationships between these constructs and the decision to divest.
Information asymmetry is a situation where the true value of an asset or a business unit is known to the insiders of the firm but not to the outsiders. In such cases, it is possible that insiders (sellers) may manipulate information to outsiders’ (buyers) disadvantage. Anticipating this problem, outsiders tend to discount the true value as offered by the insiders. Fearing this discount, insiders may not transact with outsiders at all (Akerlof, 1970).
On the other hand, a situation of high environmental uncertainty is one in which future states of the world are not known with certainty (Kogut, 1991). In such circumstances, current value (performance) of an asset or a business unit may not be a good indicator of its future value (performance). Further, linkages between assets or synergies as they appear in the present may change very drastically depending on future
105 developments (Dixit and Pindyck, 1995; Perotti and Rossetto, 2007). Legal, regulatory, economic and technological changes, among others, may bring about this uncertainty.55
In the discussion about information asymmetry, the key issue is that one party to a transaction has ‘true’ or ‘better’ information as compared to another. For information to be true/better, it has to be reasonably complete and accurate. However, the ‘true’ information would itself be dubious if it is highly uncertain. In such cases, even if the information is true, it does not matter since it is highly unpredictable anyway. This forms the basis for the rest of the arguments in this paper. In a game theoretic parlance, one could think of situations of high environmental uncertainty as those of ‘incomplete information’ and situations of high asymmetric information as those of ‘imperfect information’. So, the idea is: imperfection does not matter when information is essentially incomplete .
The first essay shows that business unit divestments may be considered ‘put’ options under conditions of high uncertainty. In particular, it has been shown that environmental uncertainty has a significant negative relationship with the decision to divest. Further, an increase in environmental uncertainty strengthens this negative relationship indicating an increase in the ‘put’ option value. On the other hand, a 55 This environmental uncertainty will be common to all firms operating in a particular environment for eg., an industry. It is also important to separate ‘subjective’ or ‘perceived’ environmental uncertainty from this construct. Perceived environmental uncertainty may vary between firms (buyers and sellers) even though the environmental uncertainty (alternatively objective environmental uncertainty) in the industry is the same for all firms. Differences in perceived uncertainty arise due to the asymmetric expectations between the firms about the future. Such asymmetric expectations may or may not arise due to asymmetric information. However, it is likely that asymmetric information may affect the formation of asymmetric expectations and consequently different levels of perceived uncertainty. In this paper, the discussion is about the environmental uncertainty that is common to all firms in an industry. It is the environment as it exists. 106 decrease in environmental uncertainty renders uncertainty irrelevant to the decision to divest, indicating loss of ‘put’ option value. Some secondary, yet important, results that emerged in the cross-sectional and longitudinal analyses were the following:
1. Information asymmetry was not significant to the decision to divest in
environments of increasing uncertainty.
2. Information asymmetry had a significant negative relationship with the
decision to divest in environments of decreasing uncertainty.
Further, the second essay shows that when the decision to divest and the decision to divest through a particular mode (staged versus complete divestment) are studied together:
1. uncertainty had a significant negative relationship with the decision to divest
whereas it had no relationship with the divestment mode choice.
2. information asymmetry had no relationship with the decision to divest,
whereas it had a significant negative relationship with the divestment mode
choice.
These results indicate that if divestment happens only under conditions of relatively less uncertainty (following the first essay), then information asymmetry concerns, that impact proceeds from the divestment, will be the primary drivers of divestment mode choice decisions. Together, these results give an impression that, perhaps, under conditions of high uncertainty, information asymmetry about true value of assets may not actually matter.
107 In order to ascertain that uncertainty and options considerations may dominate information asymmetry considerations, it is important to develop arguments and test hypotheses for the following:
i) under conditions of increasing uncertainty, whether an increase or a decrease
in information asymmetry would alter the impact of information asymmetry
on the decision to divest.
ii) under conditions of decreasing uncertainty, whether an increase or a decrease
in information asymmetry would alter the impact of information asymmetry
on the decision to divest.
These relationships help in understanding the dynamic interactions between uncertainty and information asymmetry in the context of divestments. Further, the conditions under which uncertainty and information asymmetry may be more important or less important can be understood.
4.2.2 The case of increasing environmental uncertainty and the impact of information asymmetry Divestments are analogous to ‘put’ options (Dixit and Pindyck, 1995, first essay).
This is because when the conditions surrounding divestments are highly uncertain, the true value of business units may not be known, and their current value may not be a good indicator of future value. Divesting business units under such conditions could lead to several costly-to-reverse losses, and also prevent firms from benefiting from emerging opportunities. Highly valuable tangible and intangible assets, for eg., specialized human skills and strong brand image etc., which have been accumulated over long periods of
108 time, may be lost with a divestment. If market conditions were to turn favorable and a firm wishes to re-enter the particular business, such assets may be difficult to regain or rebuild once lost (Dierickx and Cool, 1989). Even without such losses, divestments may still not be desirable under conditions of high uncertainty. It may be very important to just be in the business in order to have access to critical information. Exit may actually provide an opportunity for competitors to penetrate and firms may get locked-out of their market positions.
In view of all the above reasons, it might be worthwhile to wait rather than exit a business under conditions of high uncertainty i.e., treat divestments as ‘put’ options and deal with them flexibly as uncertainty unravels. The value of such ‘put’ options will be higher when the uncertainty about the value of the underlying businesses is higher
(Myers and Majd, 1990).
When uncertainty in a business unit’s environment increases, it becomes harder predict the true value of the business unit—i.e., the true value is unknown. This will be the likely case for both insiders and outsiders, to the extent that environmental uncertainty is objective and is common knowledge to both. Even if insiders have any private information about the value of the business unit, that value may not be correct anyway because it is highly uncertain. In such circumstances, whether or not there is information asymmetry about the ‘true’ value of the business unit should not matter to the decision to divest.
Further, since information asymmetry is, by itself, not likely to be relevant to the decision to divest, an increase in information asymmetry may not necessarily make it any 109 more important to that decision. The case of decreasing information asymmetry need not be any different. This is because a decrease in information asymmetry would only weaken the already non-existent relationship between information asymmetry and the decision to divest. Figure 4.1 shows these relationships. Therefore,
H1: Under conditions of increasing uncertainty, information
asymmetry will be insignificant to the decision to divest regardless
of whether it has increased or decreased.
4.2.3 The case of decreasing environmental uncertainty and the impact of information asymmetry On the other hand, when uncertainty in the external environment decreases, uncertainty will be less relevant to the decision to divest (first essay). This is because the option value in deferring divestment of a business unit weakens with decreasing uncertainty. And, with decrease in uncertainty, the value of a business becomes known.
However, a firm’s insiders may know more about the ‘true’ value of a business unit than outsiders (Myers and Majluf, 1984). This is even more likely when business units are part of large parent portfolios, because outsiders (buyers) may not be able to understand a particular business unit’s contribution to the overall parent’s portfolio of businesses—i.e., its linkages with other businesses and the synergies between them—as much as insiders
(sellers) do (Lang and Stulz, 1994). Therefore, buyers are more likely to distrust the value quoted by sellers and discount it. This reduces sellers’ incentive to divest due to increased adverse selection costs that are likely because of information asymmetry
(Akerlof, 1970; Myers and Majluf, 1984).
110 In these circumstances, if information asymmetry increases, these adverse selection problems get accentuated. On the other hand, if information asymmetry decreases, it is more likely that buyers will be able to assess the true value of the business units reasonably well. Sellers will have an incentive to divest the business units because of reduced adverse selection costs and better chances of realizing the true value for their units. Therefore,
H2: Under conditions of decreasing uncertainty, information asymmetry will be
more strongly negatively related to the decision to divest when information
asymmetry increases than when it decreases.
4.3 Method
The probability that a firm engages in a divestment decision is estimated by a probit. In this method, the decision of the ith firm to divest or not divest depends on an unobservable utility index I i that is determined by explanatory variables X ij in such a way that the larger the value of the index I i, the greater the likelihood that the firm engages in a divestiture. This index I i can be expressed as
I i = β0+ βj X ij where X ij refers to the different independent variables in the model, with subscript ‘i’ referring to the particular firm in question and subscript j=1…n refer to the independent or explanatory variables in the model.
This index is related to the actual decision to divest as follows. If Y=1 for a firm engaging in divestiture of a business unit and Y=0 for a firm that does not divest at all, it can be assumed that there is a critical threshold for each firm. I i * such that if I i > I i* then 111 the firm divests and otherwise it will not. This threshold I i* is not observable. However, it can be assumed to be normally distributed with the same mean and variance. With this assumption of normality, the probability that I i* is less than or equal to Ii can be computed using the standardized normal CDF as
-t2/2 Pi=Pr(Y=1)=Pr(I i*<= I i)=F(I i)= 1/ √(2 Π) ∫ e dt where the limits on integration run from (-∞, I i] or alternatively (-∞, β0+ βj X ij ] and ‘t’ is
- a standardized normal variable i.e, t~N(0,1). Ii is obtained by taking the inverse I i =F
1 -1 (I i)= F (P i)= β0+ βj X ij . For the purposes of estimation, this normit is converted into a probit and I i estimated using
Ii= β0+ βj X ij + u i
56 where u i is the stochastic disturbance term. The model for estimating the divestment decision for the ith firm can therefore be written as
Ii = β0+ β1*segment industry information asymmetry+ β3* segment
industry uncertainty+ β4-12 * controls+ u
Testing hypotheses 1 and 2 involves a comparison of the effect of particular covariates, segment industry information asymmetry and uncertainty in a segment’s environment, between groups. Comparing coefficients between groups is quite straight- forward in the case of linear regressions. The groups can be modeled together in a single regression, using dummy variables to distinguish between groups, and coefficients can be compared.
56 The probit estimation procedure is adapted from Gujarati, D. N. 1998. Basic Econometrics, Chapter 15, pg. 491-493. 112 However, in case of non-linear models such as probit, this procedure is not appropriate at least for two reasons: 1. Dummy interactions by themselves make interpretation difficult and 2. Comparing coefficients needs the assumption that each group has the same residual variation. Comparing two groups even when there are minute differences in residual variances can lead to incorrect inferences. Therefore, it is important to detect and correct for residual variation. Allison (1999) suggested methods to deal with residual variation. Hoetker (2007b) has shown that these methods may also not be very good in detecting residual variation. Regardless of such issues, the most important problem with Allison’s method is with regard to identifying the difference in the impact of a specific covariate. His method requires the very stringent assumption that other coefficients much be equal in both groups. So, in order to test the effect of any particular covariate, all others coefficients need to be assumed to be equal and this is an untestable assumption. That gets into a circular logic and limits the applicability of
Allison’s method (Hoetker, 2007b).
