# Handbook of Entrepreneurial Dynamics: The Process of Business Creation

Handbooks

### Edited by: William B. Gartner, Kelly G. Shaver, Nancy M. Carter & Paul D. Reynolds

• Chapters
• Front Matter
• Back Matter
• Subject Index
• ## Dedication

To Greg, soul mate and friend; Matt, Jason, Sonja, and Lauri, luminaries, all.

—Nancy M. Carter

To Maryse who keeps me in the larger life outside the academic world, and to Ellison who can stop time with her laughter.

—William B. Gartner

To the Barteldeses, Blakes, Reynoldses, and Schaeffers—my ancestors—who came to the new world to create new businesses.

—Paul D. Reynolds

To Jen. And to Alan, for getting me started.

—Kelly G. Shaver

## Foreword

New business creation has a significant impact on economic growth, innovation, and job creation (Reynolds, Camp, Bygrave, Autio, & Hay, 2001; Reynolds, Hay, Bygrave, Camp, & Autio, 2000). Since David Birch's 1979 study of the impact of new and small firms on creating new jobs, a considerable body of evidence has been generated that supports his conclusion that small firms are the major source of employment growth in the U.S. economy (Birch, 1979, 1987; Kirchhoff, 1994; Reynolds & White, 1997). More recent assessments, however, have indicated that the original focus was too broad. New firms, not necessarily small firms, are the dominant source of net job growth; there is a net job loss among older firms, whether small or large (Acs, Armington, & Robb, 1999). Entrepreneurial activity provides profound positive benefits across an important set of measures of social and economic well-being, much of them concentrated in new economic sectors such as information technology, when compared to service-producing or goods-producing industries (Boden, 2000).

Although entrepreneurship has been shown to provide many benefits, there has not been a systematic study of the entrepreneurial process. Although entrepreneurs contribute so much to our society, we know little about them as people. We can see the results of entrepreneurial activity in the form of new businesses and innovations, but we have limited information on how these new businesses actually came into existence. We can see the successes of entrepreneurial activity, yet we have few insights into why particular entrepreneurial efforts were successful while other efforts failed. Indeed, we have almost no information on the number and characteristics of the nascent entrepreneurs who attempt to start businesses and the likelihood that such attempts will result in the formation of new businesses.

What distinguishes the PSED from other studies of the entrepreneurial process is that it identifies individuals in the process of creating new businesses and develops systematic, reliable, and generalizable data on important features of the new business creation process, including information on the proportion and characteristics of the adult population involved in efforts to start businesses, the activities and characteristics that constitute the nature of the business start-up process, and the proportion and characteristics of those business start-up efforts that actually become new businesses.

This handbook reports on the creation of the Entrepreneurship Research Consortium (ERC), the organizing group for the PSED; the evolution of the research program; and theories, ideas, and measures for exploring and understanding factors that encompass and influence the creation of new businesses. What makes this handbook unique, among overviews of prior theory, ideas, and research on the phenomenon of business creation, is that each chapter provides the rationale used in developing questionnaires for the PSED and specifies measures that can be used to test theory, often providing evidence from the PSED data sets on these measures' validity and reliability. In addition, inherent in each chapter is the implication that the theories about business creation can be tested using a generalizable sample of both nascent entrepreneurs and a comparison group. The handbook also describes the PSED data collection process, provides documentation of the interview schedules, codebooks, data preparation, and weighting scheme, as well as offers examples of how analyses of PSED data might be conducted. The handbook is an invitation to explore theory on the nature of business creation and to test ideas through analyses of the PSED data sets that are in the public domain.

Conceptual Model

The PSED built on earlier efforts by Paul Reynolds and colleagues to study nascent entrepreneurs in Wisconsin (Reynolds & White, 1993, 1997), as well as a small national sample of nascent entrepreneurs who were identified from a study that was “piggy-backed” onto the University of Michigan Institute for Social Research Survey of Consumer Attitudes (Curtin, 1982; Reynolds, 1997). These prior studies indicated that it was technically feasible, as well as financially feasible, to locate and survey individuals from the general population of all U.S. adults who were actively engaged in starting businesses.

Conceptually, the entrepreneurial process can be thought of as involving three stages and three transition points, all continually influenced by political, social, and economic factors. As depicted on the left side of the model (Figure F.1), the first stage of the start-up process involves the population of all individuals, some of whom might decide to start a business. These individuals come from two potential sources: all those involved in the labor force and those who are employees of existing businesses.

The first transition point in the start-up process is “conception” when individuals from these two sources elect to pursue a new business start-up. If the new business is intended to be an independent start-up, those involved are referred to as nascent entrepreneurs (NE). If the start-up effort is sponsored by an existing business, those involved are considered to be nascent intrapreneurs (NI). The primary research questions at conception are two: What is the tendency of individuals to

Figure F.1 The Start-up Process and Handbook Organization

begin the business start-up process? What are the features of these individuals or their situation that lead some to enter this transition? There is a great deal of speculation that entrepreneurs are very different from other individuals in the general population. An important feature of the PSED involves the identification of a comparison group of adults in the general population who are not involved with start-up activities. Because the comparison group can be weighted to mirror the characteristics of the U.S. adult population, generalizations can be made about differences between the sample of nascent entrepreneurs and the U.S. adult population as a whole.

Unlike other studies of the entrepreneurial process, the PSED provides a detailed focus on the second stage in the process of business formation: the start-up process. This stage involves factors that affect the efforts of nascent entrepreneurs to bring their businesses into existence as well as the length of time involved in their startup efforts. The PSED describes the entrepreneurs as people, documents their activities, and summarizes the types and amounts of resources invested in the process. The primary question the PSED explores at the start-up processes stage is: How do nascent entrepreneurs go about the process of starting firms? An entrepreneur's start-up activities can take the form of four possible pathways: (1) The nascent entrepreneur creates an infant firm. (2) The nascent entrepreneur can be “still trying” to start the business. (3) The nascent entrepreneur can put the start-up effort “on hold” with expectations of continuing to pursue the start-up process later. And, (4) the nascent entrepreneur can “give up” and abandon the start-up effort.

If an entrepreneur's activities result in an infant firm, the second transition point in the model is “firm birth.” The question at this point is Why are some of these business start-up efforts successful in creating new firms?

The third transition point involves the subsequent “outcomes,” or activities, of the new firm; that is, once a new firm comes into existence, the firm may either grow, persist, or be discontinued. The question at this point is Why are some new firms more likely to grow, persist, or die?

A key insight into the PSED research process is an appreciation of the variety and diversity inherent in the phenomenon of business creation (Gartner, 1985). One can easily grasp this perspective by quickly listing the many kinds of businesses that one might encounter through a cursory scan of the Yellow Pages in the telephone book: restaurants (by type—fast food, casual, fine dining; by cuisine—ethnic, pizza, American, eclectic), services (medical, law, accounting, architecture, dry cleaning, package delivery), manufacturing (automobiles, soaps, furniture, computers, pharmaceuticals, beverages), retailing (clothing, jewelry, appliances, consumer electronics, sporting goods). How did all of these many different kinds of businesses come into existence? Indeed, how many different kinds of businesses are actually attempted? How similar (or different) is the process of starting a restaurant compared to starting a biotechnology research firm? How similar (or different) are the kinds of people who are likely to start a retail store compared to those starting a construction company? The basis for the PSED research effort is this recognition of variety of business creation efforts and the need to portray both the breadth and depth of this phenomenon.

Research Design

The research design for the PSED is presented in Figure F.2 and shows two basic features: (1) a procedure for identifying and interviewing nascent entrepreneurs and a comparison group and (2) the content of the interviews.

The first stage involves large-scale screening of households to create two samples representative of the national population of adults, those 18 years old and older. First, a sample of those involved in attempting to start a new business was identified, either nascent entrepreneurs (NE) or nascent intrapreneurs (NI). Second, a representative sample of typical adults, not involved with a business start-up, was selected to be used as the comparison group (CG). Once the screening procedures were completed, the second stage of data collection involved detailed phone interviews followed by completion of self-administered questionnaires mailed to the respondents. The third stage involved the follow-up interviews completed with nascent entrepreneurs 12 and 24 months after the first interview. Details about the research design, data collection process, the composition of the data sets, and examples of how analyses of the data might be undertaken are found in the appendixes.

Figure F.2. PSED Research Design Overviewa

Appendix A explains the PSED research design and data collection process in detail. This appendix describes the process for contacting potential respondents, the procedure and criteria for screening respondents to identify nascent entrepreneurs, the phone and mail interview schedules used for the nascent entrepreneurs and the comparison group, and the process of data collection for the follow-up interviews. Appendix B provides information on the two primary PSED data sets: the screening data set which consists of 64,622 respondents initially contacted and the detailed data sample set of the 830 nascent entrepreneurs and 431 comparison group members. This appendix also discusses the creation of weights as a way to ensure the generalizability of the detailed data sample to accurately reflect characteristics of the population. Appendix C provides examples of how analyses of both the screening data set and the detailed data sample set might be conducted using the weights.

The Entrepreneurship Research Consortium (ERC)

To adequately capture the complex phenomenon of business creation, a social, political, and collaborative process atypical among scholars studying organizations (or anything else) was put in place. Rather than one major funding source for the PSED, there was a multitude of funding sources. Rather than one group of scholars involved in the development of the PSED, there was a multitude of groups. Rather than one particular research agenda, there was a multitude of research agendas. The fundamental value of those involved in the PSED centered on an attempt to use theory to develop questions that could depict the phenomenon of business creation in as comprehensive a manner as possible. The use of theory in the development of the PSED is therefore less an effort to create a general theory about the nature of business creation and more an effort to depict business creation in a generalizable way.

The idea to develop a consortium of individuals and institutions involved in the process of research on new business formation came about because of a collaborative effort among Paul Reynolds, Nancy Carter, and Bill Gartner in 1994 to study the new venture creation activities of a random sample of less than 100 nascent entrepreneurs (Carter, Gartner, & Reynolds, 1996). They recognized that a large sample of nascent entrepreneurs would be important for ascertaining differences and similarities among various types of nascent entrepreneurs and their businesses as well as for comparing and contrasting nascent entrepreneurs to those who were not engaged in business creation. The problem was no one particular institution was willing to fund a project that would likely cost hundreds of thousands of dollars to locate individuals in the process of business creation and then follow up on their business creation efforts over a number of years. As described in Reynolds (2000), a series of meetings and communications (letters, faxes, phone calls, and e-mails) was undertaken to generate interest in creating a large (1,000+) database of nascent entrepreneurs and a comparison group that would be created from a random-digit telephone dialing procedure.

Table F.1 lists key meetings and activities that were held to generate financial support and form teams to develop the survey instruments and questionnaires. During the early part of 1995, colleagues who we thought might be interested in participating in such a project were contacted. Besides making phone calls, we staged a forum at the Babson Kauffman Entrepreneurship Research Conference in London (April) to gauge interest, and we began a mail campaign that involved sending letters to all members of the Entrepreneurship Division of the Academy of Management (May). The process of organizing the ERC moved very quickly from this point on. After presentations at the National Family Business Conference in July 1995 and the Academy of Management meeting in August 1995, there appeared to be enough interest in ERC for a meeting solely devoted to the ERC research process. This meeting was held in November at the Chicago O'Hare Hilton. More than 50 people from 6 countries attended. Design teams were formed to develop research objectives, discuss hypotheses, identify specific questionnaire items, and scholars representing their institutions made financial commitments (e.g., some scholars brought checks).

Table F.1 Key Dates in the Development of the ERC
DateEvent
January-March 1995Colleagues contacted to gauge interest in collaborative research
April 1995Forum at Babson Kauffman Entrepreneurship Research Conference, London, UK
May 1995Letter campaign to the Entrepreneurship Division of the Academy of Management members
July-August 1995Recruiting presentations at National Family Business Conference, Nashville, TN, and Academy of Management meeting, Vancouver, BC
November 1995Organizing meeting of ERC, Chicago, IL
January-April, 1996Paid membership of 22 institutions designed initial questionnaires; developed administrative structure and decision processes of ERC
August 1996Pilot study assessment and planning meeting, Cincinnati, OH
November 1996Data analysis workshop, University of Houston, TX
January 1997Review of initial survey results, Atlanta, GA
January 1998Full implementation of research process
January 1998Nat. Sci. Foundation (NSF) grant application for female oversample
February 1998–2002Review and update meetings, University of Southern California Los Angles, CA
June 1998–2002Review and update meetings, Babson Kauffman Entrepreneurship Research Conferences: Gent, Belgium; Columbia, SC; Babson Park, MA; Jönköping, Sweden; Boulder, CO
August 1998–2002Review and update meetings, Academy of Management: San Diego, CA; Chicago, IL; Toronto, Canada; Seattle, WA; Washington, DC
January 1999NSF grant application for minority oversample
2000–2003Ewing Marion Kauffman Foundation funding for data collection
November 2001ERC votes to transfer responsibility for the project to the Ewing Marion Kauffman Foundation
January 2002Preferential Rights to Scholarly Analysis (PRSA) procedure abandoned
January 2003All data shifted to the public domain on the Institute for Social Research (ISR) project Web site
December 2003Completion of final data collection for the fourth wave of the panel study

By April 1996, there were 22 institutions that had sent money to participate in the ERC, and five design teams had assembled materials for the initial household screener, and had prepared detailed phone and mail interview schedules for the nascent entrepreneurs and a comparison group, as well as a phone interview for business angels. During this time an executive committee was nominated and elected to supervise the ERC process (Candida Brush, Nancy Carter, Per Davidsson, William Gartner, Paul Reynolds, Kelly Shaver, and Mary Williams), and Babson College was selected as the host institution for the ERC. In addition, a procedure for allocating exclusive rights for analyzing data on a particular topic area was formulated (entitled: Preferential Right to Scholarly Analysis—PRSA). The PRSA process enabled scholars to submit proposals to the executive committee outlining specific research questions and hypotheses, items to be utilized, a strategy for analysis, and dissemination plans for the results. The executive committee reviewed all submitted proposals to look for overlaps. When similar topics were proposed by different groups of scholars, the executive committee asked scholars to collaborate or to narrow their research “claims” to more narrow topic issues so that other scholars could also conduct research around that topic area.

In August 1996, a “Pilot Study Assessment and Planning Meeting” was held concurrently with the national Academy of Management meeting. A “Data Analysis Workshop” of the pilot study results was held at the University of Houston in November 1996, which was followed by another meeting held in January 1997 in Atlanta. Concurrent to all of these meetings was substantial interaction among all of the scholars about their proposed PRSAs and the inclusion of questions in the questionnaires as data from the pilot studies was analyzed. By the end of 1997, two pilot studies had been conducted, all of the interview schedules were nearly complete, and a total of 34 institutions had provided funds.

As the full field data collection began to be implemented in 1998, it became apparent that the research process would require substantially more funds in order to identify the necessary number of nascent entrepreneurs, particularly women and minority nascent entrepreneurs. A proposal to supplement the nascent entrepreneur sample was submitted to the National Science Foundation (NSF) (Carter, Brush, Aldrich, Greene, & Katz, 1998) and funding was received to double the number of women nascent entrepreneurs surveyed as well as to cover the cost of a 24-month follow-up for the entire sample. In addition, a second proposal was prepared for the National Science Foundation to sponsor an oversample of Blacks and Hispanics (Greene, Carter, Reynolds, Aldrich, & Stearns, 1999). This proposal was also funded. Finally, the Kauffman Center for Entrepreneurial Leadership (now the Ewing Marion Kauffman Foundation) began to provide substantial funding for the Panel Study of Entrepreneurial Dynamics in 2000 through completion of the final and fourth wave of the data collection in 2003. The Ewing Marion Kauffman Foundation began to provide substantial funding for the PSED in 2000 through completion of the final and fourth wave of data collection in 2003. Perhaps most important, Babson College served as the host institution for the ERC, acted as the steward of the ERC funds, and allowed Paul Reynolds, the Paul T. Babson Chair of Entrepreneurship, to devote the majority of his time to service as the coordinating principal investigator for the life of the ERC and the duration of the project.

From 1998 to 2002 there were meetings and updates on the data collection process held at the Babson Kauffman Entrepreneurship Research conferences (each June), at the national Academy of Management meetings (each August), at the University of Southern California Greif Research symposiums on Emerging Organization (each February), and at a meeting at the Kauffman Center for Entrepreneurial Leadership at Kansas City, MO in July 2001. Scholars involved in the PSED had opportunities at these meetings to meet, discuss, and make decisions about this research effort.

This brief overview of some of the key events in the history of the ERC is to suggest that an emphasis on some topics in the handbook rather than others is partially a matter of who was involved in the PSED research process, as well as when they were involved and for how long they were involved. The ideas and theories used in a research program such as the PSED required scholars who were willing to champion their ideas and see them through from the inception of the ERC in 1995 through the many design meetings to formulate questions in the survey instruments, through the pilot studies, through the full field data collection, and (for longitudinal issues) through the follow-up surveys.

Development of the Handbook

The chapters included in this volume are the result of an invitation in 2001 to all scholars in the ERC who participated in the development of the PSED. A listing of the individuals is provided in Table F.2. Each scholar was asked to contribute a chapter that summarized a key theoretical perspective that was operationalized in the PSED research program and to identify variables associated with that theory. An important part of the development of the PSED questionnaires (and by implication the theory used to create the questionnaires) was a political and social process of getting agreement among scholars representing 34 sponsoring institutions. Reaching agreement on the interview schedules, particularly on specific items and groups of items on the interview schedules, was no small task. Each institution and scholar involved in the PSED had some specific interest in understanding some facet of the process of business creation. Limits on the budget, and on the time of respondents, meant that some questions about the business creation process would simply have to be omitted. The PSED reflects the spirit of compromise among differing viewpoints, as well as the persistence and tenacity of a few to see this project reach fruition. The chapters in this handbook, therefore, represent the primary theoretical viewpoints upon which the process of business creation was explored. In addition, each chapter indicates how theory about the entrepreneurial process was operationalized, and insights are offered as to the efficacy of these measures.

Table F.2 PSED Scholars

The relationship between the PSED research program and the organization of the handbook is included in Figure F.1. In broad terms, a number of factors are likely to influence a person's decision to engage in entrepreneurial activity and subsequently persist in efforts to start a new business. The model shows the entrepreneurial process as involving three major transitions. The first transition is the entry into the start-up process, and the second transition is the exit from the start-up process—either with a new firm birth or abandoning the effort itself. The major factors or processes that affect these transitions are indicated in the four dashed-line boxes. Two are seen as operating in parallel, perhaps with substantial interaction: the life context and personal background and individual cognitive characteristics or dispositions. These are covered in Parts I and II of the handbook. The actual nature of the start-up process itself, which can be quite complex, is covered in Part III. The environmental context in which all these processes operate is the focus of Part IV.

Part I on “life context and personal background” offers theory, measures, and some evidence on, primarily, demographic characteristics of both the individuals who indicated they were actively engaged in the business formation process (nascent entrepreneurs) and a comparison group of individuals who were selected from the survey of individuals who indicated they were not actively engaged in the business formation process. Chapters in this section cover such topics as age, gender, race and ethnicity, household structure, household income and net worth, labor force participation, residential tenure, work experience, educational background, functional expertise, family background, time use, and work participation history.

Part II covers cognitive characteristics that might be useful for exploring whether nascent entrepreneurs think in ways that are different from those individuals who have not attempted to start businesses, as well as ways these constructs can also be used to differentiate among types of nascent entrepreneurs. The topics covered in this section include career reasons, entrepreneurial expectancies, satisfaction with job and life, the adaptor/innovator style, perceived social support, entrepreneurial intensity, problem-solving style, attribution, locus of control, economic sophistication, and social skills.

Part III explores the factors that constitute the characteristics of the process of business formation itself. Chapters in this section describe the kinds of businesses undertaken (by legal form, type of economic activity, ownership structure, and location), the process of opportunity recognition, start-up activities, start-up problems, team composition, characteristics of firm founding, social networks, knowledge and use of assistance, funding, financial sophistication, and future expectations.

Part IV offers two chapters on the context of the business start-up effort, such as perceptions of entrepreneurial climate and environmental uncertainty, as well as two chapters that explore the strategic and technology orientations of the emerging new businesses.

Finally, there are three appendixes that describe the PSED data collection process, data documentation, data preparation, the use of weights, and examples of how analyses might be conducted.

An important aspect of this handbook is the multilevel, multidisciplinary perspective offered about the process of new business formation. Although each chapter focuses on a particular theory or ideas about an aspect of new business formation, the handbook, taken as a whole, suggests that new business formation is “over-determined” (Weick, 1979). Each factor specified in these chapters may account for a significant part of the variance that determines aspects of the new business formation process. New business formation is likely to be significantly affected by the kinds of individuals involved, kinds of businesses started, a variety of environments, and the ways in which these businesses are started. All of these dimensions matter—individual, firm, environment, and process—in the study of new business formation.

A Dynamic Program

It is inevitable that any book will represent current knowledge at the time its chapters were written. Because the PSED is an active ongoing research program, we recommend that scholars interested in state-of-the-art information about the data set (as well as the data set, itself) seek out the following two Web sites. The first Web site is maintained by the Institute for Social Research (ISR) at the University of Michigan at—http://projects.isr.umich.edu/PSED/. This site is likely to be maintained as long as the ISR is contracted to continue the collection of data on subsequent panels of nascent entrepreneurs. At some point, it is likely that all of the data, codebooks, and interview schedules will be transferred from this Web address to the University of Michigan Inter-University Consortium for Political and Social Research (ICPSR) data archives. Many scholars at U.S. universities have access to the ICPSR archives, as their home institutions are members of the consortium. There are, however, some institutions, particularly those outside North America, that are not participants in the ICPSR. A “parallel” Web site has been developed (http://www.psed.info) with the current information on the PSED data, codebooks, and interview schedules, as well as a listing of current research efforts and activities of scholars working with the PSED. This Web site is maintained through the support of one or more of the scholars and institutions involved in PSED research activities and not housed, per se, at any one specific university or research center. The http://www.psed.info site, therefore, should be a way to easily find information on the PSED. Finally, the editors of the handbook can be contacted to locate current information on the PSED, as well.

References
, , & (1999). Measures of job flow dynamics in the U.S. economy (Discussion Paper). Upper Marlboro, MD: U.S. Bureau of the Census, Center for Economic Studies.
(1979). The job generation process. Unpublished report by the Massachusetts Institute of Technology, Program on Neighborhood and Regional Change. Washington, DC: U.S. Department of Commerce, Economic Development Administration.
(1987). Job creation in America. New York: Free Press.
(2000). Employment establishment changes and survival, 1992–1996 (Discussion Paper). Upper Marlboro, MD: U.S. Bureau of the Census, Center for Economic Studies.
, , , , & (1998). The influence of founder's gender in business start-ups. Proposal to the National Science Foundation, Grant SBR-9809841.
, , & (1996). Exploring start-up event sequences. Journal of Business Venturing, 11 (3), 151–166. http://dx.doi.org/10.1016/0883-9026%2895%2900129-8
(1982). Indicators of consumer behavior: The University of Michigan Surveys of Consumers. Public Opinion Quarterly, 46, 340–362. http://dx.doi.org/10.1086/268731
(1985). A framework for describing and classifying the phenomenon of new venture creation. Academy of Management Review, 10 (4), 696–706.
, , , , & (1999). The influence of founder's race in the start-up process. Proposal to the National Science Foundation, Grant SBR-9905255.
(1994). Entrepreneurship and dynamic capitalism: The economics of business firm formation and growth. Westport, CT: Praeger.
(1997). Who starts new firms? Preliminary explorations of firms-in-gestation. Small Business Economics, 9, 449–462. http://dx.doi.org/10.1023/A:1007935726528
(2000). National study of U.S. business start-ups: Background and methodology. In (Ed.), Advances in entrepreneurship, firm emergence and growth (Vol. 4, pp. 153–228). Stamford, CT: JAI.
, , & [and others] (2003). Global entrepreneurship monitor: 2003 executive report. Babson Park, MA: Babson College.
, , , , & (2001). Global entrepreneurship monitor: 2001 executive report. Kansas City, MO: Kauffman Center for Entrepreneurial Leadership.
, , , , & (2000). Global entrepreneurship monitor: 2000 executive report. Kansas City, MO: Kauffman Center for Entrepreneurial Leadership.
, & (1993). Wisconsin's entrepreneurial climate study. Milwaukee, WI: Marquette University, Center for the Study of Entrepreneurship.
, & (1997). The entrepreneurial process: Economic growth, men, women, and minorities. Westport, CT: Quorum Books.
(1979). The social psychology of organizing (
2nd ed.
). New York: Random House.

## Acknowledgments

As with any entrepreneurial venture, the resources to support the development and implementation of the Panel Study of Entrepreneurial Dynamics (PSED) came in stages. For seed round funding and support we gratefully acknowledge the universities and research institutions, listed in Table F.3, that funded the Entrepreneurship Research Consortium (ERC). Without their initial financial support, and perhaps more importantly the intellectual capital contributed by the membership, the panel study would not have been possible.

The research design of the PSED included a number of innovations. The first was the wide range of topics and issues covered by the data collection procedures. These reflect the substantial intellectual investment of the ERC members and the design teams that acted as a focus for the assembly of ideas and issues and their conversion into items to be included in the phone and mail questionnaires. In addition, major elements of the research methodology were developed in partnership with the University of Wisconsin Survey Research Laboratory; virtually the entire staff was involved and numerous enhancements were provided by the interviewers themselves. Associate Director Charles Palit was particularly helpful in the creation of the sampling designs that led to the current version of the screening schedule. Victoria Ivey developed the corporate identity package for the ERC.

Having proven the venture concept, second-round funding for the research program came from several sources. The National Science Foundation provided funding for an oversample of women through Grant No. 9809841 (Nancy M. Carter, principal investigator; Howard E. Aldrich, Candida G. Brush, Patricia G. Greene, and Jerry Katz, coprincipal investigators) and an oversample of Blacks and Hispanics through Grant No. 9905255 (Patricia G. Greene, principal investigator; Howard E. Aldrich, Nancy M. Carter, Paul D. Reynolds, and Timothy M. Stearns, coprincipal investigators). These grants allowed us to double the size of the sample and to fund the 24-month follow-up surveys.

The Kauffman Center for Entrepreneurial Leadership (KCEL) (now the Ewing Marion Kauffman Foundation) began to provide substantial funding for the PSED in 2000 through 2003; the research office at the time, led by Michael Camp and Larry Cox, was a major source of strategic assistance to the program. The Kauffman Foundation sponsored the transfer of the entire data collection program and files from the University of Wisconsin Survey Research Laboratory to the University of Michigan Institute for Social Research (ISR) and has generously supported the PSED

Table F.3 Institutions Involved in Funding the Entrepreneurship Research Consortium
Ewing Marion Kauffman Foundation
National Science Foundation
Funding Institutions
Babson College
Boston University
Canadian Council on Small Business and Entrepreneurship
Clemson University
Coleman Foundation
EIM Small Business Research & Consultancy
Entrepreneurship Theory and Practice
Greek Entrepreneur's Club
Indiana University
Finnvera plc.
NFIB Education Foundation
New Jersey Institute of Technology
NUTEK (Swedish Board of Industrial and Technical Development)
Rensselaer Polytechnic Institute
RISEbusiness (Research Institute for Small and Emerging Business)
RWI/Essen (Rheinisch-Westfälisches Institut für Wirtschaftsforschung) Swinburne University of Technology
Syracuse University
Temple University
University of British Columbia
University of Cincinnati/Xavier University
University of Colorado, Boulder
University of Houston
University of Michigan
University of Missouri-Kansas City
University of North Carolina at Chapel Hill
University of Pennsylvania/Widener University
University of Southern California
University of St. Thomas
University of the Pacific

through the third and fourth waves of data collection in 2003. The shift of all the data and research files, the reconstruction of the project, and the comprehensive documentation developed at the University of Michigan were largely due to the experience and skill of Richard Curtin, director of the Surveys of Consumers program. In addition to support for data collection, Kauffman provided support to Kelly G. Shaver for the development of programs and processes for selecting different types of nascent entrepreneurs (Shaver, Carter, Gartner, & Reynolds, 2001), to the University of Southern California for the development of an initial “user friendly” data set called “PSED LITE,” and to William B. Gartner and the University of St. Thomas (to fund Nancy M. Carter) in the preparation of The Entrepreneur Next Door, an executive report of the initial findings of the PSED (Reynolds, Carter, Gartner, Greene, & Cox, 2002).

From the initial inception of the ERC, Babson College has served as the host institution, acting as the steward of the ERC funds in the early years of the project. In addition, Babson College provided funding that allowed Paul Reynolds, as the Paul T. Babson Chair of Entrepreneurship, to devote the majority of his time to service as the coordinating principal investigator for the life of the ERC and the duration of the PSED project.

Beginning in 1999, the Lloyd Grief Center for Entrepreneurial Studies hosted a series of all-expenses-paid workshops titled the Greif Research Symposium on Emerging Organizations to bring PSED scholars to the University of Southern California to encourage analyses of the PSED data sets and share research results. It was during the 2001 Greif symposium that the idea of a handbook on theory underlying the PSED research effort was first discussed. The 2002 Greif symposium was used to develop the outline for the book and the process for involving PSED scholars as contributors.

We are particularly indebted to Howard Aldrich, Candy Brush, and Patricia Greene for their involvement in the development of the handbook and for their willingness to provide chapters on topics that helped round out the handbook's breadth and comprehensiveness.

The publication team at Sage has significantly improved the quality of the book we submitted to them. Al Bruckner, senior acquisitions editor at Sage, immediately saw our vision for a book on the theory of the PSED and did a wonderful job of guiding us forward through the handbook's development. MaryAnn Vail has adroitly shepherded us through the publication process from author chapters to the completion of the book with expert advice and steady guidance that has made the process easy and smooth. Diane Foster evened out all of the inconsistencies in our efforts and brought forth a very coherent organizing framework and process for the book's completion. Robert Holm has done a magnificent job of melding the many different writing styles of our authors into a very readable text.

Finally, we recognize those entrepreneurs who are helping to modify and improve the economy: the 65,000 respondents who completed the PSED screening, the 1,300 respondents who provided detailed data, and the millions of future entrepreneurs who may benefit from a more complete understanding of the start-up process.

References
, , , , & (2002). The entrepreneur next door: Executive report of the Panel Study of Entrepreneurial Dynamics. (Available from the Ewing Marion Kauffman Foundation, Kansas City, MO)
, , , & (2001). Who is a nascent entrepreneur? Decision rules for identifying and selecting entrepreneurs in the Panel Study of Entrepreneurial Dynamics (PSED). In , , , , , , & (Eds.), Frontiers of entrepreneurship research 2001: Proceedings of the 21st Annual Babson Kauffman Entrepreneurship Research Conference (p. 122). Babson Park, MA: Babson College.
• ## Conclusion

…Only in departure whole. Arrival is always partial.

—Bill Knott (2004)

With our combined accumulation of over 100 years of knowledge and experience in scholarly activities devoted to research on entrepreneurs and entrepreneurship, and now with our nearly 40 cumulative years involved with the Panel Study of Entrepreneurial Dynamics (PSED), we are in a position to see that where we began then is not where we thought we would end up today, or for that matter where entrepreneurship scholarship is likely to take us tomorrow. In other words, the research process is by its nature in a constant state of flux. Like the start-up businesses we seek to study, our preliminary business plan for this project bears only a family resemblance to the enterprise that has developed. Paraphrasing Albert Einstein, “If we knew what we were doing, we wouldn't call it research.” The word research, itself, invokes an intrinsic need to “re-search,” to go back and look again at what occurred. This book is both the culmination of attempts to document the theory underlying the PSED and the foundation for its further exploration. Through use of the theory provided here, or through the application of other theory using the data our theory generated, future investigators can generate new knowledge and understanding about the phenomenon. This insight recognizes that the development of the PSED as a significant milestone along the path to a better understanding of the process of business creation. But, we do not believe we have reached the end of the journey. Even if all of the empirical tests of the theory presented in this handbook were carried out, we would still have achieved only a partial exploration of all of the information that the PSED contains.

We end this book—much as we began it—with an invitation to our readers to participate in an exploration of a unique and wonderfully rich, complicated, and comprehensive data set on the process of business creation. We believe having some familiarity with the information in this handbook is critical for anyone who intends to use the PSED. Embedded in many of these chapters are critical insights about how specific questions are linked to particular theoretical ideas, and warnings about how misapplications of the use of these questions could result in erroneous findings. These chapters then, are much like “soundings” that test the depth of the waters when sailing in unknown seas. Before diving into the PSED, a reading of these chapters will help provide a sense of how deep the water really goes. There is no reason to get “stuck in the mud” on some particular topic issue when others have made significant efforts to mark the channels.

We also recognize that although the theory presented here offers ways to understand and explore aspects of the process of business creation, the accumulation of questions across all of these various theoretical perspectives provides a variety of unintended ways in which new theory about entrepreneurship and new applications of these new theories can be used to explore the PSED. It would not be facetious to suggest that there is probably a question asked in one of the surveys that would be useful for addressing other theories (your theory, perhaps) that could be used to probe the new business formation process. But, please, pay attention to prior theoretical and empirical work that has already occurred using the questions you may be interested in using.

Although the handbook is focused on theory, the foundation of the PSED could be summed up in this aphorism: “Theory without data is speculation.” We believe that entrepreneurship scholarship is in need of more facts about what the phenomenon is, rather than more theory about what the phenomenon might be. The theory in the handbook provides direction for exploring the phenomenon of new business formation; yet these ideas are grounded in specific questions and specific facts resulting from these questions. The theory presented here can be tested. This is a significant advance for the entrepreneurship field. Offering an aphorism from Thomas Henry Huxley: “The great tragedy of science is the slaying of a beautiful hypothesis by an ugly fact.” The PSED is an ocean of ugly facts waiting for researchers to explore so that we can discover just how beautiful some of the competing hypotheses are.

Finally, the PSED is a collaborative research project. And, it is in this original, and still current, spirit of collaboration that we ask that scholars interested in using data from the PSED seek to participate in this active community of researchers. One of the implicit aspects of work in the entrepreneurship field is a realization that our efforts are not a “zero sum game.” The sharing of information about the database: insights into the nuances of particular cases, explorations of theories through factor analyses of combinations of questions, or any other work in progress toward the publication of results and findings helps benefit the entire field, as well as each scholar individually. We invite you to join us.

References
(2004, January 12). The answer. The New Yorker, 79 (42), 82.

## Appendix A

Where Do New Businesses Come from?

The origin of new businesses is the primary focus of the Panel Study of Entrepreneurial Dynamics (PSED). The design of the research is intended to document the underlying processes and factors that lead individuals to pursue the creation of a new business firm. The optimal strategy seemed quite straightforward: Locate individuals involved in the business start-up process and determine as much as possible about their initial situation and their plans and actions to create a new business. Periodic follow-ups would provide information on the outcome of their efforts. Implementation of such a complex focus has proven difficult but feasible.

Conceptualization of the Phenomena

To facilitate discussion of the start-up process, it has been convenient to conceptualize the entrepreneurial process as occurring within a political and economic context as indicated in Figure A.1.

The process is conceptualized as having four stages with three transitions. The first transition occurs when individuals, either alone or responding to an initiative of an existing firm, elect to pursue a firm creation. These persons are classified as nascent entrepreneurs (NE) if they represent an independent start-up effort, or, if they are sponsored by an existing business, the persons are considered nascent intrapreneurs (NI), a type of corporate entrepreneurship. Conception may be considered the beginning of this, the firm gestation process. The second transition in the life course of the firm occurs when the gestation process is complete, firm birth has occurred, and an infant firm is in place as an operating business. For many in the start-up phase, however, the next transition is to abandon the effort. The firm is, in a sense, stillborn. The third transition for infant firms is a passage into firm adolescence, a stage in which survival is considered more certain and not a constant challenge. A secondary feature of the third transition would be the nature of the growth trajectory of surviving new firms; a new firm pursing a high-growth trajectory may be considered to have a different character compared to those that are designed to persist as small firms.

Figure A.1 Conceptualization of the Entrepreneurial Process

Three of the critical features of the entrepreneurial process about which very little is known are represented by the question marks in Figure A.1: “?a” represents the proportion of business start-ups that complete the process with the implementation of an infant firm; “?b” represents the proportion of start-ups that never complete the process, although when they are abandoned is also a major issue about which little is known; finally, “?c” represents the features of the gestation process itself, both the length of time required to develop an infant firm and the activities that take place.

One primary objective of the research program was to provide systematic, reliable data on the basic features of the entrepreneurial or start-up process. This would include information on the nature and extent of variation in the critical aspects of the start-up process: the proportion of the adult population involved in firm conceptions, the activities that compose the start-up process, the proportion of start-ups that complete the second transition to become infant firms, and the survival and growth trajectories of the new firms.

A second objective was to provide reliable data on those factors or variables that could account for variations in these transitional events. This includes, for example,

variations in the number of start-up efforts that become operating new firms or the length of time spent in the gestation phase before the efforts are completed or abandoned. Potential influences would include, but are not limited to, the following (many are discussed in detail in the chapters of this handbook):

• Economic context, including national conditions, local conditions, and conditions in the economic sector or market of the new firm

• Personal context, including work and family situations and responsibilities, as well as the presence of supportive social networks for members of the start-up team

• Personal background and relevant experience, based on early family life, educational or occupational training, or in specific economic or market sectors for members of the start-up team

• Personal predispositions, either in terms of decision-making style, risk preferences, desire for autonomy, and personal aspirations

• Sociodemographic background, such as age, gender, race, and other characteristics

• Nature and sequence of start-up activities pursued in the firm gestation process

• Nature of the markets, competition for the new firm, and the strategic focus.

• Access to and use of resources, financial and otherwise

• Access to and use of programs designed to assist start-ups and new businesses

While assessing whether or not any one of these factors might affect the start-up process was one type of research objective, most past research in entrepreneurship has emphasized one or two of the factors from this list. A second objective was to estimate the relative impact of each of these factors holding constant the impact of all other factors. Such a comparative analysis requires that data be assembled on all factors for each start-up initiative. The major challenge in the design of the project was to collect data on all of these topics from the same set of respondents involved in starting new businesses. Such a comprehensive data set was required for the research program to provide a complete description and explanation of the critical elements of the entrepreneurial process.

Based on these objectives, and the constituencies to be served by the results, the final research design should have a number of features. First, a technique must be developed to provide systematic descriptions of the start-up process from its very beginning—the beginning is assumed to occur when somebody decides to implement a new firm, before any behavior occurs. Second, the data should be based on a representative sample of business start-ups so that estimates for the entire population—of people or firms—would be possible. Third, as the creation of a new firm appears to be very complex and multifaceted—with many small factors having direct and interactive influences—it is important that all major perspectives be represented in the data collection. This is tantamount to trying to measure a wide range of potential independent variables from a variety of perspectives. Fourth, as the focus is on the process of implementing a new business, the project would need to be designed as a longitudinal or panel study.