A way to get around these problems of residual variation, that make comparison difficult, is to find methods that do not depend on equality of residual variance. One such way is to model the two groups separately. This will give consistent coefficients and standard errors within each group (It is not hard to see that estimating two equations for the groups using dummy interactions, within a combined model, will also produce similar coefficients. However comparison cannot be as straightforward, as noted above). These coefficients can then be compared for their significance in the respective groups, and inferences can be drawn based on the patterns of significance. This method will be 113 particularly useful when the differences in significance levels are vastly different (not a comparison between p<.11 and p<.09), when the group membership is based on an exogenous variable and, the sample sizes are comparable. This method is limited in that the magnitudes of the coefficients cannot be compared if the coefficients on a particular covariate are significant in both groups in negative (positive) direction. In such cases, there will be no meaningful information beyond the fact that the coefficients are significant in both groups in negative (positive) direction.
Another alternative that does not need the assumption of equal residual variation is comparing the ratios of coefficients of any two coefficient pairs in both groups.
However, in this method 1) even very large differences in ratios may not be statistically significant particularly if one or more terms are estimated with poor precision, and 2) much larger sample sizes may be needed to compare ratios than comparing individual coefficients. Also, theoretically one needs to draw hypotheses that involve comparing ratios of coefficients.
Comparing hypotheses that are in terms of ratios of coefficients on covariates is not the purpose of this study. Therefore, comparing the coefficients between two samples by modeling them separately was more appealing.57
In this study, the membership of the sub-samples depends on exogenous variables such as whether industry uncertainty has changed from a previous period and whether industry information asymmetry changed from a previous period. Hence, there are no
57 Interested readers are referred to Hoetker’s working paper (2007b) on these matters. 114 selection concerns here. Also, the samples are of each group are sufficiently large. This made it worthwhile to model the two groups separately and compare the significance of variables.
4.3.1 Data
All forms of divestments—sell-offs, spinoffs, equity carve-outs were considered for analysis. Preliminary data were obtained from Security Data Corporation’s (SDC)
Mergers and Acquisitions and New Issues databases. The sample has been obtained after cleaning the initial list. The data filters used (See Appendix) are those that are standard in the finance literature (Chen and Guo, 2005; Powers, 2003). The sample period was from January 1980 to December 2003.
Only completed deals have been included in this list because intentions that do not materialize later on cannot be considered as decisions. Deals with no match on
Compustat were also deleted from the list. This was to ensure adequate accounting information availability for analysis. Also, since this analysis required data at the segment level, only those companies with corresponding matches in the Compustat segment database were included.
Only the first decision on a particular segment was taken into account (i.e., if a segment was found to be both in the spin-off/carve-out and sell-offs datasets, the first of the deals was taken into account to avoid any confounding factors affecting the follow-up decision). Also, only parent firms which are in the manufacturing and related categories
(as indicated in the Robins and Wiersema (1995) grouping), other than utilities (SIC
115 codes 4000-4999) were included.58 Utilities are in the regulated industry category, and therefore it would not be appropriate to treat them along with the other industrial
58 A primary SIC code is assigned by Standard & Poor’s to each company in the COMPUSTAT databases according to its primary business activity (as determined by revenues). Therefore, a company’s primary SIC code may change depending on the change in the proportion of sales contributed by particular segments. When this change is affected retrospectively, it is possible that there could be differences between what has been the primary SIC code of the firm in a particular year and the one that has been assigned retrospectively. Segments, on the other hand, are identified based on the primary and secondary products and retain their SIC codes. Segments SIC codes could change when they are merged with other segments and reorganized. However, there is no evidence that these changes in SIC codes are retrospective. Therefore, segment SIC codes are more reliable. Impact on the parent’s portfolio relatedness measure: There is NO impact of a possibly different historic SIC code for the parent firm on the calculation of the parent’s portfolio relatedness measure and the segment’s relatedness measure. This because parent SIC code does not enter the calculation of these measures at all. There are two measures of parent’s portfolio relatedness that have been calculated. Their calculations are detailed below: 1. The Robins and Wiersema (1995) measure of firm’s portfolio resource based relatedness: According to Robins and Wiersema, for each combination of two different industry categories ‘i’ and ‘j’ in a firm’s portfolio, the sales-weighted measure of interrelationship Rij is given by R ij =P i r ij + P j r ij , where Pi= percentage of sales in industry category i and Pj= percentage of sales in industry category j. These weighted measures of similarity between paired industries (Rij) are summed over all combinations of two industries that could be formed in a business portfolio of the firm resulting in an aggregate index of interrelationship of the businesses of the firm Mk= Σ Rij= Σ rij(Pki+ Pkj), where i and j represent any two different industries in which the firm k is active. The authors then introduce a correction factor to avoid double counting. This is the parent’s total portfolio interrelatedness. Segment’s interrelatedness is measured as the sum of the segment’s relatedness with other segments in the parent’s portfolio. 2. The measure of related entropy in a parent’s portfolio that reflects the level of diversification. The basic entropy index was computed by Jacquemin and Berry (1979) as E= ΣPi* ln(1/Pi). In this equation, E signifies the entropy measure and Pi is the proportion of a firm’s sale in SIC industry i. This is typically treated as a measure of total diversification when calculated at the 2-digit SIC level. The unrelated diversification is computed using the proportions at the 2-digit SIC level and subtracted from the 4-digit level to obtain a measure of ‘related’ diversification in the portfolio. As can be seen, these calculations involve the segment level SIC code information and not the parent SIC code. Therefore, changes in the historic SIC codes of a parent firm has no impact on the computation of these measures. Other measures such as uncertainty and information asymmetry are calculated at the segment industry relatedness category level. The industry relatedness category that a firm belongs to is determined by its SIC code. However, the uncertainty and information asymmetry measures are averages of all firms in the industry categories. Therefore, if firms do not get included in a category they belong to, it is also possible that some firms get included where they do not belong to and, on average, the differences may get compensated. The effect will be expected to be similar with regard to identifying the control firms that have been chosen based on the parent’s SIC code. Therefore, the potential source of error due to differences in the historic SIC codes is not expected to be of major concern. In any case, historic SIC codes have been verified for about 15 per cent of the sample. In the sub- sample, the industry relatedness category was different only for about 13 per cent of the firms. The movements from one relatedness category to another were not systematic. Therefore, it is not a major concern. 116 groupings. Only domestic parent firms were included to eliminate influences of differing institutional regimes. American Depository Receipts (ADRs), were excluded for a similar reason. Limited partnerships that operate under different legal rules were also excluded.
Spinoff and carve-out divisions fulfilled following additional criteria: 1. these transactions were confirmed by searching SEC filings and news articles on Factiva, 2. spinoff or carved out company was not organized as a limited partnership (to ensure that mere reorganization to be under different legal rules was not the objective) , 3. information related to spinoff or carve-out was available on Compustat for the closest financial year before spinoff or carve-out (to ensure that these separated entities could be matched back to obtain operating performance before being separated) and, they were also listed on the CRSP (were indeed being traded as independent entities), 4. they were not tracking stock deals (since tracking stocks do not involve any separation of ownership of the unit from the parent and are issued mainly to ensure that the accounting performance of a particular unit can be monitored more closely while it is still within a parent conglomerate), 5. incorporated in the United States (to eliminate any institutional advantages from separation), 6. were not owned by multiple parents, 7. were not a direct result of mergers (to prevent other confounding factors).
Sell-offs are transactions between two parent companies. Thus, the parent companies are not required to disclose prior operating performance of the units/divisions being sold. Since this study requires operating data at divisional level for the sell-off division prior to the sale, sell-offs were identified by analyzing segment data in 117 Compustat following Schlingemann, Stulz and Walkling (2000). All those segments of companies which were listed in year ‘t’ but were not listed in year ‘t+1’ were identified.
All such segments need not actually be segments that were sold off, since a company could cease to report a particular segment’s information for a number of other reasons eg., reorganization within the company.
Therefore, to actually isolate the sell-offs, these segments (reported in ‘t’ but not
‘t+1’) were matched against the SDC’s completed sell-offs information. These matched deals were then checked manually for corresponding newswire items in the Lexis-Nexis database.
A sample of matched firms that have not engaged in a divestiture was also obtained. Matched parents were found using the parent’s total assets as a matching criterion within the same relationship category matched at the same four-digit SIC code
(Chen and Guo, 2005; Krishnaswami and Subrahmaniam, 1999) 59 . The time period used for selecting the matched sample is the same as that for the divesting firms.60 There were some cases where there was no match found but still the original divesting firm was retained in the sample. There were 1115 matched parent-segments. Together, the final sample had 182 sell-offs, 102 spin-offs and carve-outs, and 598 non-divested firm segments with no missing values for any of the required variables. 59 Parent firm size and industry grouping at four digit level have been used as the matching criterion following common practice (Chen and Guo, 2005; Krishnaswami and Subramaniam,1999). Since parents belong to manufacturing and related industries, size measured by parent assets is more appropriate. Both the first best match and the next best matched parent firms that have not engaged in a divestiture were chosen in order to ensure a close to 2:1 ratio of matched versus divesting firms. 60 Matched firms were not necessarily chosen from the same year as the divesting firm, though matching firms in the same year were not eliminated by design. 118 The data collection procedure clearly restricts the analysis to large publicly traded firms with listed segments. This leaves out several small firms not listed in the databases used. However, inclusion of small firms would only strengthen the results because small firms are susceptible to the influences of uncertainty more than larger firms. This study is therefore a conservative test. Also, large private firms are not in this list but such firms are relatively few anyway. Therefore the results are not expected to change with or without including them.
4.3.2 Other data sources
The data to calculate uncertainty measure were obtained from the Center for
Research in Security Prices (CRSP) monthly stock price database. Data for calculating the information asymmetry measure were obtained from the summary statistics on the
Institutional Brokers Estimates System (IBES) database. Block-holder data has been obtained from Thomson Financial’s Spectrum Institutional Stockholding database.
4.3.3 Variables and measurement
Dependent variable
The dependent variable is a dichotomous variable where the divesting firm has been represented by ‘1’ and a non-divesting firm or the matching firm has been represented by ‘0’.
Control variables
The control variables for this study come from a wide variety of explanations and theories have been used to understand divestments. They are detailed below, briefly, along with the measures used. 119 Parent and segment level performance
By far, the most dominant explanations for divestments were based on parent and segment/business unit level performance. Divestments were seen as means to restore corporate efficiency. Therefore, parent performance has been argued to positively influence divestment decisions (Cho and Cohen, 1997; Harrigan, 1981, 1982; Jain, 1985;
Montgomery and Thomas, 1988). Also, poor performance at the business unit level has been found to be a key determinant for divestments (Vignola, 1974; Patton and Duhaime,
1978; Ravenscraft and Scherer, 1987; Chang, 1996; Duhaime and Grant, 1984; Hamilton and Chow, 1993; Hitt et al., 1996; Hoskisson & Johnson, 1992).
In this study, parent and segment level performance were controlled for using the return-on-asset measure. This measure has been considered more appropriate for this study since firms operating in manufacturing and related industries are asset intensive.