Design History

The distinctive methodological challenge was the creation of a procedure that would capture the very beginning of the business creation process. It was necessary to identify a start-up initiative before it had been included in business registries (such as the Dun & Bradstreet credit rating files) or lists of business activity (such as the phone book Yellow Pages). The objective was to create a procedure that could be used to systematically develop a sample that would allow estimates of the total amount of activity in the population.

Although later abandoned, the initial effort to locate a random sample of nascent entrepreneurs involved an application of multiplicative sampling. Developed with Charles Palit of the University of Wisconsin Survey Research Laboratory, this application required a systematic procedure to identify the social network of a representative sample of adults (Sudman, Sirken, & Cowan, 1988; Palit & Reynolds, 1993). This involved establishing a list of each respondent's parents, siblings, adult children, coworkers, and significant other (usually a spouse) during the interview. The respondent is then asked if any of these people are currently involved in a start-up effort. If they are, the respondent is asked for details so they may be contacted for a more complete interview.

While this may appear similar to snowball sampling—in which a respondent identifies other individuals with unique characteristics—it differs in one important respect. When persons nominated by the respondent as nascent entrepreneurs are interviewed, they are asked questions that will make it possible to determine the probability they would be nominated by more than one person in the initial sample. This allows the use of the information from the procedure to adjust for potential multiple nominations and compute the probability with which nascent entrepreneurs occur in the total population.

Although the implementation of the procedure involved the creation of relatively complete and very useful descriptions of the respondent's social networks, it was abandoned for several reasons. Most important, a person nominated as a nascent entrepreneur was actually interviewed in less than 40% of the cases. There were two main problems. In 20% of the cases, the original respondents were not able to provide the full name and phone number for the nominated nascent entrepreneur—they did not have the information. In 40% of the cases, the original respondents refused to provide the full name and location information for the nominated nascent entrepreneurs. They were, after all, talking to a stranger on the phone and may not have fully believed that the interview was related to research. As a result, in only two of five cases was a network nominee called to be interviewed. A pretest effort completed in 1997 using one household member to nominate others in the same household for a nascent entrepreneur interview was also associated with very poor results; interviews were completed with very few—less than 10%—of the within-household nominees.

Several factors led to the change in procedures from multiplicative sampling to straightforward population screening. First, as discussed above, the multiplicative sampling procedure was not working well—very few of those nominated were interviewed. Second, prevalence rates of nascent entrepreneurs were four to eight times higher than the 1% to 2% initially expected. Third, it was possible to locate a commercial marketing research firm that could include two appropriate screening questions in a national survey so that the costs and benefits of identifying each nascent entrepreneur was attractive compared with other methods. The procedure amounted to asking all respondents if they were starting a business on their own or for their employer. Other questions were then used to eliminate those not considered active in a start-up, not potential owners, or not in the start-up phase itself.

The first full application of the research procedure was with the adult population of Wisconsin in 1992 and 1993 (Reynolds & White, 1993, 1997). A second application involved the purchase of time in the monthly Survey of Consumers completed by the University of Michigan Institute for Social Research in October and November 1993 (Curtin, 1982; Reynolds, 1997). These two studies provided similar results regarding prevalence rates (about 4% were identified as nascent entrepreneurs), demonstrated the technical feasibility of the research protocol, and indicated that costs would be high but affordable. In the first case, the initial screening was part of a special-purpose survey. In the second project, the screening was one part of a multipurpose project, in which many costs were shared with other research projects.

PSED Design

The research design for the U.S. Panel Study of Entrepreneurial Dynamics (PSED) is presented in Figure A.2. It has three components and reflects data collection from two types of respondents. The first stage involves large-scale screening to create two samples representative of the U.S. population of adults 18 years old and older, excluding residents of Alaska and Hawaii. The first sample included those involved in attempting to start a new business. These respondents are either autonomous start-ups, referred to as nascent entrepreneurs (NEs), or sponsored by an existing firm, referred to as nascent intrapreneurs (NIs). Second, a representative sample of typical adults to be used as the comparison group (CG) was identified. Both types had to meet certain criteria and be willing to participate in subsequent interviews. Once the screening procedures were completed, the second stage of data collection involved detailed phone interviews followed by self-administered questionnaires mailed to the respondents. The third stage was the follow-up phone and mail interviews completed to determine the outcome of their efforts to implement a new firm. The initial design included plans for 12- and 24-month follow-up interviews. Additional funding has allowed for 36-month follow-ups to be completed in 2003.

This design involved optimizing a number of desirable, though often incompatible, features. Among these issues were choices between the sample size versus the amount of information assembled from each respondent, the scope of information to be included in the interviews versus the desire to keep respondents involved over multiple data collection activities, which items were best suited for the phone or self-administered mail questionnaire, and the simplicity of the interview items versus the complexity of the research concepts. There is, of course, no single best solution to this optimization problem. The design of this research program was the result of both technical issues and the need to provide data for use by over 100 scholars and researchers in the Entrepreneurship Research Consortium. A political process led to an acceptable solution for this group of scholars. Another research team might have developed another, equally appropriate, solution.

Figure A.2 PSED Research Design Overview

The research procedure involved the national screening, the initial round of data collection through phone and mail interviews, and the follow-up interviews. The PSED data set consists of five related samples:

• NE representative sample drawn from the original screening of the U.S. population (referred to as NE Mixed Gender)

• NE female oversample drawn from supplemental screening of the U.S. population with only females retained for sample (Carter, Brush, Aldrich, Green, & Katz, 1998)

• NE minority oversample drawn from the original screening of the U.S. population with only Blacks and Hispanics retained for sample (Greene, Carter, Reyolds, Aldrich, & Stearns, 1999)

• CG representative sample drawn from the original screening of the U.S. population of those not involved in entrepreneurship for a comparison group (referred to as CG Mixed Gender)

• CG minority oversample drawn from a screening of the U.S. population of minorities not involved in entrepreneurship for a comparison group (Greene et al., 1999)

The dates of the initial screening for each sample and the number of interviews involved at each stage of data collection are presented in Table A.1. Note that data collection for both the minority nascent entrepreneurs and minority comparison groups started 12 months after the other screening activities. So the summer 2003 period involves the third follow-up for the mixed gender and female nascent samples and the second follow-up for the minority nascent cohort. This is why no third wave follow-up data for the minority nascent entrepreneur sample will be collected.

Table A.1 Sample Selection: Cohort Size, Screening Dates, and Follow-up Samples
Screening Respondents: Basic Procedure

National screening of the adult population was initiated in 1998 by a commercial market research firm [TeleNation Program, Market Facts, Inc. (now Synovate); Arlington Heights, IL] that surveyed three random samples of 1,000 adults each week in the contiguous 48 states and the District of Columbia. Random digit dial (RDD) sampling procedures avoid the problems of high percentages of households with unlisted phones to create representative samples of private households. The phone numbers themselves are randomly created. Based on TeleNation procedures, once a residential living unit was contacted, the first individual aged 18 or older that would complete the phone interview was accepted as a respondent. Quota sampling was used to ensure that half of each sample was men and the other half women. Each sample wave of 1,000 was completed in a 3-day period, with a three-call criterion (initial call and two call-backs). However, up to 2% of the respondents were called from four to nine times to complete an interview. The interviews were controlled to be less than 30 minutes long to minimize midinterview terminations.

For locating eligible nascent entrepreneurs, an equal number of samples were screened on weekdays (Monday, Tuesday, and Wednesday) and weekends (Friday, Saturday, and Sunday). The yield of eligible nascent entrepreneurs was actually about 0.5% higher for the weekend samples. Between questions about marketing issues and consumer preferences, two items were inserted to determine if a respondent might qualify as a nascent entrepreneur:

• Are you, alone or with others, now trying to start a new business?

• Are you, alone or with others, now starting a new business or new venture for your employer? An effort that is part of your job assignment?

About 88% responded “no” to both items and were not involved further. Those that answered “yes” to either (6.9% to the first or 3.6% to the second) or both (1.2%) of these items were considered candidates for the nascent entrepreneur interview if they met three criteria:

• They expect to be owners or part owners of the new firm.

• They have been active in trying to start the new firm in the past 12 months.

• The effort is still in the start-up or gestation phase and is not an infant firm.

The first two criteria were included as part of the initial screening interview procedure. The third criterion was incorporated in the next phase of the data collection.

Those that expected to be owners of the new firm and had been active were invited to participate in “a national study of new businesses being conducted through the University of Wisconsin” and told a cash payment would be provided. From 86% to 87% of those that met these criteria provided their first name. Their first name, along with their phone number, was then provided to the University of Wisconsin Survey Research Laboratory in Madison, Wisconsin, the unit responsible for the initial waves of the data collection.

A similar procedure was used to identify candidates for the comparison group, except that all respondents in the sample were offered a chance to participate in a “study of the work and career patterns of all Americans, including those not currently working.” Again, a cash payment was mentioned as an inducement. In this case, 62% to 72% agreed to provide their first name, which was forwarded, along with their phone number, to the University of Wisconsin Survey Research Laboratory.

In addition to providing candidates for the nascent entrepreneur cohort and the comparison group, the resulting data set includes basic sociodemographic information on the respondents and their household, as well as the county and state in which the phone was located. Data on those 90% screened but not qualifying as nascent entrepreneurs were used in computing the population prevalence rates, a “screening comparison group” of about 60,000 individuals.

Completion of Nascent Screening: The Third Criteria

The first names and phone numbers of potential candidates for the nascent entrepreneur interview were relayed to the University of Wisconsin Survey

Research Laboratory. They were then assigned for a phone interview. If the respondent was involved in several start-up efforts, they were asked to focus on only the most recent start-up effort. (Up to one third reported simultaneous participation in several start-ups.) The third criterion was reflected in a series of four questions to determine the following:

• Has the start-up had a positive monthly cash flow that covers expenses and the owner-manager salaries for more than 3 months?

If the answer was yes, the activity was considered an infant business and not a startup effort. In these cases, the respondents were thanked for their time. As this takes less than 5 minutes on the phone, they are sent a token payment of $5 and dropped from the study. Approximately one fourth (27%) of the respondents failed to meet the selection criteria and were dropped at this stage because they were considered too far advanced in the start-up or gestation phase of the entrepreneurial process. This represented the ambiguity associated with the phrase “trying to start a new business.” Initial Detailed Interviews Among the remaining respondents, it was not possible to locate about 7%. Another 20% of the remaining group would not or could not complete the phone interview. Hence, the cooperation rate was 80% among those that could be contacted and 71% among those that were eligible, whether or not they were contacted. The complete phone interview took an average of 60 minutes to complete. At the completion of the interview, the respondents were sent a check for$25.

A similar procedure was followed with the comparison group. For the full population comparison group, only a randomly selected subset of respondents was taken from those that volunteered during the national screening to save costs. As most nascent entrepreneurs are under 55 years of age, those over 55 were sampled at one third the rate of those under 55. For the minority comparison group, all eligible respondents were incorporated in the procedure. The comparison group phone interview took about 25 minutes to complete and respondents were also sent a check for $25. At the completion of the phone interview, all NE and CG respondents were asked to complete a brief self-administered questionnaire for a second payment of$25. Ninety-eight percent agreed to consider completing the self-administered questionnaire. After repeated postcard reminders, mailings, and phone calls, from 53% to 71% of the nascent entrepreneurs and 76% to 83% of the comparison group respondents returned the mail questionnaire. These follow-up calls were terminated 240 days after the completion of the phone interviews when it was discovered that those that returned the mail questionnaire tended to do so within 40–50 days; 75% were received within 2 months.

The initial phone interview with the nascent entrepreneurs contained a wide range of questions, which took 60 minutes to complete. The major conceptual domains included in the nascent entrepreneur and comparison group interviews are presented in Table A.2. Full detailed interview schedules are available on the project website (http://projects.isr.umich.edu/PSED).

Table A.2 Initial Interview Topics: Nascent Entrepreneur and Comparison Group
NE ScheduleCG ScheduleTopic
Phone Interview
XIntroductory conversation on start-up, reasons, expectations
XStart-up activities
XFirm registration activities
XNature of start-up effort: legal form, economic sector, etc.
XX (selected)Start-up team: composition, background, and contributions
X(start-up related)X(career related)Social network: scope, background, and contributions
XStart-up funding: requirements, and expectations
XAssessment of market, competition
XCompetitive strategy
XKnowledge, use of assistance programs
XFuture expectations for the new firm
XXPersonal decision-making style
XXCurrent labor force activity
XXWork, career experiences
XXResidential tenure, migration (R and parents)
XXRespondent birth order
XXHousehold structure
XXHousehold income
XXHousehold net worth
XReaction to participation
Mail Questionnaire
XOpportunity recognition, information gathering assessment
XXEntrepreneurial climate scale
XStart-up problems
XEconomic sector, community context assessment
XFinancial management expectations
XXWork, training background of respondent
XItem inventory: Reasons for starting a new firm
XXAssessment of risk preferences
XXPersonal work background details
XXIndividual problem-solving orientation
XXSelf-assessment inventory I: Work and start-up orientation
XXSelf-assessment inventory II: Generalized personal domains
XXTime-use diaries: Recent work day and day off
XXWork participation history, previous 11 years
Operational Outcomes: Respondent Cooperation and Response Rates

No survey data set is perfect, but some are more complete than others. Completeness reflects the cooperation of the respondents in terms of participation in the project as well as cooperation during the interview procedure. Because of the need to collect data over an extended period of time, a high level of respondent trust and cooperation is desirable. This seems to have occurred in this project.

The basic features of the participant processing mechanism, along with indicators of respondent cooperation, are presented in Table A.3. The first column represents the results for mixed gender and female only nascent entrepreneurs, the second for the minority nascent entrepreneurs, the third for the total population comparison group, and the fourth for the minority comparison group. The first-section indicates the total counts of individuals involved at the different stages of the first round data collection process, from the thousands involved in the national screening to the hundreds completing and returning the mail questionnaire. This gives some idea of the scope of the challenge of keeping track of every individual involved, whether or not he or she is in the full data set.

The second section summarizes the level of cooperation at the different stages of the project. Potential nascent entrepreneurs were more likely to volunteer for the project than those in the comparison group, 86% to 87% versus 62% to 72%. Those in the comparison group, however, were more likely to complete all of the data collection procedure. Quite simply, the reduced cooperation from the nascent entrepreneurs probably reflected the severe time pressures on those trying to start new firms, compared to those in the comparison group.

The time and effort required to obtain completed phone interviews is indicated by the time lags between the initial screening and the phone interview, which average 41 to 62 days, with a maximum of 250 days. Further, the number of contacts required to obtain the detailed phone interviews averages seven to eight for nascent entrepreneurs and four to five for the comparison group, with a maximum of 74. Twenty-five percent of the nascent entrepreneur phone interviews required more than 9 to 10 calls, and 25% of the comparison group phone interviews required more than 5 to 7 calls. The effort required to obtain the self-administered questionnaires was reflected in the lag between completion of the phone interview and receipt of the mail questionnaire, which averaged 38 to 53 days, with a maximum of 471 days.

Reactions of the respondents were measured in several ways. Nascent entrepreneurs were asked at the end of the phone interview how the experience affected their interest in starting a new firm. As shown in Table A.3, 59% to 75% said it increased their interest, 25% to 39% said it had no effect, and 1%, 8 out of 830, indicated that it reduced their interest in starting a new firm. In fact, the positive effect may cause some problems, for some could claim that participation in the project may increase interest and, because of the content of the interview schedules, enhance the business knowledge. This may improve their chances for business success. In a sense, two “research effects” may be canceling the overall impact of the project. The Heisenberg effect in research refers to the impact of collecting data on a phenomenon; observations may absorb energy from the process under study.

Table A.3 Indicators of Respondent Cooperation

The Hawthorne effect refers to the increase in work group productivity (or other task measures) that is known to follow if the group knows it is the subject of a research effort. Hopefully, these effects offset any impact from the project and led to a neutral impact across the respondents.

A random sample of 5% of the mail questionnaire items (19 in the nascent entrepreneur questionnaire, 16 in the comparison group questionnaire) indicates that 96% of the respondents completed the items. The mail questionnaires that were returned were almost always completely filled out.

In survey research, it is well known that the hardest information to gather are details regarding household financial status. The long-running General Social Surveys (Davis & Smith, 1996) find that 97.5% of the adult respondents will report their intent to hide a defect in a used car they wanted to sell, 99.7% would report a bankruptcy, 99.4% would report a nervous breakdown and 99.6% treatment for a mental disorder, 99.5% illegal drug use, and 98.8% would report on the number of sexual partners they had in the last 12 months; but only 91.1% will report household income. The completion rates for items regarding household income and net worth in the PSED phone interview are presented at the bottom of Table A.3. From 90% to 97% of the nascent entrepreneurs provided information on both items; 97% to 99% of the comparison group members have provided the requested information. This is to be compared to the 77.7% that provided household income in the original screener completed by Market Facts or the 76.4% to 82.8% in the 1966 National Household Education Survey (U.S. Department of Education, 1997, p. 47). In terms of respondent cooperation in survey research in the United States at the end of the twentieth century, the PSED ranks among the best.

Why were the cooperation rates and interview schedule completion rates so high? A great deal of effort, involving the interviewers themselves, was devoted to modifying and adjusting the phone procedures and wording so that the entire research staff felt comfortable with the project and the procedures. Their commitment to the project and confidence in the interview procedure was transmitted to the respondent who, in turn, reciprocated with cooperation and trust. This is, obviously, not a trivial accomplishment and reflects the sustained dedication of the University of Wisconsin Survey Research Laboratory staff.

Follow-up Data Collection

The critical dependent variable associated with the project is the tracking of the start-up initiatives to determine the outcome of these efforts.

This longitudinal study was threatened when the project was to be transferred to a new institution. In the midst of the first 12-month follow-up data collection, the University of Wisconsin made an administrative decision to close the Wisconsin Survey Research Laboratory. With financial support from the Ewing Marion Kauffman Foundation, the entire data collection mechanism, all records identifying individual respondents, and all data files were transferred to the University of Michigan's Institute for Social Research (ISR) under the supervision of Richard Curtin. Although both units focus on creating clean data and high response rates, the operational procedures are slightly different.

The timing and responsibility for these efforts is presented in Table A.4. The comparison groups are omitted as no effort was made to follow their progress after the initial interview. This information makes clear, once again, that the nascent entrepreneur minority sample was initiated 12 months following the mixed gender and female cohorts; no third follow-up is anticipated for this subsample.

Table A.4 Follow-up Data Collection: Timing and Responsibility
Follow-up Interview Schedules

The follow-up data collection included both phone and mail components. Most critical was the single item that specifies the current status of the start-up effort. Previous studies indicate that four alternatives are commonly reported (Carter, Gartner, & Reynolds, 1996; Reynolds & White, 1997, p. 68):

• Active start-up, continued efforts to implement a new firm.

• Dormant start-up, no current efforts underway, but start-up has not been abandoned.

• Abandoned start-up, not successful and no further efforts are expected.

• Going concern, the start-up has become an infant business.

A different set of questions was asked depending upon the current status of the start-up effort. All respondents, however, received the same self-administered mail questionnaire, which was a reduced form of the mail questionnaire used in the initial data collection.

The nascent entrepreneur interview completed at 12 months is complicated by the different outcomes commonly reported. Table A.5 provides an overview of the material covered for each of the four options. Those that report a dormant start-up effort or one that has been abandoned (all members of the start-up team have quit working on the start-up) had a much shorter interview that those reporting infant firms or an active start-up effort.

Two other issues complicated the follow-up interview.1 First, a small proportion of start-up efforts changed the focus of their business. This was, however, relatively easy to accommodate in the follow-up phone interview by asking appropriate questions about any change in activity before continuing with the rest of the interview.

But there were a few cases among the half of the start-up efforts that were team initiatives in which the original respondent was no longer involved in the start-up but others on the team were still working on creating a business, sometimes with success. This occurrence reflected a major conceptual issue for the project: whether it is the study of (a) individuals trying to start firms or (b) firm start-ups, located by sampling individuals. (The average team size is slightly more than 2.) The procedure was to ask the respondent for the name and phone number of a current member of the start-up team so that information on the firm may be obtained from an active participant. While the project was generally successful in terms of getting the data on “replacement informants,” a lack of funds precluded any attempt to contact these individuals. These items were eventually dropped from the third follow-up interviews in year 2003.

Follow-up Data Collection: Operational Results

The combined results for the first follow-up interviews with nascent entrepreneurs are shown in Table A.6. First round follow-ups for the mixed gender and

Table A.5 Interview Topics: Nascent Entrepreneur Follow-Up Interviews

female oversample cohorts was completed by the University of Wisconsin Survey Research Laboratory, whereas the first-round follow-ups for the minority over-sample cohort were completed by the University of Michigan Survey Research Center.

Table A.6 First Follow-up Outcome Summary
OutcomeCase CountPercentage
Respondent reports an operating business18722.5
Respondent still active in start-up process19223.1
Respondent inactive in start-up process but expects to “get to it some day”12314.8
Respondent has quit or disengaged11313.6
Cannot locate or no useful information obtained21525.9
Total830100.0

It is, of course, desirable to reduce the number that cannot be located, about 26% or 215 individuals in this situation. These cases are a combination of those who could not be found to get a useful response (usually about 10%, three deaths were reported) and those that did not respond—usually by delaying the interview (by constantly rescheduling the interview appointment until the field period was terminated).

The results of the first and second follow-up efforts are combined and presented in Table A.7. The first set of columns summarizes the results of both initial follow-ups. The top section is the mixed gender and female oversamples, the middle section the minority oversample, and the bottom section combines the counts for all samples. These are unweighted raw case counts. The right columns summarize the results of the second follow-up of the mixed gender and female oversample.

The complexity of the table is due to the second follow-up effort and the phenomena itself. The University of Michigan made an effort to contact respondents not located in the initial follow-up by the University of Wisconsin. They were successful for 46 out of 168 of these cases (“no information” located in sections A:4 and B). On the other hand, they were not able to locate 62 respondents who completed the initial follow-up (section D). Further, it is of some interest that 26 of those reporting a going business in the first follow-up indicated in the second follow-up (row F) they were no longer involved, and another 5 reported that some part of a going business had been sold (section A:1). Further, no effort was made in the second follow-up to contact those 89 who reported quitting or disengagement in the first follow-up (row C).

This assessment demonstrates the complexity of the entrepreneurial process. It is clear that successful tracking of the outcomes of start-up efforts will require careful and dedicated survey research professionals. Fortunately, such individuals have been involved in the PSED.

Table A.7 First and Second Follow-up Outcome Summary

Operational Outcomes: Call-Backs and Prevalence Rates

It is quite clear, from this and other studies, that those trying to start new firms are among the busiest people in the country. Most are young adults in mid career with jobs and families, focusing on creating a new business. It would be no surprise to find they are hard to reach on the phone. The screening firm maintained a record of the number of calls required to complete an interview, and normally a potential household was dropped after three calls. However, the firm guaranteed their clients 1,000 cases per sample. When interview yields dropped, they would increase the call-backs to assure the quota of 1,000 respondents per wave; this was more cost-effective than drawing an additional sample on short notice. This operational data was provided on 53 of the samples. The prevalence rate of nascent entrepreneurs could be determined as a function of the number of calls made to complete the interview. The results are presented in Table A.8.

The results indicate that if all sample points were routinely subjected to a 10 callback operational criteria, the nascent entrepreneur prevalence rate would be somewhat higher, perhaps by as much as 20%. Although rare for commercial market research firms, call-back standards of 20, 30, or even higher are common in academic- and policy-oriented survey research projects. This pattern suggests that all prevalence rate estimates may be conservative.

Table A.8 Calls to Complete and Nascent Entrepreneur Prevalence Rates

On the other hand, once they start an interview, nascent entrepreneurs are hard to stop, as some talked for up to 90 minutes about their favorite subject, their new business.

Commentary

The procedures developed in this research program to locate and identify those active in the business start-up or entrepreneurial process have had very wide applications. The original panel design has been implemented in Argentina (de Rearte, Lanari, & Atucha, 1998), Canada (Menzies, Gasse, Diochon, & Garand, 2002), Greece and the Netherlands (Wolters, 2000), Norway (Alsos & Kolvereid, 1998) and Sweden (Delmar & Davidsson, 2000). In addition, this procedure has been the technical core for a large-scale, cross-national assessment of entrepreneurial activity. The Global Entrepreneurship Monitor (GEM) project has modified and extended the procedure to encompass identification of those in both the start-up process and managing new firms up to 42 months old (Reynolds, Hay, & Camp, 1999; Reynolds, Hay, Bygrave, Camp, & Autio, 2000; Reynolds, Camp, Bygrave, Autio, & Hay, 2001; Reynolds, Bygrave, Autio, Cox, & Hay, 2002). The procedure has been improved with the addition of a third screening item: respondents are asked if they are owner-managers of a going business. It was found in the PSED that a substantial proportion that considered themselves a start-up were actually operating a going business and did not qualify as nascent entrepreneurs. In a similar fashion, a substantial number of those who identified themselves as owner-managers of a going business in the GEM surveys turned out to have never paid any salaries and were clearly in the start-up phase (Reynolds & Hunt, 2001).

The use of the PSED screening procedures as part of the GEM project has led to the realization that start-up activity has reached an unprecedented global scope. As of 2002, close to half a billion adults in a world of 6 billion may be actively engaged in starting new businesses (Reynolds et al., 2002). The total amount of human effort and financial resources devoted to entrepreneurship is already enormous and is growing each day. Such an outcome was never anticipated when the initial design was being developed in Wisconsin in 1992.

Notes

1. The Swedish National Panel Study of Business Start-Ups completed a 6-month follow-up prior to the design of the 12-month follow-up for the U.S. study. The material provided by Per Davidsson and Henrick Hall, (Jönköping International Business School, Jönköping, Sweden) was very useful in developing the follow-up phone interview schedule.

References
, & (1998). The business gestation process of novice, serial and parallel business founders. Entrepreneurship Theory and Practice, 22 (4), 101–114.
, , , , & (1998). The influence of founder's gender in business start-ups. Proposal to the National Science Foundation, Grant SBR-9809841.
, , & (1996). Exploring start-up event sequences. Journal of Business Venturing, 11 (3), 151–166. http://dx.doi.org/10.1016/0883-9026%2895%2900129-8
(1982). Indicators of consumer behavior: The University of Michigan survey of consumers. Public Opinion Quarterly 46, 340–362.
, & (1996). General social surveys, 1972–1996: Cumulative code book [National Data Program for the Social Sciences Series, no. 15]. Chicago: National Opinion Research Center.
, & (2000). Where do they come from? Prevalence and characteristics of nascent entrepreneurs. Entrepreneurship and Regional Development, 12, 1–23. http://dx.doi.org/10.1080/089856200283063
, , & (1998, September). El proceso de creacion de empresas; Abordaje methodologico y primeros resultados de un studio regional. Argentina: Universidad Nacional de Mar del Plata.
, , , , & (1999). The influence of founder's race in the start-up process. [Grant SBR-9905255] Proposal to the National Science Foundation.
, , , & (2002). Nascent entrepreneurs in Canada: An empirical study. San Juan, PR: ICSB 47th World Conference.
, & (1993). A network sampling procedure for estimating the prevalence of nascent entrepreneurs. Proceedings of the American Statistical Association International Conference on Establishment Surveys, pp. 657–661.
(1997). Who starts new firms? Linear additive versus interaction based models. Small Business Economics, 9, 449–462. http://dx.doi.org/10.1023/A:1007935726528
(2000). National panel study of U.S. business start-ups: Background and methodology. In (Ed.), Advances in entrepreneurship, firm emergence and growth: Vol. 4 (pp. 153–228). Stamford, CT: JAI.
, , , , & (2002). Global entrepreneurship monitor: 2002 executive report. Kansas City, MO: Kauffman Center for Entrepreneurial Leadership.
, , , , & (2001). Global entrepreneurship monitor: 2001 executive report. Kansas City, MO: Kauffman Center for Entrepreneurial Leadership.
, , , , & (2000). Global entrepreneurship monitor: 2000 executive report. Kansas City, MO: Kauffman Center for Entrepreneurial Leadership.
, , & (1999). Global entrepreneurship monitor: 1999 executive report. Kansas City, MO: Kauffman Center for Entrepreneurial Leadership.
, & (2001). Global entrepreneurship monitor: Data collection-operations manual: Vol. IIA Adult population surveys [working paper]. London: London Business School.
, & (1993). Wisconsin's entrepreneurial climate study. Milwaukee, WI: Marquette University Center for the Study of Entrepreneurship.
, & (1997). The entrepreneurial process: Economic growth, men, women, and minorities. Westport, CT: Quorum Books.
, , & (1988). Sampling rate and elusive populations. Science, 240, 991–996. http://dx.doi.org/10.1126/science.240.4855.991
U.S. Department of Education, National Center for Education Statistics. (1997). Unit and item response rates, weighting, and imputation procedures in the 1996 National Household Education Survey, [working paper no. 97–40 by J. Montaquila & J. M. Brick. Project officer, Kathryn Chandler]. Washington, DC.
(2000). Nascent entrepreneurship in the Netherlands: A glance behind the scenes of business start-ups. In (Ed.), Entrepreneurship in the Netherlands: Opportunities and threats to nascent entrepreneurs (Chapter 2, pp. 3–16). Zoetermeer, NL: EIM Small Business Research and Consultancy.

## Appendix B: Data Documentation, Data Preparation, and Weights

The analysis of data from the Panel Study of Entrepreneurial Dynamics (PSED) requires a good deal of information about the design of the research. The wording of questions and how the responses were coded is just as important as how to effectively combine the answers from the different samples while preserving the confidentiality and identity of the respondents. Whereas the prior appendix focused on the sampling procedures and interview schedules, this appendix details the procedures used to collect and document the data, the development of case weights, and the procedures used to protect the confidentiality and identity of respondents.

Without a theoretical foundation, mere numbers are meaningless. Most of the descriptive and theoretical questions addressed in the PSED require multiple questions to assess. As a result, the individual questions must be interpreted within the context in which they were asked, including both preceding and subsequent questions in each sequence. Although all responses were translated into a numerical code for analysis purposes, some codes have explicit meaning—say dollars or age in years—while other coded values were assigned by convention, say “1” for a “Yes” response. Even if the coded values do not have an explicit meaning, the simple tabulation of responses will be misleading unless the responses are properly weighted to account for the differences in sample designs and nonresponse. The proper use of weighted data is critical for developing inferences regarding the population represented by the samples.

Types of Questionnaires

There were seven basic questionnaires associated with different parts of the project:

• Screening Questionnaire: A short set of questions used to locate potential nascent entrepreneurs or candidates for the comparison groups

• Initial NascentEntrepreneur Questionnaire: Detailed questions asked in phone interview schedules of all eligible nascent entrepreneurs

• Initial Self-Administered Questionnaire: Mail interview schedules sent to all nascent entrepreneurs that completed the phone interview and agreed to complete the questionnaire

• Comparison Group Questionnaire: Initial phone interviews conducted with the comparison group respondents; a revised version of the nascent entrepreneur phone interview more appropriate for those not starting a business

• Comparison Group Self-Administered Questionnaire: Mail questionnaires used with the comparison group respondents; a revised version of the nascent entrepreneur mail interview schedule more appropriate for those not starting a business

• Follow-up NascentEntrepreneur Questionnaire: Phone interview schedules used in subsequent follow-ups with the nascent entrepreneurs

• Follow-up Self-Administered Questionnaire: Mail interview schedule used for all nascent entrepreneurs that completed subsequent interviews: a reduced version of the interview schedule provided to nascent entrepreneurs in the initial interviews

There are two forms of the phone interviews. The Computer Assisted Telephone Interview (CATI) schedules are relatively complicated computer programs using programming syntax. To facilitate analysis, a user-friendly form was prepared that looks like a paper and pencil interview and attempts to reflect all the features of the CATI version. The self-administered mail questionnaires are less complex and are only in one format.

Another critical document is the Codebook, which represents a compilation of the initial and follow-up data in one 450-page document. The Codebook provides a description of all variables in the data sets, the valid responses to each item, and the numeric values assigned to each response code.

These materials are much too lengthy to include in this handbook. Copies of the CATI phone interview schedules, the self-completed questionnaires, and the Codebook are available on the project Website (http://projects.isr.umich.edu/PSED) or on a CD provided by the sponsoring organizations.

Most serious researchers will produce a printed copy of the Codebook and interview schedules for reference; they fit easily in ring binders and two-sided printing is very useful for reducing the bulk. A separate binder for each facilitates simultaneous cross-referencing—reviewing the item in the interview schedule and then considering the pattern of the responses in the Codebook.

Interview Schedules

There are two types of interview schedules used in this project. The first is a phone interview during which a trained interviewer engages in a carefully constructed interaction with the respondent. These are completed with a CATI procedure during which questions appear on a computer monitor and are read to the respondent by the interviewer who then enters the responses directly into the computer. The questions asked of each respondent are adjusted based on their responses to prior questions. The computer program automatically performs consistency checks and provides immediate feedback to the interviewer in the form of clarification questions or probes (additional questions to resolve any ambiguity in the answer) so any inconsistencies can be resolved.

The second type of interview schedule is a self-completed questionnaire that was mailed to each respondent.

The screening interview is provided in Exhibit B.1 as an example of a phone interview schedule. The actual interview schedule is a sophisticated and complicated computer program. This version has been created to guide those planning to analyze the data. Most of the items are related to the personal and household situation of the respondent. Only seven questions are related to the PSED screening procedure, starting with item 1 (variable label BSTART): “Are you, alone or with others, now trying to start a new business?”

Note the important skip pattern that follows item 2 (BJOBST). This indicates that only if a respondent says “Yes” to either or both items 1 and 2 (BSTART and BJOBST) will they continue with item 3 (OWNER). If they say “No” to both questions, the program will exit the PSED section and the interviewer will be instructed to ask about the next topic of the interview. Those that stay in the section are then asked about ownership (OWNER) of the start-up and if they have been active, in the past 12 months (SUACTS). If they will not be owners or have not been active the interviewer exits the section and the respondent is not considered eligible for the detailed PSED interview. Those that qualify are given two chances to volunteer for the project (VOLUNT1, VOLUNT2) and those that agree to participate are asked to provide their first name only (NAMEVOL). All respondents, however, are asked the standard set of a dozen sociodemographic items regarding age, educational attainment, household structure, ethnic background, and so forth.

An example of one page from the self-administered questionnaire is presented in Exhibit B.2. These items were provided to all nascent entrepreneurs and all those in the comparison groups. The specific examples are concerned with preferences for different types of businesses, those with more certain payout but less risk versus those with higher risk and payouts. This is followed by items related to the previous work experiences of the respondent. Because a wide range of respondents with very diverse educational backgrounds must be able to read and interpret these items, substantial effort was devoted to making them as simple and as direct as possible.

Exhibit B.1 Example of Phone-Administered Interview Schedule: Screening (Variable labels in brackets)

Exhibit B.2 Sample Page from Self-Administered Questionnaire

Codebook Presentations

The beginning of all analysis should start with an examination of the unweighted frequency distribution of the variables under consideration. For example, the answers of 64,622 respondents approached in the screening process to the two basic items are presented in Exhibit B.3. This data is unweighted and combines the responses of all screened respondents, including those selected for the comparison groups. These questions were answered by almost all the respondents, the interviewers were instructed to get a “Yes” or “No” answer if at all possible. The few that gave “Don't know” and “Refuse” responses were recoded to “No” responses; thus all answers in the data set are either “Yes” or “No.” This practice of coding “Don't know” and “refuse” as “No” was confined to the screening data set as the other databases include specific codes for these responses.

Exhibit B.3 Codebook Variable Example: Initial Screening Items

An example from the self-administered questionnaire, item QH9, as provided in the Codebook is presented in Exhibit B.4. The letter “Q” has been added to the variable label to indicate the first wave of data collection. In this case, a total of 6 respondents did not provide useful information on this item compared to 899 that were able to answer the question. Again, both nascent entrepreneurs and comparison group respondents are combined in this unweighted presentation.

To facilitate comparisons across the different waves of data collection—and consistent with the practice followed in other longitudinal studies—the answers to the same questions asked at different times are presented together. For example, respondents classified as nascent entrepreneurs were asked a number of questions about what they had done to get ready to start the business, including the question shown in Exhibit B.5, which asks about saving money to invest in the business.

Exhibit B.4 Codebook Variable Example: Mail Questionnaire Item
Exhibit B.5 Codebook Variable Example: Response Categories Across Waves

If they said “Yes” at any point, they were asked when this saving behavior began and were not asked the question in subsequent interviews. The results, taken from the most recent Codebook, indicate that 830 were asked this question in Wave 1 and 574 answered “Yes.” Only those that did not say “Yes” in Wave 1 were asked this question a second time in Wave 2. Apparently 99 could be located and interviewed in Wave 2, another 37 said “Yes” and 62 said “No.” Of the 62 that were contacted for the third interview, another 15 said “Yes.” (The Wave 2 and Wave 3 respondents may not be the same 62, as efforts are made at each wave to locate those that could not be contacted in the previous efforts.) Such a presentation has obvious advantages for understanding the response patterns associated with each item.

Some items in the phone interview and self-administered questionnaire involve a spontaneous comment or statement from the respondent. Less than a dozen are asked of all respondents and most have been coded. These might be the type of business being created, the person's primary occupation, or perhaps the country of birth. These are reviewed by a coding team and numbers assigned to represent different responses. The Codebook provides the number/answer correspondence for these open-ended items—as well as any issues that come up in the coding. For example, the type of business activity is coded as one of over 500 Standard Industry Classifications (SIC), and respondent occupation is classified as one of over 500 U.S. Bureau of Labor Statistics Standard Occupational Codes.

A large number, however, represent comments made when a respondent considered the responses to a fixed choice item as not appropriate to their situation. In such cases, the interviewer entered a sentence or two describing their reaction. Most of these “other” responses have not been coded.

Overview of Data Sets

There are two basic PSED data sets. The first data set contains the data collected in the screening interviews to locate eligible individuals for either the nascent entrepreneur or comparison group. The second data set contains the detailed information collected in the phone and self-administered interviews with nascent entrepreneurs and the comparison groups for all waves in the panel data set.

Screening Data Set

The screening data set contains information on 64,622 individuals. These interviews consist mostly of demographic items and the items used to determine if the respondent was a nascent entrepreneur and, hence, qualified for inclusion. In addition, because the state and county of each respondent is known, a number of county characteristics have been added to each record to provide a harmonized description of the context in which the respondent made decisions either to create a new business or not to get involved. The number and source of variables for the screening data file are provided in Table B.1.

Table B.1 Source and Count of Screening File Variables
SourceCount
Specific questions asked of respondents for project6
Transforms and recoding of PSED items10
Sociodemographic variables asked by research firm18
Transforms and recoding of sociodemographic items17
Variables characterizing the county of residence26
Transforms and recoding of county characteristics26
Variables for operational purposes: Identification number, weights, number of calls to complete, date of interview, sample cohort, etc.17
Total number of variables120

The result is a single data file that has 64,622 records and 120 variables for each case. Confidential information, first name, and the phone number have been deleted to maintain anonymity. All values for all variables are numbers, some representing specific responses (Yes or No), others ratio values (population density of the county).