Return on assets has been computed as operating income before depreciation over total assets (Powers, 2001).61
Agency theory
Another major explanation for divestment decisions was based on agency theory.
Managerial self-interest and poor governance mechanisms were argued to adversely influence the decisions to divest (Bethel and Leibeskind, 1993; Finkelstein and
Hambrick, 1989; Jensen and Murphy, 1990). Strong governance mechanisms such as large block-holder ownership, higher number of outsiders on boards etc., were shown to 61 Return on sales measure has been computed as operating income before depreciation over net sales and was used to check the robustness of results to alternative performance measures. Accounting measures of performance have been chosen as opposed to market measures since market measures are not available at the segment level. 120 reduce such tendencies and favorably influence divestments (Bethel and Liebeskind,
1993; Hoskisson, et al., 1994; Sanders, 2001).
While there have been arguments for and against the monitoring efficacy of outside board members (Baysinger and Hoskisson, 1990), block-holders seem to have an incentive to monitor firms more closely. Therefore, agency explanations of divestments were controlled in this study by using the level of stock held by block-holders as a proxy.62 The normally accepted definition of block-holders as those who control more than 5 per cent of the firm shares has been used (Bethel and Liebeskind, 1993).
Transaction costs economics and behavioral uncertainty
Transaction costs economics was yet another theory applied to understand divestments. The core of the TCE argument is that, at any point in time, uncertainty about the behavior of partners to a transaction will have an impact on whether or not a transaction is internalized. Transactions with higher behavioral uncertainty that can be efficiently managed by ‘fiat’ would be internalized, and those that involve less behavioral uncertainty for which ‘fiat’ would be relatively costly, will not be integrated or will be divested (Hill and Hoskisson, 1987; Jones and Hill, 1988; Markides, 1992).
A measure of behavioral uncertainty should capture the magnitude of the problems of coordination arising due to transaction specific investments in recurrent transactions, particularly under conditions of uncertainty (Williamson, 1979; 1985). It
62 Measuring block holder ownership only by considering institutional owners could cause a slight bias because individual block-holders are not included. 121 would thus be an interaction term between transaction specific investments, frequency of transactions, and the levels of uncertainty. Unfortunately, several TCE studies have captured only one of these three dimensions of behavioral uncertainty and, the divestment literature is no exception (David and Han, 2004; Carter and Hodgson, 2006).
In this study, it was important to capture the behavioral uncertainty between the business unit and the parent firm in order to assess the transaction costs considerations in the divestment decisions. If a business unit is very closely related to the other business units in the parent firm, one can expect greater need for coordination between the business unit and the rest of the firm. For example, if the business unit is a part of an integrated production process, there are likely to be more transaction specific investments between the parent and the business unit. Further, the frequency of interactions for coordination with other business units that are a part of the same process can also be expected to be higher. On the other hand, if the business unit is a stand-alone and is less connected with other units, there will be less need for coordination. Interactions with other units will be infrequent and also there will be fewer transaction specific investments.
Therefore, a business unit’s relatedness to the parent company can be a proxy for the level of transaction-specific investments and frequency of interactions between the business unit and the parent firm. The interaction of the business unit’s relatedness with parent firm and, the level of uncertainty in the segment’s industry can thus proxy the behavioral uncertainty. Controlling for this variable became important in order to
122 separate the real options explanation (based on industry uncertainty) from that of TCE
(behavioral uncertainty).
Parent and segment relatedness
Further, the level of relatedness between various businesses within a parent firm’s portfolio and, the level of relatedness of the business unit to be divested with the parent firm were the other important determinants of divestment decisions. Greater levels of relatedness, in general, are found to be negatively related to divestments (Hoskisson,
Johnson and Moesel, 1994; Chang and Singh, 1999).
Parent’s portfolio relatedness was measured using related entropy (Jacquemin and
Berry, 1979) 63 since parent’s resource-based relatedness was highly correlated with several other variables of the study. In the present case, the relationship between the particular division in question (the division divested via spin-off, carve-out or sell-off) and the other divisions within the parent company was important. Therefore, a sales- weighted measure of interrelationship of the focal division with the other divisions in the parent firm was calculated following Robins and Wiersema (1995) 64 and then summed up to get the extent of resource-based relatedness of the segment with the parent firm.
63 The basic entropy index was computed by Jacquemin and Berry (1979) as E= ΣPi* ln(1/Pi). In this equation, E signifies the entropy measure and Pi is the proportion of a firm’s sale in SIC industry i. This is typically treated as a measure of total diversification when calculated at the 2-digit SIC level. The unrelated diversification is computed using the proportions at the 2-digit SIC level and subtracted from the 4-digit level to obtain a measure of ‘related’ diversification in the portfolio. 64 Robins and Wiersema (1995) developed a resource-based relationship index to measure a firm’s overall portfolio interrelatedness. According to Robins and Wiersema, for each combination of two different industry categories ‘i’ and ‘j’ in a firm’s portfolio, the sales-weighted measure of interrelationship Rij is given by R ij =P i r ij + P j r ij , where Pi= percentage of sales in industry category i and Pj= percentage of sales in industry category j. These weighted measures of similarity between paired industries (Rij) are summed over all combinations of two industries that could be formed in a business portfolio of the firm resulting in an aggregate index of interrelationship of the businesses of the firm Mk= Σ Rij= Σ rij(Pki+ Pkj), where i and 123 Other controls
Other important variables that needed to be controlled, following previous literature, were parent debt position (leverage), firm size, and segment size. Parent’s debt position (leverage) was measured as total long term debt over total common equity (lev1) and long term debt over market value of equity (lev4), again to check the robustness of results to alternative measures (Chang and Singh, 1999). Firm size was measured as log of total assets and segment size was measured as the proportion of the parent’s total assets that are in the segment i.e., segment assets/parent’s total assets (Duhaime and
Baird, 1987; Bergh, 1992). 65 Again, an asset based measure was considered more appropriate due to the nature of the firms in this study.
Parent’s valuation was measured as market value to net sales. Parent’s market value was computed as the price at the end of the fiscal year multiplied by the total outstanding common stock. Another measure of valuation was the parent’s market-to- book ratio.66 Also, a measure of growth options in other business units/segments of the firm was calculated using the market-to-book value of the median firm in the industry to
j represent any two different industries in which the firm k is active. The authors then introduce a correction factor to avoid double counting. This is the parent’s total portfolio interrelatedness. Segment’s interrelatedness is measured as the sum of the segment’s relatedness with other segment’s in the parent’s portfolio. The parent’s resource-based portfolio interrelatedness measure was very highly correlated with segment relatedness calculated using the resource-based measure. Therefore, the entropy measure was used for parent’s portfolio relatedness. Results on key variables are robust when the segment relatedness is dropped and the parent’s resource based portfolio measure was used. Parent’s resource-based relatedness measure was significant itself in negative direction as expected. 65 Controlling for the total number of segments was considered. However, just that number would not provide any more information than what is obtained from firm size and the measures of relatedness. 66 Market-to-book ratio was obtained by summing market value of shares, preferred stock, parent’s long term debt, total current liabilities net of cash and divided by the parent’s operating income before depreciation. Parent’s market value was computed as the price as the end of the fiscal year multiplied by the total outstanding common stock. Another measure used was the parent’s market-to-sales ratio. 124 which the business segments belonged to. This median value was weighted with the segment’s proportion of the parent firm’s sales. The weighted sum was taken as a measure of the total growth options in other segments of the firm. All variables were taken as of the financial year preceding the event and the level of information asymmetry was considered with a one-period lag.
Independent variables
Information asymmetry
Information asymmetry about the value of the business unit/segment is a key independent variable. Generally, information asymmetry about a parent firm has been used as a measure in prior literature. However, if a business unit is being divested, whether the unit will be properly valued will depend on whether or not there is an appropriate comparison available in the segment’s industry. Therefore, the relevant measure would be the level of information asymmetry in the business unit’s industry.
Information asymmetry in the business unit’s industry was measured as the mean value of the standard deviation in analysts’ forecasts from IBES for the segment’s industry grouping in a particular year (Krishnaswami and Subramaniam, 1999). This variable has also been lagged by one year.
Uncertainty
Uncertainty is the other key independent variable in this model and a time-varying estimate of business unit’s/segment’s environmental uncertainty was needed. It is a common practice to quantify the construct of uncertainty by calculating the variance of indicators such as stock price, GDP or sales over time. Such approaches fail to account 125 for trends in data that can increase the measured variance while actually not being an element of uncertainty if they were predictable. Also, such approaches do not accommodate the possibility of variances being heteroskedastic that is very characteristic of time series in general.
Following Folta and O’Brien (2004), Carruth et al., (2000), the Generalized autoregressive conditional heteroskedasticity models (GARCH) were used and, the conditional variances generated were used as a measure of uncertainty (Bollerslev, 1986;
Engle, 2001). In particular, the GARCH-M (1,1) models were run on value-weighted industry portfolio 67 returns that were developed from monthly stock returns (adjusted for dividends and splits) for all firms in the CRSP database from 1950-2004.68
The GARCH-M model can be written as follows:
rt = α+γht-1+ρrt-1+δε t-1+εt
2 h t = κ+ρ1 h t-1+ δε t-1
εt = sqrt( ht zt) and zt ~ N(0,1)
This model represents the generalized autoregressive conditional heteroskedasticity in-mean specification, GARCH-M, with ARMA (1, 1) in the mean equation. ‘ εt’ representing the error term, is conditionally normally distributed and serially uncorrelated. ‘h t’, the conditional variance, is a linear function of the past period’s
67 All industry groupings are based on Robins and Wiersema (1995) classification. 68 The data has been checked for white noise using the Portmonteau’s Q-test, the correlograms and Bartlett’s periodogram based white noise test. Where there was evidence of white noise, an ARIMA (0,0,1) term was introduced to mitigate the situation and the tests were repeated to confirm white noise. 126 2 69 squared errors, εt-1 , and the last period’s conditional variance, h t-1, i.e., GARCH(1,1).
The ARMA (1,1) in the mean equation implies that the conditional returns in this model are a linear function of the last period’s conditional variance, past conditional returns and past disturbance. Under this rich specification, volatility can change over time and expected returns are a function of volatility as well as past returns. The monthly conditional variances were averaged to obtain annual figures. The variable business unit’s environmental uncertainty is computed as the square root of the average yearly conditional variance. The lagged variable on uncertainty was used by considering the uncertainty in the year preceding the closest financial year to the event date.
4.4 Analysis and results
Summary statistics for divesting and non-divesting parent-segment combinations are in Table 4.1. These statistics show, on average, the parents that have divested a business segment seem to have higher institutional block-holdings, greater levels of diversification in their portfolios, higher leverage ratios and larger firm sizes as compared to non-divesting firms. Also, divested segments are larger in size than non-divested segments.