The two most critical variables in the data set are the unique identification number assigned to all 64,622 respondents in the screening interviews (WAVEID02) and the unique identification number assigned to all 1,261 respondents on which additional data was obtained (RESPID).

Detailed Data Set

The detailed data set is substantially more complicated, as it reflects respondents from five different samples, data collected from two different procedures (phone and mail), as well as data collected for up to four different points in time using the same questions. The result is 3,910 variables (three data collection periods through the second follow-up) on 1,261 individuals. The data set begins with the 117 variables from the screening interviews, and additional variables are added as new data sets and variable transformations are completed.

For example, it turned out that the gender of many respondents completing the initial detailed interviews was either ambiguous or incorrect. The gender appeared to have changed from the screening interview to the first detailed interview or was recorded incorrectly. As a result, considerable effort by Professor Nancy Carter was required to clarify all gender ambiguous cases—often reviewing the operational comments of the interviewer (discussed in Chapter 2). This revised gender variable is called NCGENDER. Although USGENDER is accurate and appropriate for analysis of the screening data set, NCGENDER would be the preferred indicator for the detailed data set. Similar problems developed with specification of ethnic background—the classification from the screening interview might be different from that in the initial detailed data collection. A careful review and correction of discrepancies has been completed by Professor Patricia Greene; her work is reflected in the variable PGRACE (Chapter 3).

Further complications developed with regard to the exact nature of the start-up activity. The details about the ownership of the potential new firm assembled in the initial interview indicated a substantial proportion of the ownership would be by some other entity than a natural person, for example, ownership by a different operating business or a financial institution. Careful examination of the ownership patterns (done by Professors Kelly Shaver and Paul Reynolds) led to the creation of a variable, AUTONSU, which reflects the nature and extent of major ownership patterns among the start-up efforts, as presented in Exhibit B.6. The discovery that only 721 of 830 start-ups (or 87%) will be wholly owned by natural persons is an important finding in its own right.

Exhibit B.6 Codebook Variable Example: Constructed Variable—Start-up Type

Following the initial screening interviews, detailed data was collected from respondents on four different occasions. Each occasion involved a different set (phone and self-completed mail) of interview schedules. Because many of the same items were asked of the respondents in different data collection periods, the first letter of the variable names have been adjusted to provide a guide to the data collection administration. The mixed-gender and the female oversample are represented by the initial “Q” and follow-up data collection efforts by the letters “R,” “S,” and “T”. However, for the minority oversample, the original interview schedules were used for the initial detailed data collection (“Q”), but the first follow-up was done with the interview schedule used for the second follow-ups with the other cohorts (“S”); the second minority follow-up is completed with the third follow-up of the other two samples in which variable names begin with “T.” These allocations are presented in Table B.2.

The major consequence of this is that assessments that combine follow-up data from the mixed-gender and female cohorts and the minority cohort require a reallocation of minority data in a standardized form to produce a new set of variables. For most variables this can be easily done by changing the variable labels for the minority cohort to match the other waves (rename “S503” as “R503;” rename ‘T503” as “S503”). The items and response frequencies should be checked before these changes are implemented to make sure the wording and response alternatives are the same for both administrations of the questions.

Table B.2 Variable Names First Letter and Follow-up Stage by Cohort
Representative Samples and Respondent Weights

The design of the sample is based on two critical factors: the definition of the population of interest and the method by which elements in that population are selected. The definition of the population needs to be precise but could encompass nearly any collection of elements. Although this study is ultimately concerned with the population of nascent entrepreneurs, there is no comprehensive listing of such people that could be used to exhaustively define the population. Other alternative lists, such as new firms listed in commercial credit-rating registries (such as Dun & Bradstreet) were not suitable for the purposes of this project since only established firms are eligible to be included. The method selected was to define the population of all adults and then from that population precisely define the eligibility criterion for nascent entrepreneurs.

Controlled selection of population elements by probability methods is required for representative samples. Probability sampling designs require that each member of the population have a known nonzero probability of selection. An important characteristic of probability sampling is that it allows relatively few observations to generate the same results that would be obtained if interviews were conducted with the entire population. The match will not be perfect, but the accuracy of probability samples can be statistically assessed. Although members of the comparison groups were drawn in proportion to the entire population, the probability of selection for nascent entrepreneurs was the product of the overall selection probability times the probability of selection among the eligible nascent entrepreneurs.

Given a representative sample, the results drawn from the survey can be generalized to the entire population. For example, if the typical new firm is established about 12 months following the beginning of the start-up process, this finding can be generalized to the entire population of nascent firms. Importantly, a representative sample allows the calculation of standard errors of the estimates and confidence intervals. For example, assume that a representative sample of 2,000 is interviewed from a population of 200 million and that it is found that 80 individuals (4%) were nascent entrepreneurs. This would imply that the best population estimate of nascent entrepreneurs would be 4% of 200 million or 8 million in the total population. Moreover, the 95% confidence interval would be from 3.2% to 4.9% so that the population estimate would range from 6.4 million to 9.8 million nascent entrepreneurs.

In practice, no sample design is perfectly representative of the intended population. There is a wide range of potential errors in any survey. The most commonly cited are sampling errors due to the fact that not all members of the population are interviewed. With probability samples, these errors can be calculated to determine confidence intervals for any estimate, as the example above indicates.

This is not the only source of potential error. The most important of these are coverage errors due to the fact that samples do not include the entire population (such as the homeless or those without telephones). Although some of the problems of coverage errors can be avoided by simply redefining the population (for example, those living in dwellings with telephones), to an important extent there will always be some coverage bias. Good samples are designed to keep such coverage bias to a minimum, especially if the coverage errors would affect estimates of the population of interest.

Another type of error involves bias due to the question wording, the inability of respondents to articulate their answers, language, or hearing problems. Finally, some errors are due to respondent refusal to participate or the inability of the interviewers to locate the respondent. Unfortunately, there are no standard measures of the effects of these errors or any standard method to correct for any potential bias.

The sample was based on telephone ownership among households within the United States, excluding Alaska and Hawaii. Although landline phones were owned by approximately 96% of all U.S. households at the time this sample was selected, in more recent years the number of households with only cell phones has grown. All subsequent waves of interviews were conducted by contacting respondents on the phone of their choice so that the growing cell phone ownership has not had an impact on this study. Nonownership of phones is typically concentrated among lower income households, younger households, and those that recently changed residences. Those factors can be mitigated by the use of survey weights.

Most dwellings have several adult residents. If it were possible to select one adult resident at random for the interview, this would eliminate another potential bias in the sample. This is, however, relatively time consuming and expensive. A typical solution is to interview the first person 18 and older who can be reached by phone. This frequently leads to a higher number of female respondents, as women are more likely to be at home and answer the phone. This was the reason that TeleNation used gender as a control in the sample selection. Again, adjusting the weights assigned to the sample can help to mitigate this problem.

Sample Weights

Although the sampling procedures used were expected to exhibit no systematic bias, confidence that the sample represents the population increases if the sample favorably compares to a more precise description of the population provided from another credible source. Rather than simply comparing sample distributions to information contained in independent surveys, the typical procedure is to correct the sample distributions so that they match information contained in the U.S. census. Such a procedure would correct for any coverage bias as well as correct for any systematic bias due to nonresponse, panel attribution, or other reporting errors.

The four panels in Exhibit B.7 indicate how sample distributions are made identical to independent information. The example shows how a sample is made consistent with a population based on age and gender.

The first panel of Exhibit B.7, labeled “Sample,” assumes the selection of 100 respondents distributed by age and gender selected from a target population of 100,000. The second panel of Exhibit B.7, labeled “Population,” gives the known age and gender distribution in the target population of 100,000. The third panel of Exhibit B.7 provides the weights that transform the sample distributions into the population totals. For example, there are 5 females aged 18 to 24 in the sample but 10,000 in the population, so that each of these respondents represented 2,000 in the population. Other weight values are also the ratio of the population to the sample totals. Of course, the more important aspect of the weights is how they differ among the various population groups. For example, the largest weight difference is between the youngest and oldest respondent, indicating that the older respondents were more likely to be overrepresented and thus have the lowest weights. Alternatively, the youngest respondents were underrepresented and had the largest weights.

Although the use of the raw weights would provide a correct estimate of means and other sample statistics, it would not provide a correct estimate of the variances. Most statistical programs assume that the sum of the weights is equal to the actual cases and use that total to compute the estimates of variance and standard errors. In the example in Exhibit B.7, the sum of the raw weights equals 100,000—exactly equal to the total population. This would lead to much higher estimates of precision than justified by the size of the sample. In order to prevent this problem, it is common to normalize the weights so the average value is one, and, as a result, the sum of the weights equals the sum of the cases. For the Exhibit B.7 example, this requires dividing all weights by 1,000, to produce a sum of the weights (100) equal to the sum of the cases (100). The bottom panel of Exhibit B.7 shows the normalized weights for each cell, which average to 1.00. Examples of analysis using the weights are provided in Appendix C.

Initial Weights

In order to provide precise weights, it is important to have the same measures on both the sample respondents and on the total population. As screening samples are gathered in 3 days of phone interviewing and the data delivered to the clients within 48 hours of completing the field work, each sample is weighted in relation to the most recent Current Population Survey estimates for the United States (http://www.bls.census.gov.cps). This recurring U.S. census survey—done four times a year—is considered the most accurate and timely estimate of the U.S. population characteristics. In order to compute weights, the prevalence of sample cases in each of 160 cells from a four-way

Exhibit B.7 Example of Weight Calculation

table is determined. The table reflects age, gender, four regions of the United States, and household income. As household income is missing for about 25% of the cases, special adjustments were made for these households.

Although this has the advantage of providing timely data for the clients, the weights that result have a wide range of values. When the original weights for the 64 PSED screening samples are combined, the range is from 0.1 to 10—some respondents were given a weight that was 100 times others. The use of a four-dimensional table with 160 cells with a sample of 1,000 would lead to an average of five to six cases per cell. Due to random and uncontrollable events, there will be samples with few or no cases in some of these cells, which lead to extreme weights.

The major disadvantage of such a spread of weights is that it can add considerable variance to the data set. Efforts to find systematic relationships between variables may be less successful because of additional variation created by the weighting scheme.

University of Michigan Revised Weights

The University of Michigan was able to recompute the weights, making several adjustments in the process. First, the entire screening data set was treated as one sample of 64,622. Second, the Current Population Survey estimates were combined for the 2-year period in which the screening took place. Third, the sample population match was based on age, gender, ethnic background, and educational attainment. Both household income and educational attainment provide estimates of socioeconomic status, but there are many fewer missing values for educational attainment (1.8% vs. 23.7%) which reduced the need to estimate weights for cases with missing values. As a major objective for the project is a more complete understanding of ethnic differences, standardizing weights on ethnic background were used to compute the revised weights. The final set of University of Michigan weights was based on age, gender, ethnic background, and educational attainment as reported in the screening interview.

Both the original and new weights have an average value of 1.00 and provide almost identical point estimates, as shown in Table B.3 for the population age distribution.

The revised weights have several major advantages:

• Range is from 0.7 to 1.7, less than a factor of 3, compared to 0.1 to 10.0, a factor of 100.

• Estimated variation contributed to estimated percentages due to weighting was reduced from 34.2% to 4.5%.

Extreme values—when there is a large range of weights—can cause problems in detailed analysis as a single case may have dramatic effects on the results. Most analysis is designed to “explain” or “account for” variation in variables and assumes that the research and measurement procedures will not be a major source of variation; if variation due to research procedures can be reduced, it facilitates analysis, for the remaining variation is due to the phenomena itself. Hence, the revised weights provide substantial advantages and all weights in the data files are based on the new weights. The original weights—based on the weights assigned by the screening firm for each sampling wave—are not included in the data sets. In all cases, the weights

Table B.3 Impact of Alternative Weights on Screening Sample Nascent Entrepreneurs Age Distribution
AgeOriginal WeightsUniversity of Michigan Weights
18–24 Years12.62%12.99%
25–34 Years28.97%28.29%
35–44 Years30.01%30.19%
45–54 Years19.49%19.80%
55 years and older8.88%8.73%
Total99.97%100.00%
Table B.4 Weights in the Data Set
Variable LabelData SetDescription
WT_SCRNScreeningWeight assigned to all cases
WTW1DetailedWeight assigned to all entrepreneurs in initial data collection
WTW2DetailedWeight assigned to cases providing Wave 2 data
WTW3DetailedWeight assigned to cases providing Wave 3 data
WTW4DetailedWeight assigned to cases providing Wave 4 data
WTCGDetailedWeight assigned to all comparison group cases in Wave 1

have been standardized so that the average value is 1.00; the sum of the cases will equal the sum of the weights.

There are five sets of revised weights in the two data sets, as presented in Table B.4.

Weights and Analysis

Weights should be used in all types of analyses. The PSED data set allows for two basic types: assessments of samples that represent the total population and comparison of the subsample of nascent entrepreneurs with a subsample of typical adults, the comparison group.

The most straightforward assessments reflecting the entire population would be analysis using the screening data set of 64 thousand cases. For example, one may wish to develop predictive equations using age, gender, household income, and educational attainment to estimate which will be associated with, and perhaps influence,

those who would become active in the entrepreneurial process. Those who qualified as a start-up owner and were active in the past 12 months might be considered two-criteria nascent entrepreneurs (SUOWNACT = 1 in the screening data set). Once the variables had been assembled into a suitable file for analysis and those cases with missing values excluded, the weights should be “recentered” to an average value of one (1.000) by dividing the weight (WT_SCRN) by the average of the weights for the reduced sample. The analysis could be completed with confidence that the range of variables and scope of impact in the sample reflected the U.S. adult population.

But the variables in the screener data set are few, limited to the sociodemographic data collected for marketing purposes and some characteristics of the counties where the respondents were located. The detailed data sets allow more precise comparisons on a wide range of personal and situational attributes to determine those ways in which nascent entrepreneurs will differ from typical U.S. adults. Perhaps it would be a straightforward comparison of the household size or number of children at home for nascent entrepreneurs and typical adults, perhaps with a two-column table. In such an analysis, the sum of the weights for the nascent entrepreneurs (WTW1) should equal the number of nascent entrepreneurs included in the comparison; the sum of the weights for the comparison group (WTCG) should equal the sum of the number in the comparison group. A new variable, say WT_ANAL, equal to the two weight variables for the two subgroups should be computed. Once adjusted such that the total of all cases is equal to the sum of all weights, the patterns of difference between the subsamples will be accurately described and the statistical significance will be correctly computed.

Examples of analysis procedures using various weights are presented in Appendix C, including details of comparison group respondents considered active nascent entrepreneurs at the time of the initial data collection.

Respondent Rights and Welfare

As an observational study, there is virtually no chance that participation in the PSED data collection would lead to harm for the participants. In fact, it has been found that most nascent entrepreneurs are more interested in pursuing a start-up after completing the interview—it is assumed this is a good thing. On the other hand, a great deal of information is obtained during the interview, and the right of persons to control information about themselves is respected in two ways. Two elements are critical.

The first, an informed consent statement, is read to all participants at the beginning of the phone interview, which is as follows:

Before we begin, I want to assure you that all of the information you give is completely confidential. None of it will be released in any way that would permit identification of you or your household. Your participation, of course, is voluntary.