On the other hand, parents that have divested a business segment seem to have lower uncertainty in the segment industry, poorer performance, lower behavioral uncertainty between the parent and segments, lower information asymmetry in the 69 In fitting time series models, the simplest models are fitted first. Higher models with more number of autoregressive terms call for greater number of parameters. They are not used unless there is a good reason to believe that the model specified is not capturing the process sufficiently. Here, the GARCH-M(1,1) captures the underlying process well, as in case of several other studies cited, and therefore there was no need to use an alternative specification. 127 segment industry, and lower parent firm valuations as compared to non-divesting firms.
Further, the segments that have been divested seem to be less related to their parent firms’ than those segments that have not been divested. Some of these differences are significant whereas others are not. The t-tests for significance of differences between means are also reported in table 4.1.
The correlation matrix in Table 4.2 shows no major problems of multi-collinearity among the variables except for behavioral uncertainty and segment relatedness (0.9646) indicated by * in the table. 70 The result on the key variable, segment’s environmental uncertainty, holds with and without the measure of behavioral uncertainty. Therefore the results or interpretation of this study do not change with the presence or absence of this variable.
Tables 4.3 and 4.4 show results from probit and marginal effects, respectively, for firms experiencing an increase in uncertainty in the business unit’s environment. Tables
4.5 and 4.6 show these results for firms experiencing decrease in uncertainty in the business unit’s environments. 71 Only marginal effects have been interpreted.
In each table model 2 is a variation on model 1 with a different leverage measure.
Models 3 and 4 are variants of models 1 and 2, respectively, controlling for the transaction costs economics explanation by introducing the behavioral uncertainty 70 Parent’s resource-based portfolio relatedness measure is significantly correlated with several other variables, including segment’s relatedness. Therefore, this variable was replaced by the parent’s related entropy measure of portfolio relatedness to avoid the problem. Even parent performance and segment performance are correlated upto 0.66. However, results are robust to alternative measures of parent and segment performance. 71 For non-linear models such as probit, it is only the marginal effects that are relevant for interpretation (Train, 1986, Hoetker, 2007a). Therefore, only the marginal effects have been interpreted. All marginal effects are at the mean. 128 variable.72 Models 5-8 are counterparts of models 1-4 with controls for growth options in other businesses.
It can be seen from Tables 4.4 and 4.6 that the factors that influence the decision to divest may be different depending on whether environmental uncertainty and information asymmetry in a segment’s industry are increasing or decreasing. For example, under conditions of increasing uncertainty, segment’s relatedness was insignificant when information asymmetry increased. However, it was negative and significant to the decision to divest under conditions of decreasing information asymmetry. Similarly, parent’s valuation was a significant factor under conditions of increasing information asymmetry whereas it was not significant when information asymmetry decreased. These results indicate that the importance of different factors to the decision to divest differ under different circumstances.
Table 4.4 shows the marginal effects for testing hypothesis 1—i.e., for the set of firms that experienced an increase in uncertainty in business units’ environments. The results show that the coefficient on information asymmetry was not significant to the decision to divest regardless of whether information asymmetry increased or decreased.
Thus, hypothesis 1 has been supported.
Further, table 4.6 shows the marginal effects for testing hypothesis 2—i.e., for the set of firms that experienced a decrease in uncertainty in business units’ environments.
These results show that the coefficient on information asymmetry was significant and 72 Conditioning diagnostics are important in the presence of polynomial terms in the regression equation or when there are interaction terms. The model here has an interaction term, but the results hold regardless of the presence of the interaction term. Also, the standard errors are not inflated much with the introduction of new variables. Together, these suggest there are no serious multi-collinearity problems in this data. 129 negatively related with the decision to divest, when information asymmetry increased.
On the other hand, the coefficient of information asymmetry was significant and positively related with the decision to divest, when information asymmetry decreased.
Since both these coefficients are of different signs, they can be compared. These results thus support hypothesis 2 that when uncertainty decreases, information asymmetry has a stronger negative impact on the decision to divest when it increases than when it decreases.
In order to detect whether the degree of change in uncertainty and the degree of change in information asymmetry matters for information asymmetry concerns to surface as important for the decision to divest, the sample of firms that experienced a decrease in uncertainty have been split by the median into two sub-samples. The firms were separated into those that experienced small decreases in uncertainty and those that experienced large decreases in uncertainty. Probit analyses were repeated for these two sets of firms. Results show that for the set of firms that experienced small decreases in uncertainty, information asymmetry was not significant to the decision to divest. On the other hand, for the set of firms that experienced large decreases in uncertainty, information asymmetry had a significant negative relationship with the decision to divest.
Similarly, the sample was also split by the degree of change in information asymmetry. For the set of firms that experienced a decrease in uncertainty, whether information asymmetry increases were large or small were considered. Again, for the set of firms that experienced large increases in information asymmetry, information
130 asymmetry had a significant negative association with the decision to divest. On the other hand, information asymmetry was not significant when the increases were small.
A further split has been done in the set of firms experiencing decreases in uncertainty whether they experienced small or large decreases in information asymmetry.
The marginal effects of information asymmetry were positive in these samples.
Together, these results show that within the sample of firms experiencing decreases in uncertainty, the impact of information asymmetry varies between the firms experiencing large increases in information asymmetry as compared to those experiencing large decreases in information asymmetry. This further shows the non- linear nature of this relationship.
4.5 Discussion and conclusion
This study examined the conditions under which information asymmetry matters for the decision to divest. The core idea of the study has been that information asymmetry concerns should be less relevant when the external environment itself is highly uncertain. This relationship has been explored under conditions of increasing and decreasing environmental uncertainty and increase and decrease in information asymmetry in a business unit’s environment.
This essay differs from the first essay in the following manner. The focus of the first essay was on investigating the implications of real options theory in the context of divestments. It involved investigating the relationship between uncertainty and the decision to divest and how an increase or decrease in uncertainty affected the impact of uncertainty on the decision to divest. While information asymmetry was a control 131 variable, increases or decreases in information asymmetry were not considered and were not of concern in that essay. This essay, on the other hand, is concerned primarily with the impact of information asymmetry when information asymmetry increases or decreases, given that environmental uncertainty has increased or decreased. This essay thus investigates the dynamic interactions between environmental uncertainty and information asymmetry which were not considered in the first essay.
The results of this essay showed that under conditions of increasing environmental uncertainty, information asymmetry was not significant to the decision to divest regardless of whether information asymmetry has increased or decreased. On the other hand, when environmental uncertainty decreased, information asymmetry was strongly negatively related with the decision to divest when information asymmetry increased than when it decreased.
Taken together, these results indicate that information asymmetry concerns may be less relevant, when the value of a business is itself uncertain and such uncertainty continues to increase. However, information asymmetry concerns may surface when uncertainty about the future of a business unit decreases. Further, the extent to which information asymmetry adversely affects the decision to divest differs based on whether information asymmetry has increased or decreased.
These results, therefore, suggest that environmental uncertainty and options considerations may dominate information asymmetry concerns with regard to the decision to divest, under certain circumstances.
132 This study highlights that models including only information asymmetry without taking uncertainty into consideration may not provide a complete picture of the dynamics between the two. It is possible that uncertainty in external environment may dominate information asymmetry concerns in other situations as well. For example, several labor market studies employ models based on information asymmetry (Alvi, 1988;
Dewatripont, 1989). The focus of such studies is on principal-agent problems and principals designing effective contracts to create incentives for agents to reveal true information. However, if the value of such information itself is highly uncertain, it might be interesting to question whether creating incentives to extract ‘truthful’ information is a worthy exercise. There may be several other such applications where it might be interesting to study the impact of uncertainty and information asymmetry in a comprehensive framework.
133 t-test for difference in Firm type Non-Divesting Firms Divesting Firms means Variable Obs Mean Std. Dev. Obs Mean Std. Dev. Uncertainty in business unit’s environment 598 0.06 0.01 284 0.06 0.01Significant Parent ROA 598 0.12 0.17 284 0.08 0.16 Significant Segment ROA 598 0.09 0.25 284 0.03 0.29 Not significant Institutional Block-holder Share 598 10.45 13.71 284 12.13 13.19Significant Behavioral Uncertainty 598 0.02 0.03 284 0.01 0.02Significant Business unit’s industry information asymmetry 598 0.33 2.28 284 0.21 0.28Not significant Parent's Valuation (sales measure) 598 1.95 6.66 284 1.07 1.44Not significant Parent’s Level of Diversification (Related Entropy) 598 0.05 0.18 284 0.09 0.2Significant
Parent's portfolio Resource-based Relatedness 598 0.51 0.47 284 0.3 0.4Significant Business unit's resource-based relatedness 598 0.37 0.55 284 0.19 0.36Significant Parent Debt (lev1) 598 0.19 0.16 284 0.23 0.16Significant Parent Debt (lev4) 598 0.43 0.68 284 0.77 1.9Significant Parent Size 598 8.19 1.98 284 8.38 1.89Not Significant Segment Size 598 0.53 0.39 284 0.57 0.39 significant Table 4.1: Summary statistics for divesting and non-divesting firms
134 Information Uncertainty in asymmetry in Institutional segment's segment block-holder Parent Segment Variable Divesting firm industry industry share performance performance Leverage (lev1) Divesting firm 1 Uncertainty in segment's industry -0.0781 1 Information asymmetry in segment industry -0.0306 -0.0036 1 Institutional blockholder share 0.0578 0.0596 -0.0105 1 Parent performance -0.0983 -0.0406 -0.0133 0.0483 1 Segment performance -0.0981 -0.0272 -0.0094 0.0131 0.6462 1 Leverage (lev1) 0.1069 -0.0114 -0.0085 0.0404 0.0133 -0.0386 1 Leverage (lev4) 0.1303 0.0258 0.0008 0.0635 -0.0459 -0.0997 0.4673 Level of diversification in parent's portfolio (Related entropy) 0.0898 -0.0042 -0.0208 0.1236 0.0294 0.0502 0.056 Segment's resource-based relatedness to parent firm -0.166 0.1008 0.0024 -0.002 0.1119 0.1918 0.0238 Behavioral uncertainty -0.161 0.2541 0.0035 0.0225 0.104 0.193 0.0285 Parent's valuation -0.0741 -0.0189 -0.0056 0.0032 -0.2123 -0.1327 -0.1061 Firm size 0.0469 -0.0089 0.0383 0.0795 0.3475 0.1897 0.132 Segment size 0.0573 -0.0732 -0.005 -0.0219 -0.1776 -0.1342 -0.0871 Growth options in other businesses -0.0717 0.0246 -0.0104 -0.0001 0.0646 -0.0034 0.1403 Table 4.2: Correlation matrix Table 4.2 (continued)
135 Table 4.2 (continued)
Level of diversification Segment's in parent's resource- Growth portfolio based options in Leverage (Related relatedness to Behavioral Parent's Segment other Variable (lev4) entropy) parent firm uncertainty valuation Firm size size businesses Leverage (lev4) 1
Level of diversification in parent's portfolio (Related entropy) -0.026 1
Segment's resource- based relatedness to parent firm -0.006 0.0507 1
Behavioral uncertainty -0.0005 0.0364 0.961* 1
Parent's valuation -0.0959 -0.0575 -0.0911 -0.0857 1 Firm size 0.0819 0.1136 0.2073 0.2077 -0.1572 1
Segment size -0.0663 -0.254 -0.2748 -0.2736 0.1579 -0.3753 1 Growth options in other businesses 0.0311 -0.0341 0.0919 0.09 -0.0185 0.0247 -0.0645 1
136 Variable Name Model1 Model2 Model3 Model4 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry Business unit industry information asymmetry -0.053 0.741 -0.08 1.162 -0.055 0.893 -0.082 1.18 [0.21] [0.64] [0.32] [1.03] [0.22] [0.77] [0.32] [1.04]
Uncertainty in a business unit's environment -17.676*** -1.496 -18.153*** -1.242 -18.959** -12.414 -19.306** -13.161 [2.86] [0.15] [2.92] [0.13] [2.52] [1.19] [2.57] [1.26]
Parent performance -3.246** 0.651 -3.201** 0.268 -3.218** 0.64 -3.173** 0.325 [2.31] [0.46] [2.33] [0.19] [2.29] [0.45] [2.31] [0.23] Segment performance -0.017 -0.331 0.066 -0.354 -0.051 -0.395 0.031 -0.428 [0.02] [0.39] [0.08] [0.42] [0.06] [0.48] [0.04] [0.52] Institutional block-holder share -0.002 0.004 -0.002 0.005 -0.002 0.003 -0.002 0.004
[0.28] [0.51] [0.27] [0.62] [0.32] [0.42] [0.30] [0.50]
Behavioral uncertainty 4.315 73.893** 3.98 79.344** [0.35] [2.01] [0.33] [2.20]
Parent's diversification (entropy measure) 0.638 1.072* 0.671 1.002* 0.651 1.244** 0.683 1.203** [1.15] [1.86] [1.22] [1.71] [1.18] [2.16] [1.25] [2.08] Business unit's relatedness with parent -0.32 -0.769*** -0.333 -0.741*** -0.598 -5.424** -0.59 -5.739** [1.41] [3.11] [1.47] [2.99] [0.69] [2.23] [0.68] [2.41] Parent's leverage (lev1) 0.317 0.886 0.303 0.873 [0.48] [1.33] [0.46] [1.31] Parent's leverage (lev4) -0.086 0.429** -0.086 0.361* [0.44] [2.12] [0.44] [1.81] Parent's valuation (sales measure) -0.204** 0.04 -0.184** -0.012 -0.206** 0.025 -0.185** -0.016 [2.38] [0.47] [2.32] [0.15] [2.37] [0.29] [2.30] [0.20] Firm size 0.059 0.036 0.054 0.027 0.058 0.035 0.053 0.028 [0.93] [0.47] [0.85] [0.37] [0.91] [0.46] [0.83] [0.37] Segment size 0.078 0.46 0.102 0.447 0.075 0.518* 0.099 0.511* [0.28] [1.50] [0.37] [1.48] [0.27] [1.69] [0.36] [1.68]
Constant 0.926 -1.115 0.859 -0.999 1.017 -0.434 0.944 -0.294 [1.28] [1.19] [1.18] [1.08] [1.28] [0.43] [1.18] [0.29] Observations 214 169 214 169 214 169 214 169 LogLik -128.07 -100.11 -128.06 -101.52 -128.01 -97.18 -128.01 -97.93 Pseudo R2 0.12 0.13 0.12 0.12 0.12 0.16 0.12 0.15 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 4.3: Probit results for firms facing an increasing in uncertainty in business unit’s environment Table 4.3 (continued) 137 Table 4.3 (continued)
Variable Name Model5 Model6 Model7 Model8 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry Business unit industry information asymmetry -0.061 0.85 -0.099 1.441 -0.055 0.975 -0.095 1.447 [0.24] [0.71] [0.40] [1.22] [0.22] [0.80] [0.39] [1.19]
Uncertainty in a business unit's environment -27.338*** 1.107 -27.269*** 0.258 -25.357*** -11.848 -25.459*** -13.522 [3.57] [0.11] [3.55] [0.03] [2.69] [1.10] [2.71] [1.28]
Parent performance -5.231*** 0.933 -5.094*** 0.665 -5.286*** 0.704 -5.136*** 0.49 [3.32] [0.60] [3.31] [0.43] [3.33] [0.45] [3.33] [0.32] Segment performance -0.351 -0.873 -0.379 -0.918 -0.286 -0.952 -0.322 -1.001 [0.37] [0.87] [0.39] [0.93] [0.29] [0.99] [0.33] [1.05] Institutional block-holder share -0.013 0.004 -0.012 0.004 -0.012 0.004 -0.012 0.005
[1.61] [0.50] [1.58] [0.57] [1.59] [0.61] [1.55] [0.64]
Behavioral uncertainty -7.231 99.520** -6.693 105.003** [0.53] [2.28] [0.50] [2.47]
Parent's diversification (entropy measure) 0.952* 0.517 0.958* 0.479 0.937* 0.643 0.945* 0.632 [1.80] [0.84] [1.81] [0.77] [1.76] [1.08] [1.78] [1.05] Business unit's relatedness with parent -0.14 -0.749*** -0.157 -0.728*** 0.307 -7.162** 0.257 -7.495*** [0.56] [2.83] [0.63] [2.75] [0.33] [2.46] [0.28] [2.63] Parent's leverage (lev1) 0.156 0.7 0.162 0.577 [0.22] [0.97] [0.23] [0.80] Parent's leverage (lev4) -0.095 0.459** -0.102 0.383* [0.46] [2.18] [0.49] [1.90] Parent's valuation (market measure) 0.061*** 0.001 0.060*** -0.001 0.061*** 0.004 0.060*** 0.002 [4.14] [0.13] [4.19] [0.07] [4.13] [0.36] [4.18] [0.23] Firm size 0.057 0.099 0.055 0.092 0.057 0.105 0.055 0.098 [0.80] [1.20] [0.78] [1.12] [0.80] [1.27] [0.78] [1.19] Segment size -0.413 0.102 -0.395 0.041 -0.412 0.1 -0.394 0.05 [1.26] [0.30] [1.22] [0.12] [1.27] [0.29] [1.22] [0.15] Growth options in other businesses -0.028** -0.035*** -0.029** -0.035*** -0.027** -0.037*** -0.028** -0.037*** [2.14] [2.80] [2.27] [2.82] [2.09] [2.85] [2.23] [2.88] Constant 1.590* -1.351 1.527* -1.168 1.473 -0.559 1.417 -0.338 [1.88] [1.38] [1.83] [1.20] [1.61] [0.54] [1.56] [0.33] Observations 210 165 210 165 210 165 210 165 LogLik -109.57 -91 -109.65 -93.01 -109.47 -86.87 -109.56 -88.19 Pseudo R2 0.24 0.2 0.24 0.18 0.24 0.24 0.24 0.23 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1%