(OPTIONAL: To show our appreciation for your help, we'll send you $25 upon completion of this interview.) The fact that some respondents decide not to participate suggests they are exercising their rights not to be involved. The cover page of the mail questionnaire includes the following statements: • All information on specific firms or individuals will be kept confidential. • The identity of firms and people involved in the survey will remain anonymous. And supervisory control and operational procedures ensure that these promises will be honored. The second element, and perhaps more important, is ensuring that respondents remain anonymous in the data sets. All respondents are promised anonymity and their names, addresses, and phone numbers are not provided to anyone outside the research team collecting the data. It is possible, however, that respondents with very unique profiles might be identified through comparison with public data, such as phone directories, which sometimes list occupations, and commercial credit-rating services. This has been a realistic concern with regard to reports of personal wealth, as several respondents in low population counties reported a household net worth in excess of$10 million dollars. Since their county, occupation, and work history are included in the data set, it is possible that their specific identities could be determined. For these cases, the net worth was adjusted to equal the largest values in the continuous distribution, approximately $2.5 million. These features are consistent with the spirit and letter of current guidelines associated with the inclusion of human participants in research (http://www.nihtraining.com/ohsrsite/guidelines). Commentary The PSED data set developed for the study of the business start-up process is not without problems. On the other hand, most of the problems have been identified and solutions developed. Perhaps the most significant problem is the complexity of the data set itself. This reflects—more than anything else—the complexity of the phenomena. Many of these complications were not anticipated when the project was designed. Some of them, and the solutions that mitigated these problems, are discussed in the different chapters of this handbook and the many analyses and papers that have been published based on the PSED data set. Now that these complications are recognized and more fully appreciated, they should lead to improved procedures in subsequent research on this topic. ## Appendix C: Examples of Analysis: Work File Preparation, Comparisons, and Adjustment of Weights The data sets assembled as part of the Panel Survey of Entrepreneurial Dynamics (PSED) project are complex and extensive. Analysis of such material can be a challenge, particularly for those not accustomed to large data sets based on representative samples. The complete PSED data set consists of five different samples and over 4,000 variables. The data collection procedures have been reviewed in Appendix A and the structure of the data sets in Appendix B. This appendix is designed to provide a guide to data analysis. Most critical are discussions about different subsamples may be selected from the detailed data set for different purposes and how weights should be adjusted for different options. Proper weighting will provide both accurate descriptions of the phenomena as well as appropriate inferences when statistical tests are involved. Any analysis will involve the following steps: • Determine and enumerate the objective of the analysis. • Identify the appropriate unit of analysis and specify the aspects or characteristics relevant to the proposed analysis in terms of their conceptual features. • Locate the appropriate units of analysis in the data set. • Locate the specific variables that will provide indicators of the conceptual model. • Assemble a working data file based on the appropriate units of analysis which includes the variables required for the analysis and excludes the thousands of extraneous variables. • Using the working data file, determine those cases with complete data on the relevant variables and recenter the weights (transform the weights such that the average value is 1.00). • Perform the analysis and examine the relationship among variables and consider the strength of the relationship and the level of statistical significance. • Interpret the empirical results. • Proceed to the next issue. Experienced analysts may do many of these steps simultaneously, or almost simultaneously; experienced users of the data will develop the facility to quickly determine which research questions can be approached with the existing data. The interview schedules and questionnaires provide the best information on what data is available; the Codebook provides an overview of the type of information included in the data set as well as the frequency distributions of the unweighted data. Several practices can improve the efficiency of the analysis. Perhaps the most important is the use of batch or production processing based on explicit command or syntax files. In this mode, the analyst writes a program that will complete all the critical steps—locate and merge data files, select variables, transform variables, adjust weights, and run the analysis procedure. Although the creation of such files may take some time, they have the major advantage of providing an explicit record of all stages of the procedure and can be easily modified as the work progresses. An explicit record is critical when the analysis is to be described to others in a presentation or in a professional report or article. The ease of modification and repeating all processing procedures is invaluable as the analysis proceeds. No one is able to get all these steps right the first time, and it is never possible to predict exactly where the problems and mistakes will occur. By having an explicit program that can be adjusted during the analysis, it becomes relatively easy, although it is never simple, to locate and correct problems, add variables, adjust transformation, and the like. The alternative, attempting cursory or immediate comparisons based on spreadsheets or temporary procedures implemented by pull-down menus, can provide “instant results” but is difficult to systematize and replicate when a series of complicated adjustments is required. They are not satisfactory for any analysis with complex data sets that require sophisticated transformations and assessments. The second practice is very basic: document everything. Make sure every variable has a label, every value has a label, and the reason for any adjustment or transform or procedure is written down. This can often be done with comments inserted into the command or syntax file. With complicated assessments, the rationale for programming completed even in the last week may be lost. Documenting the reasons for all coding and transformations as they are produced is the only way to minimize lost time recreating the rationale for a given set of procedures or transforms. This is not much fun, but a necessary part of a useful, efficient data analysis that can be described to others. The number of examples highlights a range of processes and issues associated with analysis, with a focus on adjustment of the weights. They include the following: • Example 1: Estimate entrepreneurial activity in the total population • Example 2: Consider the joint impact of educational attainment and ethnic background on entrepreneurial activity in the population • Example 3: Determine the impact of household net worth on entrepreneurial activity • Example 4: Preference for risk and entrepreneurial activity • Example 5: Follow-up outcomes as affected by gender and ethnic background • Example 6: Multivariate analysis with screening and detailed data sets The first two examples utilize only the data from the screening file, the third only data from the phone interview in the initial round of data collection, the fourth uses data from the first-round phone interview and self-administered questionnaire, the fifth provides an assessment based on the initial data and the first- and second-round of follow-up interviews, and the last example reviews procedures for preparing data for multivariate analysis using both the full screener data set and the detailed data set for nascents and the comparison group. As a set, these examples cover many of the issues that should be encountered in most types of analysis. In the interest of conserving space, details on the assembly of the working analysis file are not presented; emphasis is on the selection of respondents, proper selection of and adjustments of weights, and variable transforms. These examples of analysis are provided as demonstrations on the use of the data set. They are not presented as the definitive or final analysis of these issues. A number of these topics are receiving intensive attention by teams of scholars, and their results provide more sophisticated assessments; many are found in the previous chapters of this handbook. Selection of Cases for Analysis It may appear straightforward to compare nascent entrepreneurs to the comparison groups, but three issues reflecting the conceptual frameworks and data collection procedures require attention. These are related to • The expected ownership of the nascent firm • Criteria for start-up versus a new firm • Nascent entrepreneur included in the comparison group Decisions on each of these issues can affect the cases chosen for the analysis and, in turn, the size of the subsample and the adjustments of the weights. Legal Person Ownership: To determine whether the respondent qualifies as a nascent entrepreneur involved in a start-up activity, several assessments were made of the data to produce specific criteria for inclusion in the sample. Careful attention was given to the nascent firms in relation to two issues: first, reports (based on Q190) regarding outside sponsorship of the new business and, second, the extent to which those expected to own part of the new firm were considered to be nonhuman or legal persons (financial institutions, venture capital firms, other businesses, etc.). Based on this assessment, a five-category typology was developed for all respondents in the sample (AUTONSU). It is presented in Table C.1. Table C.1 Expected Ownership of New Firm by Legal Persons [AUTONSU] This was subsequently recoded into a second typology, which involved combining all start-ups in which legal persons were expected to have 1–50% of the ownership into one category, presented in Table C.2. This distinction is quite relevant for those that conceptualize new firm creation as an individual phenomenon, one that reflects the efforts of one or more natural persons. Most efforts to explain and understand new firm creation have focused on natural persons. There is less systematic information about new business ventures implemented as collaborative efforts of natural persons and an existing business or financial institution and almost none when the new venture will be completely owned by existing legal persons. Those who wish to focus only on natural person-created new ventures may wish to restrict analysis to only the 721 entities that will be owned by one or more natural persons (AUTONSU4 = 100). Those focusing on business sponsored start-ups may wish to restrict analysis to those 109 cases where some part of the ownership will be with legal persons (AUTONSU4 = 300). Those who wish to emphasize any new business creation can focus on all 830 cases. Criteria for Start-up Versus a New Firm: One of the criteria for identifying a start-up effort that was still in the gestation period was evidence of business activity. The criterion separating start-up efforts from operating firms was the presence of positive monthly cash flow that covered all monthly expenses, including owner-manager salaries, for more than 3 months. Despite the care devoted to measuring this criterion in the interview and excluding those with new firms, subsequent analysis indicated Table C.2 Expected Ownership of New Firm by Legal Persons [AUTONSU4] some oversights in this procedure. It was determined that six of the start-up efforts had positive monthly cash flow that covered all monthly expenses and owner-manager salaries for more than 3 months (91 days) prior to the completion of the phone interview. These cases, with their identification numbers, subsample, and reported duration of positive monthly cash flow, are provided in Table C.3. Another 19 cases were found to have positive monthly cash flow for more than 1 but less than 92 days and, therefore, met the criteria. The remaining 805 cases did not report any period of positive monthly cash flow. All six of these cases are those where natural persons are expected to own 100% of the new firm. Table C.3 Cases With Excessive Monthly Positive Cash Flow Identification NumberNascent SubsamplePositive Monthly Cash Flow RESPIDRTYPECFPHLAG4 3281001241092–183 days 3281001451092–183 days 3281003951192–183 days 3281005411092–183 days 3378001371292–183 days 32810060110184–275 days Nascent Entrepreneurs in the Comparison Group Subsamples: The comparison groups were selected to be a representative sample of all adults in the United States. Some may wish to use the comparison as a representative of U.S. adults, in which case no adjustments are required. Others may wish to exclude from the comparison group those individuals who have been identified as active in entrepreneurial activity. An assessment was completed of those individuals included in the comparison groups to determine their participation in start-ups at the time of the phone interview. For those in the original mixed-gender comparison group, this was obtained in a follow-up interview that was completed 10 to 14 months after the initial phone interview. Eighty percent (179) of the 223 in the original sample were recontacted, and 4 appeared to be actively engaged in a start-up at the time of the original phone interview. For the minority oversample comparison group, questions were included in the screening and phone interview to determine if the respondent would qualify as a nascent entrepreneur at the time of the phone interview; five appeared to so qualify. The identification numbers of these nine respondents are given in Table C.4. Table C.4 Control Group Members Known to Qualify as Nascent Entrepreneurs at Initial Interview It was also possible to identify those in the minority comparison group who reported participation in a start-up in the initial screening interview but did not meet the criteria as a nascent entrepreneur eligible for a full entrepreneur data collection (active in start-up, expected to own part of the new firm, no positive cash flow for more than 3 months). These additional 23 cases can be identified by the start-up involvement variable (SUINVOL) that is greater than 1 and the qualified nascent entrepreneur variable (CGSUACT) that is not equal to 1. Those doing analysis may then create three types of comparison group subsamples: one that includes all those respondents with data, for a total of 431; one that excludes those who would have qualified as nascent entrepreneurs in the initial interview, for a total of 422; or one that excludes all cases that appeared to have entrepreneurial activity of any kind during the first interview, for a total of 399 cases. The choice will depend on the purpose of the analysis. Overview of Subsample Choices: The net result of these variations in definitions is to provide a range of options for those completing different analyses with the sample. These are outlined in Table C.5. Including all nascent entrepreneur and comparison group cases leads to the full sample of 1,261. Excluding those six cases where the nascent appeared to be reporting on a business with positive monthly cash flow for more than 3 months (91 days) reduces the nascent firm sample to 824. Excluding the seven that will be 51% or more owned by legal entities further reduces this subsample to 817 cases. If those cases in which there is any ownership by a legal entity are excluded, the nascent subsample is reduced to 716. In a similar fashion, excluding those nine individuals who appeared to qualify as nascent entrepreneurs at the initial interview would reduce the comparison group from 431 to 422 cases; excluding individuals who reported any entrepreneurial activity at the initial interview would further reduce this group to 399. In the following examples, different subsamples will be chosen and weights adjusted for the analysis. Developing different subsamples of cases can easily be achieved based on the variables included in the data files. Examples of how this can be achieved are provided below. Subsample Selection—Operational Procedures: Reducing the sample by eliminating the cases that are not consistent with the conceptual framework is achieved by using syntax statements that eliminate certain types of respondents. Using the SPSS syntax, those with positive monthly cash flow of over 91 days can be deleted with the following commands. Saving the working analysis file will permanently delete these cases. It should be noted that “NE” stands for “not equal to” and instructs the program to skip cases that meet this criterion. This would reduce the nascent subsample from 830 to 824 cases. Those nascent firms expecting more than 50% ownership by legal persons can be eliminated with the following command: This would reduce the nascent subsample from 824 to 817 cases. And all those nascent firms with any legal person ownership may be dropped from the file with the following syntax commands: Table C.5 Alternative Conceptual Choices and Impact on the Sample Sizes This would reduce the subsample further from 817 to 715 cases. If the commands are added as a set, then all specified cases are excluded. In a similar fashion, various comparison group cases can be eliminated. All those known to be nascent entrepreneurs at the time of the first interview can be deleted with This would reduce the comparison subsample from 431 to 422 cases. Eliminating all those who reported any entrepreneurial activity in the screening interview can be accomplished by the following three commands. This would reduce the comparison group subsample from 422 to 399 cases. This procedure applies only to the data file distributed at the time the handbook was being prepared. Some versions of analysis programs, such as SPSS, may require that all missing value assignments be deleted; the procedures may not allocate cases with a missing value designation leading to an undercount. Example 1: Estimate Entrepreneurial Activity in the Total U.S. Population One of the most important issues associated with the representative samples of nascent entrepreneurs is estimating the prevalence of the activity in the total population. In this case, the total population is all U.S. residents 18 years of age and older. Careful development of this estimate requires several steps: • Identify that part of the screening sample with full data on age, gender, and participation in entrepreneurial activity. • Adjust the weights for this part of the sample such that the average weight is 1.00. • Determine the average prevalence rate and standard error of the mean for each gender and age category. • Compute the number of individuals involved in entrepreneurial activity for each gender and age category. • Sum the computations across all categories to create an estimate for the entire U.S. population. In this case, the measure of participation in start-up activity as a nascent entrepreneur is based entirely on the screening interview and consists of only two criteria: reports of active efforts to implement a new business and expectation of full or partial ownership of the new firm. This two-criterion measure of entrepreneurial activity omits the third criterion used in the selection of those for the detailed PSED data set. Specifically, the lack of monthly positive cash flow that covers all expenses including wages and salaries was assessed to select nascent entrepreneurs for the detailed data collection. This additional criterion, however, was measured at the beginning of the detailed interview and no information on this criterion is available in the screening data set. The full screening sample of 64,622 is reduced for this analysis for two reasons. There are five different cohorts in the screening data set. The items used to identify those active in start-ups were used in four. As these items were not included in the screening of the initial comparison group sample (RTYPE = 20), these 2,010 cases are dropped from the analysis. The second factor is the absence of age data (USAGE7C) on about 3% of all cases; another 1,988 cases are dropped from the working file. The average weights (WT_SCRN) for the 14 gender (USGENDER) and age cells can then be computed as shown in the top of Exhibit C.1. The average value for each cell is adjusted, as shown in the computation of new weights (WTAGESEX) in the center of Exhibit C.1. The results are shown at the bottom of Exhibit C.1; the average value for each cell is 1.000. It is important to stress that this weight adjustment has been done to maximize the precision of the means and standard error of the 14 gender, age groups. If the purpose was to compare the prevalence rates using a means test or analysis of variance, it would be more appropriate to adjust the overall weight to equal one with one computation. The next step in the procedure is to compute the mean and standard error of the mean for each of the 14 cells. The results are presented in columns 3 and 4 of Table C.6 to estimate the total number of U.S. adults involved in start-up activities. The first stage is assembling a count of the total eligible individuals, presented in Column 2 of Table C.6; data were from the U.S. Census Bureau Projections of Resident Population by Age, Sex, Race, and Hispanic Origin for 1999 (NP-D1-A). These two sets of information, U.S. population in each cell and prevalence rates and standard errors of the mean from the PSED screening sample, are used to compute the lower and upper bounds of the 95% confidence interval as well as the mean estimate, shown in columns 5 to 7 of Table C.6. The results are in terms of hundreds of thousands of individuals. As shown in the bottom row of Table C.6, the mean is about 11.9 (5.80 %) million with a 95% confidence interval from 10.6 (5.19 %) to 13.1 (6.41%) million. Example 2: Educational Attainment, Ethnic Background, and Entrepreneurial Activity among Men This example uses the same two-criteria measure of entrepreneurial participation based on the screener questions but focuses on the interaction of two factors—educational attainment and ethnic identity—among men. To reduce the cohort effects associated with educational attainment among minorities (older minorities Exhibit C.1 Example 1: Adjustment of Population Screening Weights have less education), the age is standardized at 25 to 54 years of age. Younger men are excluded, since those under 25 may not have finished all their education. The other ethnic category—composed of American Indians, a diversity of Asians, and many with complex ethnic backgrounds—is very diverse and hard to interpret and is also excluded. Weights are computed in two ways for two different assessments. First they are computed as above, with each of 15 cells adjusted so the average weight in each cell is 1.00. This is appropriate for an assessment of the mean values and the confidence intervals associated with each ethnic, educational group. A second assessment focuses on the capacity to determine if there are statistically significant differences between the groups. A single adjustment is used to create an overall weight with an average value of 1.00. Both assessments start with identifying all the cases where full information is available on the critical variables: age, gender, educational attainment (USEDUC5), ethnic identity (USRACE4), and nascent entrepreneurship. The resulting sample has a total of 17,755 cases. Exhibits C.2a and C.2b present the three stages of the first weight adjustment. The initial presentation shows the cell average of the initial screening weight (WT_SCRN) for all 15 cells. The second section indicates the adjustments for each individual cell. The final section indicates the results in terms of the average weights for each of the 15 cells. Following this, the mean prevalence of participation in entrepreneurial activity and the standard error of the mean can be computed and used to compute the 95% Table C.6 Estimates of Nascent Entrepreneurs (NEs) in the United States: 1999 Exhibit C.2a Example 2: Weight Adjustment for 15 Ethnic, Educational Attainment Cells Exhibit C.2b Example 2: Weight Adjustment for 15 Ethnic, Educational Attainment Cells confidence interval. This is displayed visually in Figure C.1. The vertical bars indicate the confidence interval. If the vertical bars for any two groups overlap, there is no statistically significant difference between the groups. This visual display facilitates comparisons among any two groups and clearly shows the lack of difference in entrepreneurial activity among White men related to educational attainment. It also makes clear the higher levels of participation among Black and Hispanic men compared to White men and the dramatic increase among Black and Hispanic men with higher levels of educational attainment. The confidence intervals also illustrate how the smaller sample sizes lead to less precise estimates for the Black and Hispanic men. The alternative strategy is to focus on the impact of educational attainment on participation in start-ups for each ethnic group. The question of the presence of any Figure C.1 Example 2: Visual Display of Means and Confidence Intervals: Entrepreneurial Activity by Ethnic Background and Educational Attainment Exhibit C.3 Example 2: Adjustment of Weights for Full Subsample impact can be determined with a cross-tabulation analysis. In this, it is preferred that variation across cells is maintained and the overall weight needs only to have an average of 1.00 for the total sample. The adjustment requires only a single correction, as shown in Exhibit C.3. In this case, a more precise adjustment is made by multiplying the original weight by the ratio of the total number of cases to the sum of the weights (17755/16691.22 = 1.064). The resulting analysis, presented in Exhibits C.4a and C.4b, shows the three cross-tabulations, one for each ethnic group. Because of the large sample size (14,442) the educational impact among Whites is statistically significant, although the substantive impact is small. Among Blacks, the substantive impact is substantial and leads to a statistically significant difference although the sample size is rather small (2,119). Among Hispanics, the effect is less than Blacks and because the sample size is even smaller (1,194), the statistical significance is borderline, not quite at the 0.05 level. These results are consistent with the visual display provided in Figure C.1; the row values in the right columns in Exhibit C.4 are identical to the mean values for each group in Figure C.1. Example 3: Household Net Worth and Entrepreneurial Activity This example focuses on a single hypothesis: Does household net worth affect the tendency to participate in business start-ups? This is sometimes called the liquidity hypothesis among economists, and hypothesizes that the more funds one has personally available (liquidity), the greater the tendency to be involved in new firm creation. Data from the initial detailed phone interview provides information on household net worth for both the nascent entrepreneurs and the comparison groups. Estimates from the respondents have been aggregated into a six-level measure of household net worth (HHNETR6) for 96% of the cases. It ranges from negative net worth (5.6%) to those reporting$1 million or more in net worth (2.8%).

Several other issues need to be resolved in order to analyze this question: which of those identified as nascent entrepreneurs are to be included in the analysis and which members of the comparison group are to be excluded.

The expectation that household net worth may affect participation in business start-ups reflects assumptions about the sources of financial support. It assumes limited initial support from other businesses or banking institutions. The most appropriate test of this hypothesis would restrict analysis to nascent firms that are being started by natural persons, excluding any ownership by legal entities such as other businesses or financial institutions. In addition, the character of the subsample would be less confusing if all nascent start-up's positive monthly cash flow for over 3 months were excluded. In a similar fashion, the comparison group can be adjusted to exclude any person reporting any effort to participate in entrepreneurship. As discussed above, this is achieved with the following commands:

Exhibit C.4a Example 2: Alternative Assessments of Educational Attainment, Ethnic Background on Entrepreneurial Activity Among Men

Exhibit C.4b Example 2: Alternative Assessments of Educational Attainment, Ethnic Background on Entrepreneurial Activity Among Men

The resulting sample has 1,115 cases, 716 nascent entrepreneurs and 399 comparison group members. Data on household net worth (HHNETR6) is missing for 41 cases within this subsample, reducing the total number of cases to 1,074.

The next issue to confront is the assembly of a common weight for all cases. The weights were computed separately for the first round of nascent entrepreneurs (WTW1) and the comparison group (WTCG). Creation of a new weight variable easily solves this problem.

It turns out that almost no adjustment is required for the new weight variable for this analysis. As shown below, the sum of the weights is almost exactly equal to the number of cases without adjustment. To be consistent, however, it is adjusted so the sum of the weights equals the number of cases and the average weight is equal to 1.00.

The result of the comparison is presented in Exhibit C.5, showing the distribution of wealth among those active in start-ups with no expected business or institutional ownership with a representative sample of U.S. adults not involved in business start-ups. There is no statistically significant difference between these two groups in terms of household wealth. The bar graph presented as Figure C.2 makes the similarity of the two distributions clear.

Household net worth is highly correlated with the age of the major wage earners in a household, and since nascent entrepreneurs tend to be younger adults, a comparison that controls for age may be more appropriate. As most people involved in entrepreneurship are less than 45 years old, the comparison was done

Exhibit C.5 Example 3: Weighted Assessment of Household Wealth and Entrepreneurial Participation

separately for those 18 to 44 years and those 45 years and older. The results, presented in Figures C.3 and C.4, make clear that there is a statistically significant difference among the younger group: the comparison group tends to have higher household net worth than does the nascent entrepreneur group. Among those 45 and older there is no statistically significant difference.

Figure C.2 Example 3: Weighted Assessment of Household Wealth and Entrepreneurial Participation
Figure C.3 Example 3: Weighted Assessment of Household Wealth and Entrepreneurial Participation (18 to 44 years old only)
Figure C.4 Example 3: Weighted Assessment of Household Wealth and Entrepreneurial Participation (45 years and older only)

Not only is there no explicit support for the “liquidity” hypothesis; this assessment suggests a reverse effect. Less household net worth may encourage those under 45 years old to become involved in starting a new firm. This has been an example of a basic analysis with no adjustments to control or account for any other factor except age. A more sophisticated assessment of the relative impact of household net worth and its effect on participating in a start-up in relation to 16 other factors can be found in Kim, Aldrich, and Kiester (2003).

Example 4: Preference for Risk and Entrepreneurial Activity

Perhaps no personal trait associated with starting a business gets more attention than a “preference for risk” It is widely assumed that entrepreneurs are more risk-oriented than ordinary individuals with “regular jobs.” For this reason, a number of items were included in the self-completed mail questionnaire related to preferences for risk. One item, QH9 (presented in Exhibit B.2 of Appendix B) asks which of two businesses the respondent would prefer to own: Alpha, with high earnings and high risk, or Beta, with lower earnings but less risk. Following the famous assessment of the banker Rothschild, these are referred to as the “eatwell” versus “sleepwell” choices.

The analysis requires that the initial data be assembled from two sources, the initial detailed phone interviews and the self-completed mail questionnaire. Although a very high proportion (74%) of those completing the initial phone interview returned the self-completed mail questionnaire—one in four (26%) did not. Using the procedures discussed above to consolidate the weights for the initial wave as well as excluding those cases with a lack of data on the relevant variables, the total eligible case count is 793. The appropriate weights for all cases are those developed for the initial round of data collection, WTW1.

As with the previous analysis, it seems most useful to restrict this analysis to firms that will be owned only by natural persons, and it would sharpen the contrast if the analysis is restricted to those in the comparison group that are not involved in entrepreneurship in any way. Hence, the total sample was reduced using the syntax commands discussed earlier. Further, only those cases for which valid data is available for QH9 are included (values of 1 or 2); 6 cases without QH9 data are also excluded.

As with previous cases, the weights must be computed for this comparison and then standardized such that the average weight is 1.00.

Once the normalized weights are computed, the analysis can proceed.

The outcome is presented in Exhibit C.6, and it would appear that there is no statistically significant difference in a preference for a high-payoff, high-risk venture—the “eatwell” cases—when compared to a lower-payoff, lower-risk venture—the “sleepwell” cases. The overall sample of nascent entrepreneurs is very similar to the comparison group in this respect. The difference is marginally statistically significant at the 0.10 level.

Exhibit C.6 Example 4: Preference for Risk—Nascent Entrepreneurs and Comparison Group

The phone interview, however, contains a question about preferences for growth. Item Q322 asks if the nascent would prefer the business to “be as large as possible” or “a size I can manage myself or with a few key employees.” For most respondents, this was answered several months before they completed the self-administered questionnaire and thus provides an assessment of their growth aspirations before their risk orientation was determined. It turns out that 464 qualifying nascent entrepreneurs answered both questions. The weight for this 464 is recomputed to equal an average of 1.00 and the impact assessed. As shown in Exhibit C.7, the result is statistically significant (at the 0.00003) level and there is substantial substantive significance. Those entrepreneurs interested in high-growth new firms are more than twice as likely (35% vs. 15%) to prefer a higher-payout, riskier venture.

The impact of this classification of nascent entrepreneurs and their comparison to the control group can be presented in graphic form, show in Figure C.5. This clearly indicates that “comfortable size” nascents are identical (not just close) to the comparison group in preference for high- versus low-risk ventures. The “growth oriented” nascent entrepreneurs are quite different from both low-growth nascents and the comparison group. For some, only those with aspirations for high growth would be considered “real” entrepreneurs. It is not possible, however, to determine which comes first—a preference for growth or a tolerance for risk; they both may be adopted at about the same time.

Exhibit C.7 Example 4: Preference for Risk and Firm Growth—Nascent Entrepreneurs

Figure C.5 Preference for Risk: Nascent Entrepreneurs by Growth Orientation and Comparison Groups

Example 5: Start-up Transitions as Affected by Gender and Ethnic Background

A primary objective of the PSED project was to determine the outcome of those who were actively involved in a business start-up. Once identified in the initial assessment as active nascent entrepreneurs, the major purpose of the follow-up data collection was to assess the current status of their efforts. All start-up efforts could be classified in one of four categories based on reports from the nascent entrepreneurs in the follow-up interviews:

• New business in place

• Active effort to complete the start-up process still underway

• No active effort to complete the start-up process, but not abandoned

• Start-up is terminated, no one is still actively trying to start the business

The first round of follow-up interviews took place 12 to 18 months after the initial interview. As an example of how analysis might proceed, the outcome information for those active in a start-up will be considered. First year outcome data is available for 74% (615 of 830) of those identified as nascent entrepreneurs and interviewed in the first follow-up after their initial interviews. This assessment will use as many of these firms as possible. Cases considered operating firms in the initial interview, with positive monthly cash flow for more than 3 months and with more than 50% of future ownership with legal persons were excluded. As before, the following commands were used to reduce the sample:

In addition, none of the comparison group cases is relevant, and they are all excluded with a single command:

Several features of the data set need to be taken into account in completing this analysis. First are the differences in the location of the follow-up data. For the mixed-gender and female-only cohorts, the initial status is part of the second-wave data (variables starting with “R”). The first follow-up data for the minority oversample are in the third wave of data collection (variables starting with “S”). Standardizing the relationship between the initial and first follow-up data can be accomplished by creating a new set of variables associated with the outcome. This is done for the different cohorts as follows: The variable “RTYPE” distinguishes between the cohorts of nascent entrepreneurs: “10” for mixed-gender, “11” for female oversample, and “12” for the minority oversample. New variable names and value labels are also applied.

Because of the importance of the status of the start-up effort in the follow-up, the question was actually asked twice. Those who were unsure of the correct response the first time they were asked the question (R502 or S502) were asked the question a second time (R503 or S503). By asking the question twice, interviewers were almost always able to obtain a response appropriate to one of the first four categories. This means, however, that the responses to the second question must be recoded into the first to provide a single variable representing the first follow-up outcome. Those that gave a “something else” response (value 5) to the first question are adjusted to match their response to the question when it was asked the second time with the following commands:

An additional issue occurs with relation to the weights that were recalculated for each data collection wave. This can be resolved by developing a new set of weights, based on the wave in which the follow-up data is located. The programming is relatively straightforward, as follows:

By assigning a weight of 999 to all cases in which no relevant follow-up data is provided, and classifying this as a missing value, these cases will be automatically excluded from any analysis using FU1_WT as the weighting variable.

The remainder of the assessment is rather straightforward. The most precise measure of gender, NCGENDER (see Chapter 2), and ethnic background, PGRACE (See Chapter 3), are used for this analysis. In each analysis the sum of the weights of the relevant cases is determined and adjusted so that the average weight equals one. The following represents the correction for the gender comparison weights.

The analysis related to ethnic background, in which only the White, Black, and Hispanic cases are included, required a second adjustment of the weights.

The analysis that was completed resulted in cross-tabulations but is easiest to present as bar graphs. The comparisons based on gender and ethnic background are presented in Figure C.6. There was no statistically significant difference related to gender, and the substantive differences are quite small: about 30% report an operating business, 30% report an active start-up effort, 20% report an inactive start-up effort, and 20% report that the start-up has been abandoned.

Differences related to ethnic background as shown in Figure C.6 are much greater and were just marginally statistically significant at the 0.05 level.

This leads to an examination of the interaction between gender and ethnic background, as shown in Figure C.7. In this case, both assessments are statistically significant at the 0.05 level, and the patterns are somewhat different for men and women.

Substantially fewer Black and Hispanic men report a going business in the first follow-up interview (18 to 22%), compared to White men (34%), and substantially more report continuing efforts in the start-up phase, active or inactive. The percentage of men who report abandoning the effort is about the same for all three groups (15% to 20%).

Among women, however, the differences are far more dramatic. Black and White women report the same proportions of going businesses, and more Black women report efforts in the start-up phase. Only 10% of Black women report that a startup has been abandoned, the smallest proportion of any group of men or women. Hispanic women, however, are most distinctive. Exactly half report the effort has been abandoned, and only one in six (15%) reports a going business. Although there are only 14 Hispanic women, these results clearly reflect a quite different situation compared to any of the other five gender-ethnic groups. Ideally, this assessment of Hispanic women would be replicated with a larger representative sample.

Perhaps most dramatic is the contrast between the evidence of participating in the start-up process and the data on completing the process with a new firm. It was

Figure C.6 Start-up Process First Follow-up Outcomes: By Gender and Ethnic

Background

Figure C.7 Start-up Process First Follow-up Outcomes: Gender by Ethnic Background

clear in Figure C.1 (presented in Example 2) that Black and Hispanic men are substantially more active than White men in entering the start-up process. But the evidence in Figure C.5 suggests Black and Hispanic men report less early success—although they may be simply taking longer to implement new firms than White men. More details on the extent of this “minority men effect” will be required to assess the reasons for this difference. This difference is not present when Black and White women are compared. Black women are just as successful as White women in completing the start-up with a new firm, and a very small proportion quit the start-up effort. Larger samples of Hispanic women will be required before their outcome patterns are clearly established.

A final comment is related to the nature of the procedure used to locate nascent entrepreneurs in the screening procedure. It takes months—sometimes years—to complete the start-up process. The screening procedure identifies nascent entrepreneurs at a random point in the start-up process. Measures of the time to complete the process and successful transitions from start-up to new business should be based on the point of firm conception, that date when the start-up process began. The determinations are possible with the PSED data set, but they are complex and go beyond this introductory discussion. The implementation of more precise measures of the start-up timing may affect inferences regarding the outcomes.

Example 6: Multivariate Analysis: Screening and Detail Samples

The PSED data set makes it clear that the entrepreneurial process is complex and multifaceted. The data cover a wide range of processes and factors that may affect the creation of a new business. This would lead many to consider multivariate analysis, by which the relative impact of a number of independent variables on the process might be explored. There are two options for such an analysis. The first is the most straightforward, the use of the screening data set to explore a range of factors and their relative impact on those identified as active in the start-up process. The second is to use the detailed data set for such analysis. Each strategy has advantages and disadvantages.

Screening Data Set and Multivariate Analysis. The screening data set provides two advantages. First is the large size, 64,000 cases, which allows for precise estimates of the impact of various variables. The second is the fact that it is a representative sample, which allows inferences about the entire U.S. population. This means that the average values of the independent variables as well as their dispersion (range or variation) may be considered to represent the U.S. population. So if one wanted to consider the relative impact of gender, age, and educational attainment, the procedure is relatively straightforward. These variables are identified and coded into appropriate categories, cases with missing values may be deleted, and their univariate impact can be considered.

The PSED screener, as discussed above, characterizes most respondents regarding active involvement with two criteria. First, did the respondents report that they were active in starting a new firm, and, second, did they expect full or partial ownership of the new firm? Those that satisfied these two criteria were coded as 1 for variable SUOWNACT, all others were coded zero. The only adjustment regarding the weights would require recentering weights for those cases with complete data on all variables in the analysis. Of the 64,622 cases in the PSED screener, those in the mixed-gender comparison group (RTYPE = 20), 2,010 cases, are dropped from the data set. Another 1,988 cases are dropped due to missing data on age and a further 270 due to missing data on educational attainment. (Because the screening sample was collected to have an equal number of cases on men and women, there is no missing data on gender.

Following identification of all eligible cases, the weights are “recentered” such that the average value is 1.00, using the following command syntax:

This, once again, produces an average weight of 1.00 and the sum of the cases equals the sum of the weights.

The univariate impact of these three variables is presented in Table C.7. Age and educational attainment have been recoded to match the values for the following discussion.

Given the discussions in previous examples and a sample size of over 60,000, it should be no surprise that the results are highly statistically significant.

As the dependent variable, participation in a business start-up is dichotomous, and all three independent variables are categorical. Logistic regression is appropriate for an initial analysis. Analysis was completed using the SPSS Windows 11.5 logistic regression procedure with forced entry of all variables. Forced entry reflects the assumption that all variables will be important and that the procedure will be used to determine relative significance, reflected in the Beta weights. The results are presented in Table C.8. All three sociodemographic variables are treated as sets of dummy variables (values of 1 or 0) compared to a base value that does not appear in the assessment: women, those 65 and older, and those with graduate educational experience.

Table C.7 Screening Sample: Univariate Impact on Two-Criteria Nascents

This makes clear that all three factors have a unique and statistically significant contribution to explaining the variation in participating in a business start-up. The overall model has a chi-square value of 1420.298 (p < 0.000) and has a Nagelkerke R-squared estimate of 0.064. These three variables may account for 6.4% of the variation in participation in business start-ups as a nascent entrepreneur.

Detailed Data Set and Multivariate Analysis. The major advantage of the detailed data set is the large number of variables that have been assembled for representative samples of nascent entrepreneurs and typical adults not involved in business start-ups. This reflects the decision that many factors may be of importance and worth serious consideration. Cost considerations, however, limited the number of cases that were collected for the comparison group, which is about one half the size of the nascent entrepreneur group. This is not a major concern when there are

Table C.8 Screening Sample: Logistic Regression Forced Entry Model, Three IV's

direct comparisons between the two groups. A comparison of the household net worth—discussed in Example 3—illustrates such an analysis. Those identified as active in the entrepreneurial process either had about the same household net worth as those not so active or if they were under 45 years of age, typically with less household net worth.

But a different challenge arises for assessing the relative or joint impact of various independent variables. The primary complication is that the total number of cases does not provide a representative sample of the U.S. adult population. This means that the average and range of the variables for the entire sample would clearly not represent the U.S. adult population.

One solution is to adjust the weights to create a sample that has the same proportion of nascent entrepreneurs as expected among the U.S. population. Based on the analysis of the screening data as presented in Table C.7, this is 6% of the adult population. The following strategy can be adopted. Once the variables for analysis are chosen and decisions made about who is an appropriate candidate for the nascent entrepreneur and comparison groups, the nascent entrepreneurs' weights are adjusted so they are 6% of the total weighted detailed sample.

For this example, the variables chosen from the detailed sample are gender, age, educational attainment, and total years of work experience. Although the first three were gathered in the screening interview, they were obtained again in the detailed interview. The most accurate version of each (NCGENDER, EX6AGE, ITR-WEDU4) were utilized, along with the phone interview item related to total years of work experience (Q340). For this assessment, only nascent entrepreneurs involved in firms in which only natural persons would be owners were retained, and all comparison group members with any indication of entrepreneurial activity were excluded (See Example 3 for the command syntax). Once all the cases with missing values on the independent variables were excluded, a total of 1,095 cases remained, 397 comparison group members and 698 active in the start-up process where only natural persons would own the new firm.

The procedure for creating, assessing, and adjusting the weights is outlined in Exhibit C.8. The first step is to create a single weight for all cases, WT_EX6. The weight is then assessed in terms of the average value for each subgroup. Use of the means procedure provides a convenient way to assess the average and sum of the weights. The weight variable WT_EX6 is then adjusted so the average value is equal to 1.00 for both the comparison and nascent subsamples. A new weight is then created so the total sample will represent the U.S. adult population, WT_EX6P. This is done by reducing the average value of the weights assigned to the nascent subsample to achieve this end. The total number of nascent cases should equal 6% of the total weighted sample, or 6/94ths of the number of comparison group cases.

The next step is to recenter the weights so that the sum of the weights equals the case count. Given that the sum of the weights is 422 and the total case count is 1,095, each weight is multiplied by 1,095/422. The procedure does not affect the proportion of nascents in the sample but corrects the sample size for statistical calculations. The last section of Exhibit C.8 shows that the proportion of nascents in the sample weighted by WT_EX6P is exactly 6%. The weighted number of nascent cases is 66, and the weighted number of comparison group cases is 1,029. It is now possible to consider the relative impact of different independent variables on variation in nascent entrepreneurs among the sample.

The impact of this adjustment on the basic patterns is presented in Table C.9, which has the same form as three of the variables presented in Table C.7, based on the full screening sample. In this case, the prevalence rate for all cases is not based on the screening data but has been adjusted to match the prevalence based on the screening data, 6 per 100 adults.

But the impact of gender, age, and educational attainment are of considerable interest. For comparison, the prevalence rates from the full screening sample in Table C.7 are provided in Table C.9 next to the population-weighted detailed sample in parentheses. The prevalence rates are very close for gender and age, with a slight difference related to educational attainment. Prevalence related to education is the highest for college degrees with the population-weighted detailed sample, and for those with graduate experience from the screening sample.

The use of corrected population weights is the only way to determine the relative impact of the wide range of variables in the detailed sample data. The focus is on identifying the relative impact of different variables on the decision to engage in starting a new business. This is done by creating alternative multivariate models and considering the overall impact of the models. For example, the same procedures discussed as the simple model with the screening data were applied to the population-weighted detailed data, again using SPSS Windows 11.5 and forced entry of all variables. The results are presented in Table C.10. Again, all three sociodemographic variables are treated as sets of dummy variables (values of 1 or 0) compared to a base value that does not appear in the assessment: women, 65 and older, and graduate educational experience. Note that although the beta weights are very similar to those reported for the full screening sample in Table C.8, none have a statistically significant contribution to the explained variance, as the standard

Exhibit C.8 Transforming Variables and Weights to Create a Population Representative Sample

Table C.9 Detailed Sample, Population Matching Weights: Univariate Impact on Three-Criteria Nascents

errors of the beta weights were substantially higher. Both reflect the shift from a sample of 60,354 to 1,095.

The overall model has a chi-square value of 13.261 (7 degrees of freedom, p < 0.058) and has a Nagelkerke R-squared estimate of 0.034. These three variables may account for 3.4 % of the variation in participation in entrepreneurial activities.

Table C.10 Detailed Sample: Logistic Regression Forced Entry Model—Three IV's

A second model that included the fourth sociodemographic variable, years of work experience, was also assessed with exactly the same procedure. In this case, the years of work experience was represented by 5 dummy variables (values of 1 or 0) compared to those with 31 to 60 years of work experience. The results are presented in Table C.11. Again, the beta weights are very similar to those for the full screen model or the three IV variable model with regards to the constant, gender, age, and educational attainment. Once again, the standard errors of the betas are quite large, and none are making a statistically significant contribution.

The enhanced overall model, however, has a chi-square value of 15.246 (12 degrees of freedom; p < 0.228) and has a Nagelkerke R-squared estimate of 0.038. These four variables may account for 3.8% of the variation in participation in entrepreneurial activities, an increase of 0.4% from the three variable model.

The four variable model fits about as well, or as poorly, as the three variable model, with a slight improvement in the explained variance. It would seem appropriate to assume that a modest number of years of work experience, from 1 to 20, may have an independent positive contribution to a decision to participate in a new business start-up. No experience or over 20 years of experience may reduce the tendency to participate.

Table C.11 Detailed Sample: Logistic Regression Forced Entry Model—Four IV's

An alternative objective is to determine whether or not a single independent variable has any statistically significant impact on the dependent variable. In that case, it is possible to simplify most independent variables to dichotomous variables and use procedures that answer only this question. Two criteria are used to assess success. Is there evidence that a given variable has any statistically significant relationship to a dependent variable? Is there an indication that the multivariable model has a statistically significant match with the data? With these objectives in mind, and using an odds-ratio measure of impact in a logit regression model, it was possible to consider a 10 variable model with 7 control variables; there was evidence that 7 had a statistically significant impact (Kim et al., 2003).

Commentary

This has been an introductory discussion of selected issues associated with analysis of the PSED data set. It has assumed some familiarity with standard data processing procedures, such as SAS or SPSS. While the initial training investment required to develop the skills to explore the start-up process in detail is not trivial, such detailed assessments can provide a greater appreciation of the diverse ways in which new firms come into being.

These examples have dealt with less than a dozen of the variables in the PSED data sets. The various chapters in the PSED handbook have dealt with hundreds of variables. There is plenty left over for everybody else. It is unlikely that the full potential of this resource will ever be exhausted.

References
, , & (2003, August). Does wealth matter? The impact of financial and human capital on becoming a nascent entrepreneur. Paper presented at the American Sociological Association Annual Meetings, Atlanta, GA.