138 Variable Name Model 1 Model 2 Model 3 Model 4 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry Business unit industry information asymmetry -0.02 0.275 -0.03 0.433 -0.021 0.324 -0.031 0.429 [0.21] [0.64] [0.32] [1.03] [0.22] [0.77] [0.32] [1.04] Uncertainty in business unit's environment -6.611*** -0.555 -6.791*** -0.463 -7.097** -4.508 -7.229** -4.789 [2.86] [0.15] [2.92] [0.13] [2.52] [1.19] [2.57] [1.26] Parent performance -1.214** 0.241 -1.198** 0.1 -1.205** 0.233 -1.188** 0.118 [2.31] [0.46] [2.33] [0.19] [2.29] [0.45] [2.31] [0.23] Segment performance -0.006 -0.123 0.025 -0.132 -0.019 -0.143 0.012 -0.156 [0.02] [0.39] [0.08] [0.42] [0.06] [0.48] [0.04] [0.52] Institutional blockholder share -0.001 0.001 -0.001 0.002 -0.001 0.001 -0.001 0.001 [0.28] [0.51] [0.27] [0.62] [0.32] [0.42] [0.30] [0.50] Behavioral Uncertainty 1.615 26.835** 1.49 28.869** [0.35] [2.01] [0.33] [2.20] Parent's diversification (entropy measure) 0.238 0.398* 0.251 0.373* 0.244 0.452** 0.256 0.438** [1.15] [1.86] [1.22] [1.71] [1.18] [2.16] [1.25] [2.08] Business unit's relatedness with parent -0.12 -0.285*** -0.125 -0.276*** -0.224 -1.970** -0.221 -2.088** [1.41] [3.11] [1.47] [2.99] [0.69] [2.23] [0.68] [2.41] Leverage (lev1) 0.118 0.33 0.114 0.318 [0.48] [1.33] [0.46] [1.31] Leverage (lev4) -0.032 0.159** -0.032 0.131* [0.44] [2.12] [0.44] [1.81] Parent valuation (sales measure) -0.076** 0.015 -0.069** -0.004 -0.077** 0.009 -0.069** -0.006 [2.38] [0.47] [2.32] [0.15] [2.37] [0.29] [2.30] [0.20] Firm size 0.022 0.013 0.02 0.01 0.022 0.013 0.02 0.01 [0.93] [0.47] [0.85] [0.37] [0.91] [0.46] [0.83] [0.37] Segment size 0.029 0.171 0.038 0.167 0.028 0.188* 0.037 0.186* [0.28] [1.50] [0.37] [1.48] [0.27] [1.69] [0.36] [1.68] Observations 214 169 214 169 214 169 214 169 LogLik -128.07 -100.11 -128.06 -101.52 -128.01 -97.18 -128.01 -97.93 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 4.4: Marginal effects for firms under conditions of increasing uncertainty in business unit’s environment Table 4.4 (continued) 139 Table 4.4 (continued) Variable Name Model 5 Model 6 Model 7 Model 8 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry Information asymmetry in business unit's industry -0.004 0.194 -0.006 0.328 -0.004 0.19 -0.006 0.276 [0.24] [0.71] [0.40] [1.22] [0.22] [0.80] [0.39] [1.19] Uncertainty in business unit's environment -1.724*** 0.253 -1.551*** 0.059 -1.676*** -2.304 -1.505*** -2.579 [3.57] [0.11] [3.55] [0.03] [2.69] [1.10] [2.71] [1.28] Parent performance -0.330*** 0.213 -0.290*** 0.151 -0.349*** 0.137 -0.304*** 0.093 [3.32] [0.60] [3.31] [0.43] [3.33] [0.45] [3.33] [0.32] Segment performance -0.022 -0.2 -0.022 -0.209 -0.019 -0.185 -0.019 -0.191 [0.37] [0.87] [0.39] [0.93] [0.29] [0.99] [0.33] [1.05] Institutional blockholder share -0.001 0.001 -0.001 0.001 -0.001 0.001 -0.001 0.001 [1.61] [0.50] [1.58] [0.57] [1.59] [0.61] [1.55] [0.64] Behavioral Uncertainty -0.478 19.349** -0.396 20.027** [0.53] [2.28] [0.50] [2.47] Parent's diversification (entropy measure) 0.060* 0.118 0.054* 0.109 0.062* 0.125 0.056* 0.121 [1.80] [0.84] [1.81] [0.77] [1.76] [1.08] [1.78] [1.05] Business unit's relatedness with parent -0.009 -0.171*** -0.009 -0.166*** 0.02 -1.392** 0.015 -1.430*** [0.56] [2.83] [0.63] [2.75] [0.33] [2.46] [0.28] [2.63] Leverage (lev1) 0.009 0.159 0.01 0.11 [0.22] [0.97] [0.23] [0.80] Leverage (lev4) -0.006 0.105** -0.007 0.074* [0.46] [2.18] [0.49] [1.90] Parent's valuation (market measure) 0.004*** 0 0.003*** 0 0.004*** 0.001 0.004*** 0 [4.14] [0.13] [4.19] [0.07] [4.13] [0.36] [4.18] [0.23] Firm size 0.004 0.023 0.003 0.021 0.004 0.02 0.003 0.019 [0.80] [1.20] [0.78] [1.12] [0.80] [1.27] [0.78] [1.19] Segment size -0.026 0.023 -0.022 0.009 -0.027 0.019 -0.023 0.01 [1.26] [0.30] [1.22] [0.12] [1.27] [0.29] [1.22] [0.15] Parent's valuation (market measure) 0.004*** 0 0.003*** 0 0.004*** 0.001 0.004*** 0 [4.14] [0.13] [4.19] [0.07] [4.13] [0.36] [4.18] [0.23] Growth options in other segments -0.002** -0.008*** -0.002** -0.008*** -0.002** -0.007*** -0.002** -0.007*** [2.14] [2.80] [2.27] [2.82] [2.09] [2.85] [2.23] [2.88] Observations 210 165 210 165 210 165 210 165 LogLik -109.57 -91 -109.65 -93.01 -109.47 -86.87 -109.56 -88.19 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% 140 Variable Name Model 1 Model 2 Model 3 Model 4 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry
Business unit industry information asymmetry -0.063* 3.421*** -0.066 3.376*** -0.062*** 3.645*** -0.063*** 3.583*** [1.89] [2.77] [1.54] [2.71] [2.98] [2.92] [2.73] [2.85] Uncertainty in a business unit's environment -1.199 -5.335 -1.028 -5.387 -19.301 5.587 -19.34 6.168 [0.10] [0.73] [0.09] [0.75] [1.40] [0.59] [1.41] [0.65] Parent performance -3.847** -4.870** -3.548** -5.115*** -3.795** -4.814** -3.525** -5.087*** [2.27] [2.51] [2.12] [2.75] [2.28] [2.52] [2.15] [2.75] Segment performance 0.809 0.885 0.854 0.902 0.798 1.096 0.829 1.142 [0.82] [1.00] [0.84] [1.02] [0.79] [1.13] [0.80] [1.17] Institutional block- holder share -0.007 -0.001 -0.007 -0.001 -0.009 -0.002 -0.008 -0.002 [0.73] [0.18] [0.70] [0.14] [0.87] [0.23] [0.83] [0.19]
Behavioral uncertainty 93.067** -38.189 93.737*** -39.142 [2.57] [1.43] [2.63] [1.45] Parent's diversification (entropy measure) -0.229 -0.244 -0.141 -0.251 -0.61 -0.356 -0.531 -0.365 [0.26] [0.30] [0.16] [0.31] [0.75] [0.44] [0.65] [0.44]
Business unit's relatedness with parent -0.813** -0.252 -0.798** -0.248 -5.780*** 1.984 -5.792*** 2.037 [1.97] [1.11] [2.01] [1.09] [2.94] [1.41] [3.02] [1.43]
Parent' s leverage (lev1) -0.089 0.694 -0.112 0.838 [0.11] [0.74] [0.13] [0.89]
Parent's leverage (lev4) -0.235 0.196 -0.219 0.217 [1.02] [0.66] [0.92] [0.73] Parent's valuation (sales measure) 0.003 -0.274*** 0.036 -0.288*** 0.01 -0.263*** 0.04 -0.275*** [0.03] [2.66] [0.44] [3.12] [0.12] [2.61] [0.49] [3.03]
Firm size 0.122* 0.094 0.123* 0.082 0.109 0.098 0.11 [1.69] [1.12] [1.69] [0.98] [1.50] [1.18] [1.51] Segment size 0.306 0.255 0.332 0.237 0.143 0.277 0.17 0.085 [0.89] [0.76] [0.97] [0.72] [0.41] [0.84] [0.49] [1.02] Constant -0.738 -0.62 -0.939 -0.532 0.387 -1.342 0.213 0.26 [0.76] [0.58] [0.94] [0.54] [0.38] [1.22] [0.20] [0.80]
Observations 154 145 154 145 154 145 154 LogLik -89.42 -77.04 -89.97 -76.94 -86.8 -75.41 -87.24 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 4.5: Probit results for firms under conditions of decreasing uncertainty in business unit’s environment Table 4.5 (continued) 141 Table 4.5 (continued) Variable Name Model 5 Model 6 Model 7 Model 8 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry
Business unit industry information asymmetry -0.062*** 2.799** -0.060*** 2.601** -0.062*** 2.989** -0.060*** 2.783** [2.65] [2.29] [2.77] [2.13] [3.82] [2.41] [3.81] [2.26] Uncertainty in a business unit's environment 3.654 -0.791 3.386 -1.672 -15.424 8.85 -16.064 9.256 [0.29] [0.11] [0.27] [0.24] [1.08] [0.94] [1.12] [0.98]
Parent performance -4.991*** -3.214* -4.583** -3.719** -4.929*** -3.276* -4.593** -3.811** [2.67] [1.75] [2.52] [2.01] [2.63] [1.78] [2.52] [2.04] Segment performance 0.693 1.041 0.786 1.02 0.698 1.271 0.778 1.303 [0.69] [1.12] [0.77] [1.09] [0.67] [1.24] [0.72] [1.25] Institutional block- holder share -0.004 -0.003 -0.005 -0.003 -0.005 -0.003 -0.006 -0.003 [0.46] [0.35] [0.47] [0.35] [0.55] [0.33] [0.56] [0.33] Behavioral uncertainty 94.700*** -34.196 96.794*** -36.618 [2.63] [1.29] [2.69] [1.37] Parent's diversification (entropy measure) -0.222 0.14 -0.168 0.166 -0.63 0.001 -0.582 0.016 [0.23] [0.19] [0.17] [0.22] [0.71] [0.00] [0.65] [0.02] Business unit's relatedness with parent -0.777* -0.306 -0.790* -0.293 -5.829*** 1.688 -5.949*** 1.828 [1.77] [1.22] [1.87] [1.17] [3.04] [1.20] [3.10] [1.29] Parent' s leverage (lev1) -0.488 1.683* -0.482 1.773** [0.57] [1.96] [0.55] [2.06] Parent's leverage (lev4) -0.322 0.652*** -0.299 0.649*** [1.37] [2.61] [1.23] [2.64] Parent's valuation (market measure) 0.021* -0.022* 0.018 -0.022* 0.023* -0.021* 0.021* -0.022* [1.73] [1.78] [1.60] [1.81] [1.89] [1.76] [1.79] [1.81] Firm size 0.128* 0.145* 0.119 0.117 0.114 0.146* 0.107 0.117 [1.65] [1.74] [1.58] [1.44] [1.47] [1.73] [1.40] [1.43] Segment size 0.278 0.399 0.321 0.309 0.101 0.376 0.139 0.294 [0.78] [1.15] [0.90] [0.90] [0.28] [1.08] [0.38] [0.86] Growth options in other businesses -0.001 0.005 0 0.004 -0.001 0.004 0 0.003 [0.06] [1.50] [0.03] [1.24] [0.12] [1.32] [0.04] [1.07] Constant -1.002 -1.865* -1.034 -1.531 0.159 -2.419** 0.153 -2.177** [1.04] [1.87] [1.04] [1.60] [0.15] [2.33] [0.14] [2.17] Observations 148 139 148 139 148 139 148 139 LogLik -83.79 -76.25 -84.73 -77.31 -81.09 -74.82 -81.85 -75.61 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% 142 Variable Name Model 1 Model 2 Model 3 Model 4 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry
Business unit industry information asymmetry -0.023* 1.168*** -0.024 1.149*** -0.022*** 1.228*** -0.022*** 1.202*** [1.89] [2.77] [1.54] [2.71] [2.98] [2.92] [2.73] [2.85] Uncertainty in business unit's environment -0.433 -1.822 -0.371 -1.834 -6.806 1.882 -6.827 2.07 [0.10] [0.73] [0.09] [0.75] [1.40] [0.59] [1.41] [0.65]
Parent performance -1.389** -1.663** -1.281** -1.741*** -1.338** -1.622** -1.244** -1.707*** [2.27] [2.51] [2.12] [2.75] [2.28] [2.52] [2.15] [2.75] Segment performance 0.292 0.302 0.308 0.307 0.282 0.369 0.293 0.383 [0.82] [1.00] [0.84] [1.02] [0.79] [1.13] [0.80] [1.17] Institutional blockholder share -0.003 0 -0.002 0 -0.003 -0.001 -0.003 -0.001 [0.73] [0.18] [0.70] [0.14] [0.87] [0.23] [0.83] [0.19] Behavioral uncertainty 32.820** -12.863 33.089*** -13.134 [2.57] [1.43] [2.63] [1.45] Parent's diversification (entropy measure) -0.083 -0.083 -0.051 -0.085 -0.215 -0.12 -0.187 -0.123 [0.26] [0.30] [0.16] [0.31] [0.75] [0.44] [0.65] [0.44] Business unit's relatedness with parent -0.294** -0.086 -0.288** -0.085 -2.038*** 0.668 -2.044*** 0.683 [1.97] [1.11] [2.01] [1.09] [2.94] [1.41] [3.02] [1.43] Leverage (lev1) -0.032 0.236 -0.039 0.281 [0.11] [0.74] [0.13] [0.89] Leverage (lev4) -0.085 0.067 -0.077 0.073 [1.02] [0.66] [0.92] [0.73] Firm size 0.044* 0.032 0.044* 0.028 0.038 0.033 0.039 0.028 [1.69] [1.12] [1.69] [0.98] [1.50] [1.18] [1.51] [1.02] Segment size 0.11 0.087 0.12 0.081 0.051 0.093 0.06 0.087 [0.89] [0.76] [0.97] [0.72] [0.41] [0.84] [0.49] [0.80] Parent's valuation (sales measure) 0.001 -0.094*** 0.013 -0.098*** 0.004 -0.089*** 0.014 -0.092*** [0.03] [2.66] [0.44] [3.12] [0.12] [2.61] [0.49] [3.03] Observations 154 145 154 145 154 145 154 145 LogLik -89.42 -77.04 -89.97 -76.94 -86.8 -75.41 -87.24 -75.21 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1% Table 4.6: Marginal effects for firms under conditions of decreasing uncertainty in business unit’s environment Table 4.6 (continued) 143 Table 4.6 (continued) Variable Name Model 5 Model 6 Model 7 Model 8 Increasing Decreasing Increasing Decreasing Increasing Decreasing Increasing Decreasing Divesting firm=1 asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry asymmetry Segment industry information asymmetry -0.023*** 0.989** -0.022*** 0.918** -0.022*** 1.044** -0.021*** 0.970** [2.65] [2.29] [2.77] [2.13] [3.82] [2.41] [3.81] [2.26] Uncertainty in business unit's environment 1.326 -0.279 1.23 -0.59 -5.456 3.091 -5.688 3.225 [0.29] [0.11] [0.27] [0.24] [1.08] [0.94] [1.12] [0.98]
Parent performance -1.811*** -1.135* -1.665** -1.313** -1.743*** -1.144* -1.626** -1.328** [2.67] [1.75] [2.52] [2.01] [2.63] [1.78] [2.52] [2.04] Segment performance 0.252 0.368 0.286 0.36 0.247 0.444 0.275 0.454 [0.69] [1.12] [0.77] [1.09] [0.67] [1.24] [0.72] [1.25] Institutional blockholder share -0.002 -0.001 -0.002 -0.001 -0.002 -0.001 -0.002 -0.001 [0.46] [0.35] [0.47] [0.35] [0.55] [0.33] [0.56] [0.33] Behavioral uncertainty 33.495*** -11.943 34.271*** -12.759 [2.63] [1.29] [2.69] [1.37] Parent's portfolio relatedness (entropy measure) -0.08 0.05 -0.061 0.059 -0.223 0 -0.206 0.006 [0.23] [0.19] [0.17] [0.22] [0.71] [0.00] [0.65] [0.02] Segment's relatedness -0.282* -0.108 -0.287* -0.103 -2.062*** 0.59 -2.106*** 0.637 [1.77] [1.22] [1.87] [1.17] [3.04] [1.20] [3.10] [1.29] Leverage (lev1) -0.177 0.594* -0.171 0.618** [0.57] [1.96] [0.55] [2.06] Leverage (lev4) -0.117 0.230*** -0.106 0.227*** [1.37] [2.61] [1.23] [2.64] Firm size 0.046* 0.051* 0.043 0.041 0.04 0.051* 0.038 0.041 [1.65] [1.74] [1.58] [1.44] [1.47] [1.73] [1.40] [1.43] Segment size 0.101 0.141 0.117 0.109 0.036 0.131 0.049 0.103 [0.78] [1.15] [0.90] [0.90] [0.28] [1.08] [0.38] [0.86] Parent's valuation (market measure) 0.007* -0.008* 0.007 -0.008* 0.008* -0.007* 0.008* -0.008* [1.73] [1.78] [1.60] [1.81] [1.89] [1.76] [1.79] [1.81] Growth options in other segments 0 0.002 0 0.001 0 0.001 0 0.001 [0.06] [1.50] [0.03] [1.24] [0.12] [1.32] [0.04] [1.07] Observations 148 139 148 139 148 139 148 139 LogLik -83.79 -76.25 -84.73 -77.31 -81.09 -74.82 -81.85 -75.61 Robust Z-statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1%
144 Increase in Decrease in information asymmetry information asymmetry Information Information asymmetry not asymmetry not Increase in uncertainty significant to significant to the decision to the decision to divest divest
Figure 4.1: The case of increasing uncertainty and the impact of information asymmetry
Increase in Decrease in information asymmetry information asymmetry Information Information asymmetry has asymmetry has Decrease in a relatively a relatively uncertainty stronger and weaker effect on negative effect the decision to on the decision divest to divest Figure 4.2: The case of decreasing uncertainty and the impact of information asymmetry
145 REFERENCES Akerlof, G. A. 1970. The market for “lemons”: Quality uncertainty and the market mechanism. Quarterly Journal of Economics , 84(3): 488-500. Allen, J., & McConnell, 1997. Equity carve-outs and managerial discretion. Journal of Finance, 53: 163-186. Allison, P. 1984. Event History Analysis. Beverly Hills, CA: Sage. Allison, P. D. 1999. Comparing logit and probit coefficients across groups. Sociological Methods and Research , 28(2): 186-208. Alvi, E. 1988. Information revelation and principal-agent contracts. Journal of Labor Economics, 6(1): 132-145. Aron, D. 1991. Using the capital market as a monitor: corporate spin-offs in an agency framework. RAND Journal of Economics , 22: 505-518. Bayless, M., & Chaplinsky, S. 1996. Is there a window of opportunity for seasoned equity issuance? Journal of Finance , 51: 253-278. Baysinger, B. D., & Hoskisson, R. E. 1989. The composition of board of directors and strategic control: Effects on corporate strategy. Academy of Management Review , 15: 72-87. Berger, P. G. and Ofek, E. 1995. Diversification’s effect on firm value. Journal of Financial Economics, 37(1): 39-65. Berger, P. G., Ofek, E., & Swary, I. 1996. Investor Valuation of the Abandonment Option. Journal of Financial Economics, 42 (2): 257-287. Bergh, D. D. 1995. Size and relatedness of units sold: An agency theory and resource- based perspective. Strategic Management Journal, 16(3): 221-239.
146 Bergh, D. D., Johnson, R. A., Dewitt, R. 2008. Restructuring through spin-off or sell- off: Transforming information asymmetries into financial gain. Strategic Management Journal , 29(2): 133-148. Bergh, D.D., & Lawless, M. 1998. Portfolio restructuring and limits to hierarchical governance: The effects of environmental uncertainty and diversification strategy. Organization Science , 9(1): 87-102. Bethel, J. E., & Liebeskind, J. 1993. The effects of ownership structure on corporate restructuring. Strategic Management Journal , 14(4):15-31. Bollerslev, T. 1986. Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics , 31: 307-321. Boot, A. W. A. 1992. Why hang on to losers? Divestitures and takeovers. Journal of Finance , 47(4): 1401-1423. Bowman, E. H. and Hurry, D. 1993. Strategy through the option lens: An integrated view of resource investments and the incremental-choice process. Academy of Management Review, 18(4): 760-782. Bragger, J. D., Bragger, D., Hantula, D. A., & Kirnan, J. (1998). Hysteresis and uncertainty: The effect of uncertainty on delays to exit decisions. Organizational Behavior and Human Decision Processes , 74(3), 229-253. Campa, J. M. 1993. Entry by foreign firms in the United States under exchange rate uncertainty. Review of Economics and Statistics , 75 (4), 614-622. Carruth, A., Dickerson, A., & Henley A. 2000. What do we know about investment under uncertainty? Journal of Economic Surveys, 14(2): 119-153. Carter, R., & Hodgson, G. M. 2006. The impact of empirical tests of transaction cost economics on the debate on the nature of the firm. Strategic Management Journal , 27(5), 461-476. Chang, S. J. 1996. An evolutionary perspective on diversification and corporate restructuring: Entry, exit, and economic performance during 1981-1989. Strategic Management Journal , 17: 587-611. Chang, S. J., & Singh, H. 1999. The impact of modes of entry and resource fit on modes of exit by multi-business firms. Strategic Management Journal , 20(11): 1019- 1035. 147 Chen, H. L., & Guo, R. J. 2005. On Corporate Divestiture. Review of Quantitative Finance And Accounting, 24: 399–421. Chi, T. 2000. Option to Acquire or Divest a Joint Venture. Strategic Management Journal , 2 1(6): 665-687. Chi, T., & McGuire, D. J. 1996. Collaborative ventures and value of learning: Integrating the transaction cost and strategic option perspectives on the choice of market entry modes. Journal of International Business Studies , 27(2): 285- 308. Chi, T., & Nystrom, P. C. 1995. Decision dilemmas facing managers - Recognizing the value of learning while making sequential decisions. Omega-International Journal of Management Science , 23(3), 303-312. Cho, M. H., & Cohen, M. A. 1997. The economic causes and consequences of corporate divestiture. Managerial and Decision Economics, 18: 367-374. Cuypers, I. R. P., & Martin, X. 2007. Joint Ventures and Real Options: An Extended Perspective. In J. Reuer & T. Tong (Eds.) Advances in Strategic Management: Real options theory . 24: 107-148. David, R. J., & Han, S. K. 2004. A systematic assessment of the empirical support for transaction cost economics. Strategic Management Journal, 25(1), 39-58. Dewatripont, M. 1989. Renegotiation and information revelation over time: The case of optimal labor contracts. Journal of Economics , 104(3): 589-619. Dierickx, I., & Cool, K. 1989. Asset stock accumulation and sustainability of competitive advantage. Management Science , 35(12): 1504-1511. Dixit, A. K. 1992. Investment and Hysteresis. Journal of Economic Perspectives, 6(1), 107-132. Dixit, A. K., & Pindyck, R. S. 1995. The Options Approach to Capital Investment. Harvard Business Review , 73(3), 105-115. Dubin, J. A., & Rivers, D. 1989. Selection bias in linear regression, logit and probit models. Sociological methods and research , 18(2&3):360-390. Duhaime, I. M., & Grant, J. H. 1984. Factors influencing divestment decision-making: Evidence from a field study. Strategic Management Journal , 5(4), 301-318. 148 Duhaime, I. M. & Baird, I. S. 1987. Divestment decision-making: The role of business unit size. Journal of Management , 13 (3): 483-498. Engle, R. 2001. GARCH 101: The Use of ARCH/GARCH Models in Applied Econometrics. Journal of Economic Perspectives , 15(4), 157-168. Finkelstein, S., & Hambrick, D. C. 1989. Chief executive compensation: A study of the intersection of markets and political processes. Strategic Management Journal, 10(2): 121-134. Folta, T., & Leiblein, M. 1994. Technology acquisition and the choice of governance by established firms: Insights from option theory in a multinomial logit model. Academy ofManagement Proceedings : 27-31. Folta, T. B. 1998. Governance and uncertainty: The tradeoff between administrative control and commitment. Strategic Management Journal , 19(11), 1007. Folta, T. B., & Miller, K. D. 2002. Real options in equity partnerships. Strategic Management Journal , 23(1): 77-88. Folta, T. B., & O'Brien, J. P. 2004. Entry in the presence of dueling options. Strategic Management Journal, 25(2), 121-138. Gilroy, B. M. & Lukas, E. 2006. The choice between greenfield investment and cross- border acquisition: A real option approach. Quarterly Review of Economics and Finance, 46(3): 447-465. Greene, W. 2007. Econometric Analysis, 5 th edition. Upper Saddle River, NJ: Prentice- Hall. Guiso, L., & Parigi, G. 1999. Investment and demand uncertainty. Quarterly Journal Of Economics, 114(1), 185-227. Hamilton, R., & Chow,Y. 1993. Why managers divest—Evidence from New Zealand’s largest companies. Strategic Management Journal , 14: 479-484. Harrigan, K. R. 1981. Deterrents to divestiture. Academy of Management Journal, 24(2): 306-323. Harrigan, K. R. 1982. Exit decisions in mature industries. Academy of Management Journal , 25(4): 707-732.
149 Hill, C. W. L., & Hoskisson, R. E. 1987. Strategy and structure in the multiproduct firm. Academy of Management Review, 12(2), 331-341. Hitt, M., Hoskisson, R., Johnson, R., & Moesel, D. 1996. The market for corporate control and firm innovation. Academy of Management Journal , 39(5): 1084- 1119. Hoetker , G. 2007a. The use of logit and probit models in strategic management research: Critical issues. Strategic Management Journal, 28(4): 331-343. Hoetker, G. 2007b. Confounded coefficients: Accurately comparing logit and probit coefficients across group. Working paper, University of Illinois at Urbana- Champaign, IL. Hoskisson, R. E., Johnson, R. A., & Moesel, D. D. 1994. Corporate divestiture intensity in restructuring firms: Effects of governance, strategy and performance. The Academy of Management Journal , 37(5): 1207-1251. Jacquemin, A. P., & Berry, C. H. 1979. Entropy measure of diversification and corporate growth. Journal of Industrial Economics, 27(4): 359-369. Jain, P. 1985. The effect of voluntary sell-off announcements on shareholder wealth. Journal of Finance, 40: 209-224. James, S. D. & Shaver, M. 2008. Strategic motivations for voluntary public R&D disclosures. Working paper, The Ohio State University. Jensen, M. C. & Murphy, K. J. 1990. Performance pay and top management incentives. Journal of Political Economy , 98(2). Jones, G. R. & Hill, C. W. L. 1988. Transaction cost analysis of strategy-structure choice. Strategic Management Journal, 9: 159-172. Kester, W. C. 1984. Today's Options for Tomorrow's Growth. Harvard Business Review , 62(2), 153-160. Kogut, B. 1983. Foreign direct investment as a sequential process. In C. Kindleberger and D. B. Audretsch (Eds.), The Multinational Corporation in the 1980s : 35-56. Cambridge, MA: MIT Press. Kogut, B. 1985. Designing global strategies: Profiting from operational flexibility. Sloan Management Review, 27(1): 27-38. 150 Kogut, B. 1991. Joint ventures and the option to expand and acquire. Management Science , 37(1), 19-33. Kogut, B., & Kulatilaka, N. 2001. Capabilities as Real Options. Organization Science, 12(6), 744. Krishnaswami, S., & Subramaniam, V. 1999. Information asymmetry, valuation, and the corporate spin-off decision. Journal of Financial Economics, 53(1), 73-112. Kulatilaka, N. 1995. Operating flexibilities in capital budgeting: Substitutability and complementarity in real options. In L. Trigeorgis (Ed.). Real Options in Capital Investment: Models, Strategies, and Applications. Westport, Conn: Praeger. Kulatilaka, N., & Perotti, E. C. 1998. Strategic growth options. Management Science, 44(8), 1021-1031. Kumar, M. V. S. 2005. The value from acquiring and divesting a joint venture: A real options approach. Strategic Management Journal, 26(4), 321-331. Lang, L. H. P., & Stulz, R. M. 1994. Tobin’s q, corporate diversification and firm performance. Journal of Political Economy, 102(6): 1248-1281. Lang, L., Poulsen, A., & Stulz, R. M. 1994. Asset sales, firm performance, and the agency costs of managerial discretion. NBER Reporter, Spring 94. Leiblein, M. J. & Ziedonis, A. A. 2007. Deferral and growth options under sequential innovation. In J. Reuer & T. Tong (Eds.) Advances in Strategic Management: Real options theory , 24: 225-245. Li, Y., James, B., Madhavan, R., & Mahoney, J.T. 2007. Real options: Taking stock and looking ahead. In J. Reuer & T. Tong (Eds.) Advances in Strategic Management: Real options theory , 24. MacMillan, I. C., & McGrath, R. G. 2002. Crafting R&D project portfolios. Research- Technology Management, 45(5): 48-59. Markides, C. 1992. The economic characteristics of de-diversifying firms. British Journal of Management, 3: 91-100. Masulis, R., and Korwar, A. 1986. Seasoned equity offerings: An empirical investigation. Journal of Financial Economics , January/February: 91-118. 151 McGrath, R. G. 1997. A Real Options Logic for Initiating Technology Positioning Investments. Academy of Management Review , 22(4): 974-996. McGrath, R. G. 1999. Falling forward: Real options reasoning and entrepreneurial failure. Academy of Management Review , 24(1): 13-30. Merton, R. C. 1998. Applications of Option-Pricing Theory: Twenty-Five Years Later. American Economic Review , 88 (3), 323-349. Mikkelson, W., and Partch, M. 1986. Valuation effects of security offerings and the issuance process. Journal of Financial Economics , 15: 31-60. Mittendorf, B. 2004. Information revelation, incentives, and the value of a real option. Management Science , 50(12): 1638-1645. Montgomery, C., & Thomas, A. 1988. Divestment: motives and gains. Strategic Management Journal , 9: 93-97. Myers, S. C. 1977. Determinants of corporate borrowing. Journal of Financial Economics , 5(2): 147-176. Myers, S. C., & Majd, S. 1990. Abandonment Value and Project Life. Advances in Futures and Options Research, 4:1-21. Myers, S. C., & Majluf, N. S. 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics , 13(2), 187-221. Nanda, V. 1991. On the Good News in Equity Carve-Outs. Journal of Finance, 46(5), 1717-1737. Nanda, V. and Narayanan, M. P. 1998. Disentangling value: Misvaluation and divestitures. Working paper, University of Michigan Business School. O'brien, J. P., Folta, T. B., & Johnson, D. R. 2003. A real options perspective on entrepreneurial entry in the face of uncertainty. Managerial and Decision Economics, 24(8): 515-533. Ozbilgin, M. and Penno, M. 2005. Corporate disclosure and operational strategy: Financial vs. operational success. Management Science , 51(6): 920-931. 152 Patton, G. R., & Duhaime, I. M. 1980. Divestment as a strategic option: an empirical study. Presented at the 38 th annual meetings of the Academy of Management , San Francisco, August 1978. A related summary is in ‘Selling off’, Wharton magazine , Winter 1980, 43-47. Perotti, E. & Rossetto, S. 2007. Unlocking value: Equity carve outs as strategic real options. Journal of Corporate Finance , 13(5): 771-792. Petersen, M. A. 2007. Estimating standard errors in finance panel data sets: Comparing approaches. Northwestern University Working Paper (Forthcoming in review of financial studies). Powers, E. 2001. Spinoffs, Selloffs, and Equity Carveouts: An Analysis of Divestiture Methods Choice, working paper, University of South Carolina. Powers, E. A. 2003. Deciphering the motives for equity carve-outs. Journal of Financial Research , 26(1): 31-50. Ravenscraft, D. J., & Scherer, F. M. 1991. Divisional Sell-off: A Hazard Function Analysis. Managerial & Decision Economics , 12(6): 429-438. Reuer, J. J., & Leiblein, M. J. 2000. Downside risk implications of multi-nationality and international joint ventures. Academy of Management Journal , 43: 203-214. Reuer, J. J. & Tong, T. W. 2005. Real options in international joint ventures. Journal of Management , 31(3): 403-423. Reuer, J. J. & Tong, T. W. 2007. Real options in strategic management. In J. Reuer & T. Tong (Eds.) Advances in Strategic Management: Real options theory , 24. Robins, J., & Wiersema, M. F. 1995. A resource-based approach to the multi-business firm: empirical analysis of portfolio interrelationships and corporate financial performance. Strategic Management Journal , 16(4): 277-299. Sanchez, R. 1993. Strategic flexibility, firm organization, and managerial work in Dynamic markets: A strategic-options perspective. Advances in Strategic Management , 9: 251-291. Sanchez, R., & Mahoney, J. T. 1996. Modularity, flexibility, and knowledge management in product and organization design. Strategic Management Journal , 17: 63-76. Sanders, G. 2001. Behavioral responses of CEOs to stock ownership and stock option pay. Academy of Management Journal , 44(3): 477-492. 153 Schlingemann, F. P., Stulz, R. M. & Walkling, R. A. 2002. Divestitures and the liquidity of the market for corporate assets. Journal of Financial Economics, 64: 117- 144. Shimizu, K. & Hitt, M. A. 2005. What constrains or facilitates divestitures of formerly acquired firms? The effects of organizational inertia. Journal of Management , 31(1): 50-72. Smit, H. T. J., & Ankum, L. A. 1993. A real options and game-theoretic approach to corporate investment strategy under competition. FM: The Journal of the Financial Management Association , 22(3): 241-250. Train, K. 1986. Qualitative choice analysis: Theory, econometrics, and an application to automobile demand. Cambridge, MA: MIT Press. Trigeorgis, L. 1998. Real options: Managerial flexibility in strategy and resource allocation . MIT Press: Cambridge, MA. Vassolo, R. S., Anand, J., & Folta, T. B. 2004. Non-additivity in portfolios of exploration activities: a real options-based analysis of equity alliances in biotechnology. Strategic Management Journal , 25 (11), 1045-1061. Vignola, L. 1974. Strategic Divestment. American Management Association , New York. Vijh, A. M. 2002. The positive announcement-period returns of equity carve-outs: Asymmetric information or divestiture gains? Journal of Business , 75(1): 153- 190. Warner, A. G., Fairbank, J. F., & Steensma, H. K. 2006. Managing uncertainty in a formal standards-based industry: A real options perspective on acquisition timing Williamson, O. E. 1979. Transaction-cost economics: the governance of contractual relations. Journal of Law & Economics , 22(2), 233-261. Williamson, O. E. 1985. The Economic institutions of capitalism , Free Press. Yosha, O. 1995. Information disclosure costs and the choice of financing source. Journal of Financial Intermediation , 4(1): 3-20. Zingales, L. 1995. Insider ownership and the decision to go public. Review of Economic Studies , 62: 425-448. 154 APPENDIX
List of Filters used: 1. Only completed deals from SDC 2. For sell-offs, deals from SDC to be matched with dropped segments in COMPUSTAT segment database 3. Transactions confirmed by searching SEC filings, news articles on Factiva and/or Lexis-Nexis 4. Companies should match with COMPUSTAT Industrial annual and segment databases 5. Only first divestment decision on a segment is considered 6. Parent firms and divested segments belong to manufacturing and related categories 7. Utilities are excluded 8. Parent firms and divested segments are located in the United States 9. No American Depository Receipts 10. Parents are not limited partnerships 11. Divested units are not organized as limited partnerships 12. Spin-offs/equity carveouts listed on CRSP, not owned by multiple parents and not direct result of mergers 13. Not tracking stock deals
155