## About the Editors

William B. Gartner is the Arthur M. Spiro Professor of Entrepreneurship at Clemson University. Prior to joining University of Southern California, he was on the faculty at Georgetown University, the University of Virginia, San Francisco State University, and the University of Southern California. He is one of the cofounders of the Entrepreneurship Research Consortium, which initiated, developed, and managed the Panel Study of Entrepreneurial Dynamics. His service to the entrepreneurship field has included two consecutive terms as Chair of the Academy of Management Entrepreneurship Division (1985, 1986), special issue editorships for the Journal of Business Venturing (JBV) and Entrepreneurship Theory and Practice (ETP), and Editorial Board memberships with the Academy of Management Review (AMR), Journal of Management (JOM), JBV, ETP, and the Journal of Small Business Management (JSBM). His research has been published in AMR, JBV, ETP, JOM, and JSBM; won awards from the Academy of Management, ETP, and the Babson Kauffman Entrepreneurship Research Conference; and has been funded by the Kauffman Center for Entrepreneurial Leadership, Coleman Foundation, U.S. Department of Education, Small Business Foundation of America, the Los Angeles Times, the Pacific Gas and Electric Company, the Corporate Design Foundation, and the National Endowment for the Arts. His research on nascent entrepreneurs explores how they find and identify opportunities, recognize and solve start-up problems, and undertake actions to successfully launch new ventures. He is also collecting and analyzing the stories entrepreneurs tell about their entrepreneurial adventures.

Paul D. Reynolds is Professor of Entrepreneurial Studies at Babson College (Wellesley, Massachusetts) and a Research Professor of Entrepreneurship at the London Business School. He served as the director of the annual Babson Kauffman Entrepreneurship Research Conference (1996–2000) and was the Coleman Foundation Chair in Entrepreneurial Studies at Marquette University (Milwaukee, Wisconsin; 1990–1995).

He was part of the team that created the Entrepreneurial Research Consortium (ERC) and has served as the Coordinating Principal Investigator through the life of the project, 1995 to 2003. He has also served as the Coordinating Principal Investigator of the Global Entrepreneurship Monitor (GEM) program since its inception in 1998. He has been the Principal Investigator on more than two dozen funded research projects.

In 2004 he was the recipient of the International Award for Entrepreneurship and Small Business Research sponsored by the Swedish Business Development Agency (NUTEK) and the Swedish Foundation for Small Business Research (SFS) in recognition of his role in developing and implementing the Panel Study of Entrepreneurial Dynamics and the Global Entrepreneurship Monitor research programs.

His educational background includes degrees in engineering (BS, University of Kansas), business (MBA, Stanford University), psychology (MA, Stanford University), and sociology (PhD, Stanford University). He has been on the faculty or staff of the University of California, Riverside; University of Minnesota; Wharton School, University of Pennsylvania; Nanyang Technical University, Singapore; INSEAD (Fontainebleau, France); and the University of Michigan Institute for Social Research. He is the author or coauthor of five conference proceedings; four books; four data sets in the University of Michigan ICPSR public archives; over 30 project reports and research monographs; over 60 peer review journal articles, chapters, or refereed conference proceeding reports; and several hundred professional conference presentations.

Kelly G. Shaver is Professor of Psychology at the College of William & Mary. From 1977 to 1979, he was Program Director for Social and Developmental Psychology in the Division of Behavioral and Neural Sciences at the National Science Foundation. He currently serves as an advisor to http://FamilyCareAmerica.com, is a founding director of http://MBATechConnect.org, and serves as a member of the international advisory board of the Entrepreneurship and Small Business Research Institute (ESBRI) in Stockholm, Sweden, where he was a Visiting Professor from 1999 to 2000.

For 5 years he was Editor of Entrepreneurship Theory and Practice and has served on the editorial boards of the Journal of Personality and Social Psychology and the

Journal of Personality. He currently serves on the editorial boards of the Journal of Applied Social Psychology, Entrepreneurship and Regional Development, and the Journal of Developmental Entrepreneurship. He is the author of seven books, coauthor or coeditor of five others, and is author or coauthor of over 140 papers and research articles on attribution processes and entrepreneurship. His paper on the motivations of nascent entrepreneurs was the winner of the Babson Kauffman Entrepreneurship Research Conference Best Paper Award for 2000, and his course on the psychology of entrepreneurship won the 2000 McGraw-Hill/Irwin Award for Innovation in Entrepreneurship Pedagogy. He is a Fellow of the American Psychological Society, a member of the Society of Experimental Social Psychology, and the current (2003–2004) Chair of the Entrepreneurship Division of the Academy of Management. His e-mail is kgshav@netscape.net; his web pages are at http://www.wm.edu/PSYC/shaver.html.

## About the Contributors

Howard E. Aldrich is Kenan Professor of Sociology at the University of North Carolina, Chapel Hill, where he won the Carlyle Sitterson Award for Outstanding Teaching in 2002. He is chair of the Department of Sociology and Adjunct Professor of Management in the Kenan Flagler Business School. In 2000, he received two honors: the Swedish Foundation of Small Business Research named him the Entrepreneurship Researcher of the Year, and the Organization and Management Division of the Academy of Management presented him with an award for a Distinguished Career of Scholarly Achievement. His latest book, Organizations Evolving (Sage, 1999), won the Academy of Management George Terry Award as the best management book published in 1998 to 1999 and was cowinner of the Max Weber Award from the American Sociological Association's Section on Organizations, Occupations, and Work. He is currently engaged in three research projects: (1) the process by which entrepreneurial teams are founded, focusing on similarity and differences between team members; (2) the contribution that voluntary association membership makes to entrepreneurial success; and (3) how to design courses and classroom activities to promote active learning by students.

Kathleen R. Allen, PhD, is a Professor in the Greif Entrepreneurship Center of the Marshall School of Business at the University of Southern California and the author of Entrepreneurship and Small Business Management (2nd Edition), Launching New Ventures (3rd Edition), Bringing New Technology to Market, and Growing and Managing an Entrepreneurial Business, as well as several trade books. She is the Director of the USC Technology Commercialization Alliance, which focuses on commercializing USC's technologies, and principal investigator on a National Science Foundation grant to build a national technology commercialization network (N2TEC) that will link universities, allowing them to share knowledge and resources and to collaborate across geographic boundaries to commercialize their technologies. As an entrepreneur, she was active in commercial real estate development for 10 years, owning two businesses, is cofounder of two technology ventures, and is a director of a NYSE company.

Marne L. Arthaud-Day is a doctoral candidate in management at the Kelley School of Business at Indiana University. Her research concerns the influence of values and related social attitudes (e.g., job and life satisfaction) on strategic decision making, organizational culture, and strategic implementation. She teaches in the area of strategic management. She received her BA from Wake Forest University, M.Div. from Princeton Theological Seminary, and her MBA from the University of Texas at San Antonio.

Robert A. Baron, PhD, University of Iowa, 1968, is the Dean R. Wellington Professor of Management and Professor of Psychology at Rensselaer Polytechnic Institute. He has held faculty appointments at Purdue University, University of Minnesota, University of Texas, University of South Carolina, University of Washington, Princeton University, and Oxford University (Visiting Fellow, 1982). He served as a Program Director at the National Science Foundation (1979–1981) and was appointed as a Visiting Senior Research Fellow by the French Ministry of Research (2001–2002) at the Université des Sciences Sociales, Toulouse. He is a Fellow of both the American Psychological Association and the American Psychological Society.

He has published more than 100 articles and 35 chapters in edited volumes and he is the author or coauthor of more than 40 books in the fields of management and psychology, including Behavior in Organizations (8th Edition) and Social Psychology (10th Edition). His latest book is Entrepreneurship: A Process Perspective. He holds three U.S. patents and was founder, President, and CEO of Innovative Environmental Products, Inc. (1993–2000). His current research focuses primarily on social and cognitive factors in entrepreneurship.

Candida G. Brush is Associate Professor of Strategy and Policy, Director of the Council for Women's Entrepreneurship and Leadership (CWEL), and Research Director for the Entrepreneurial Management Institute at Boston University. She teaches Entrepreneurship and Strategy courses in the undergraduate, MBA, doctoral, and executive MBA programs. Her research investigates the role of resources in emerging organizations and growth strategies of women-led ventures. With four other researchers, she investigates women's access to growth capital, referred to as the Diana Project. This research is sponsored by the Kauffman Foundation and ESBRI (Swedish Research Foundation). Her most recent book is Clearing the Hurdles: Women Building High-Growth Businesses (2004).

Ana Cabezuelo is Associate Professor in the Business Organization Department of the Autonomous University of Madrid (Spain). She worked on the PSED as a Visiting Professor in the Department of Management, Cook School of Business, Saint Louis University. Currently she is the Project Manager of the Center of Entrepreneurship and Businesses Initiatives (C.I.A.D.E.) in Madrid. The mission of this institution is to promote the establishment of private enterprises in Madrid with special emphasis on the student community and the faculty of the niversidad Autonoma de Madrid. Her latest research, funded by Spain's State Department of Small and Medium Enterprises, identified the essential competencies to become a successful entrepreneur in the small and medium size business sector.

Richard T. Curtin is a Research Professor at the University of Michigan and the Director of the Surveys of Consumers at the Survey Research Center. His research on consumer expectations and behavior is widely utilized by businesses and financial institutions, by federal agencies responsible for monetary and fiscal policies, as well as by academic researchers. He has published more than 500 articles on trends in consumer expectations and their implications for changes in consumer spending and saving behavior and regularly consults with business firms and government officials in dozens of countries. He is a member of the American Economic Association, the National Association for Business Economics, the Association for Consumer Research, the International Association for Research in Economic Psychology, and a member of the Center for International Research on Economic Tendency Surveys. He is the Associate Editor of the Journal of Business Cycle Measurement and Analysis. He received his PhD in economics from the University of Michigan in 1975.

Per Davidsson is Professor of Entrepreneurship at the Jönköping International Business School (JIBS), Sweden. In addition to his professorship, he holds academic appointments in Australia, Latin America, and the United States. He researches entrepreneurship and small business from a variety of perspectives (economic, business, psychological, geographical, sociological). His numerous theoretical, empirical, and methodological contributions have appeared in a multitude of scholarly journals and books. He has served as manuscript editor for Entrepreneurship Theory and Practice and currently serves on the editorial review board for the Journal of Business Venturing and three other scholarly journals.

Amy E. Davis is a PhD candidate at the University of North Carolina at Chapel Hill, Department of Sociology. Her dissertation uses PSED data to examine the conditions under which participation in start-up teams facilitates or impedes mothers' participation in entrepreneurship. She received her MA in sociology at University of North Carolina in 2001. Her thesis examined how voluntary associations affect resource access of business owners in the Research Triangle Park area of North Carolina. She is currently working with Arne Kalleberg on a paper investigating organizations' adoption of work/family employment programs, using data from the National Organizations Study (NOS). She along with Howard E. Aldrich and Linda Renzulli, is also studying how voluntary associations shape the social networks of entrepreneurs and nascent entrepreneurs.

William J. Dennis, Jr., is a Senior Research Fellow at the NFIB Research Foundation in Washington, D.C. and directs the foundation's activities. He has been employed since 1976 in a research capacity by the National Federation of Independent Business, the nation's largest small and independent business trade association. Prior to his affiliation with NFIB, he spent 6 years as a professional staff member in the United States House of Representatives.

His research activities focus on small business and public policy. He is a former President of the International Council for Small Business, recipient of a Special Advocacy Award for research from the United States Small Business Administration, and has served on three panels for the National Academies of Science.

Marc J. Dollinger is Professor of Business Administration in the Management Department at the Kelley School of Business, Indiana University. He received his MBA and PhD from Lehigh University (1978, 1982) in Pennsylvania and spent 5 years at the University of Kentucky before his appointment to Indiana. He is also a Visiting Professor in the International Management Department of the International University of Japan and has taught numerous times at Hong Kong University of Science and Technology.

Currently he is the Chairman of the Kelley School of Business's undergraduate program. He is also a member of the editorial board of Entrepreneurship Theory and Practice and a former board member of the Academy of Management Review. His 1990 paper, The Evolution of Collective Strategies in Fragmented Industries, was awarded the Best Paper Award by the Academy of Management Review. His textbook, Entrepreneurship: Strategies and Resources, was first published in 1995 and is now in its third edition.

Matthew W. Ford is Assistant Professor of Management in the College of Business at Northern Kentucky University. He holds a PhD from the University of Cincinnati. His research interests include organizational self-assessment, small firm planning, and the management and control of change. His published work includes entries in Quality Management Journal, Business Horizons, Frontiers of Entrepreneurship, Center for Quality of Management Journal, and Journal of Engineering & Technology Management.

Patricia G. Greene is Dean of the Undergraduate School at Babson College where she holds the President's Chair in Entrepreneurship. She earned a PhD from the University of Texas at Austin, an MBA from the University of Nevada, Las Vegas, and a BS from the Pennsylvania State University. Her research focuses on the identification, acquisition, and combination of entrepreneurial resources, particularly by women and minority entrepreneurs. She is a founding member of the Diana Project, a research group focusing on women and their business growth strategies.

Gerald E. Hills is holder of the Coleman Foundation Chair in Entrepreneurship at the University of Illinois at Chicago. His MBA and doctorate are from Indiana University. He has written and edited 17 books and written more than 75 articles in entrepreneurship and marketing journals. He has served on the editorial boards of all the leading entrepreneurship journals, including, currently, the Journal of Business Venturing. He is a past President of the International Council for Small Business and cofounder and first President of the United States Association for Small Business and Entrepreneurship. He served as President of the AMA Academic Council, the closest equivalent in the marketing discipline to President of the Academy of Management.

Under his leadership, UIC was ranked by Success Magazine number four nationally in entrepreneurship and most recently by Entrepreneur magazine as number two nationally. He was named a Wilford L. White Fellow of the International Council for Small Business, was named a Fulbright Scholar, and was given the Advocate Award by the Academy of Management for outstanding contributions to the field of entrepreneurship.

Sherrie E. Human is Associate Professor and an Academic Director in the Management and Entrepreneurship Department at Xavier University. She teaches courses at the undergraduate, MBA, and executive MBA levels on new venture creation, the business-planning process, and contemporary management skills. Prior to receiving her PhD in 1995 from the University of Kentucky, she founded and managed technology and technical education companies (along with a sailboat business in the Bahamas). Her research focuses on areas in which she has professional experience such as nascent entrepreneurship, new venture creation, managerial skills development, and areas that she has identified for conceptual and empirical contributions to the literature, such as interorganizational networks and entrepreneurship and ethics. Her research findings have been published in the Academy of Management Journal, Administrative Science Quarterly, Journal of Management Education, Journal of Business Venturing, Entrepreneurship Theory and Practice, Journal of Small Business Management, and Journal of Small Business Strategy.

Kevin LaMont Johnson is an Associate Instructor and PhD candidate in Strategic Management and Entrepreneurship at Indiana University, Kelley School of Business. He received his BA in engineering sciences from Dartmouth College and his MBA from Indiana University. He also served as a Senior Business Development Specialist for several years for a Fortune 500 corporation and has direct independent venture start-up experience. His research involves corporate entrepreneurship and understanding the performance of new ventures within established corporations. He teaches courses in strategic management, leadership, and entrepreneurship at the undergraduate level. He has the honor of being profiled in Who's Who in America® (2004).

Jerome Katz developed the financial sophistication questions of the PSED with Richard Green, PhD, CPA, of the University of the Incarnate Word, San Antonio, Texas. He participated in the PSED as a representative of the Research Institute for Small and Emerging Business. He is the Mary Louise Murray Endowed Professor of Management at the Cook School of Business, Saint Louis University. His research interests include organizational emergence, cognitive and career models of entrepreneurship, entrepreneurship education, and advanced secondary analysis approaches in entrepreneurship. His works have appeared in the Academy of Management Review, the Journal of Management, the Journal of Business Venturing, Entrepreneurship Theory and Practice, and the Journal of Small Business Management. He is Editor of Sage Publications's Entrepreneurship and Management of Growing Organizations series, as well as Senior Series Editor of Elsevier's Advances in Entrepreneurship, Firm Emergence and Growth.

Lisa A. Keister is Associate Professor of Sociology at The Ohio State University. She received her PhD from Cornell University in 1997 and is the recipient of the National Science Foundation's Faculty Early Development Career Award. She conducts research on wealth inequality in the United States and is the author of Wealth in America (2000) published by Cambridge University Press. She is currently conducting a study of wealth mobility in the United States, including an in-depth exploration of the lives of the most influential people in Columbus, OH. When she is not researching wealth ownership, She studies firm behavior during China's transition. Her second book, Chinese Business Groups (2000) published by Oxford University Press reported on that research.

Phillip H. Kim is a doctoral candidate in the Department of Sociology at the University of North Carolina at Chapel Hill. His research interests focus on the study of nascent entrepreneurs and new firm creation processes. Current research projects include the role of financial and human capital resources on achieving operating status and developing improved measures for understanding the new firm creation process.

Bruce A. Kirchhoff is Distinguished Professor of Entrepreneurship and Director of the Technological Entrepreneurship Program at New Jersey Institute of Technology in Newark, New Jersey. His prior credentials include service as Chief Economist for the U.S. Small Business Administration, Director of the Center for Entrepreneurship and Public Policy at Fairleigh Dickinson University, and Director of Research in Babson College's Entrepreneurship Center. He earned his PhD in business administration from the University of Utah, where he also earned an MBA. He received a bachelor of science degree in chemical engineering from Case Institute of Technology.

Prior to receiving his PhD, he spent 7 years in sales and marketing and 3 years as area manager of international operations for Envirotech Corporation. He has served on the faculties of Chalmers Institute of Technology in Sweden; Jönköping International Business School, Sweden; Fairleigh Dickinson University; Babson College; University of Nebraska at Omaha; Purdue University; and California Polytechnic University.

Dr. Jianwen Liao is currently Assistant Professor in the Department of Management, School of Business and Management, at Northeastern Illinois University. He has served on the faculties at DePaul University; Hong Kong University of Science and Technology (HKUST), China Europe International Business School (CEIBS). His research expertise and interests are in the areas of strategic formulation and implementation, management of technological innovation, venture creation process, and entrepreneurial growth strategies. His research has been published in academic journals such as Entrepreneurship Theory and Practice, Family Business Review, Journal of High Tech Management Research, and Frontier of Entrepreneurship Research. He received his doctorate in strategic management from Southern Illinois University at Carbondale.

Tatiana S. Manolova (DBA, Boston University) is Assistant Professor of Management at Suffolk University's Sawyer School of Management. Her current research interests include competitive strategies for new and small companies, international entrepreneurship, and organizational formation and transformation in transitional economies. Recent articles were published in the Journal of Business Venturing (forthcoming), Thunderbird International Business Review (forthcoming), International Small Business Journal, and Journal of Small Business Strategy.

Charles H. Matthews, PhD, is Professor of Strategic Management and Director, U.C. Center for Entrepreneurship Education & Research, College of Business, University of Cincinnati. Dr. Matthews is an internationally recognized scholar and innovative teacher in the field of entrepreneurship. His teaching and research interests include strategic management, decision making, and leadership succession. His research has been published in the Journal of Small Business Management, Journal of Small Business Strategy, Entrepreneurship & Regional Development, Frontiers of Entrepreneurship Research, Family Business Review, and Center for the Quality of Management Journal. An award-winning teacher, he has taught over 5,000 students from freshmen to executives, from individual instruction to classes over 500. In addition to industry experience, he is the founder of the U.C. Center for Entrepreneurship (1997), which was named one of the Top 50 Entrepreneurship Programs in the United States in 2001 (Success magazine) and a Top Tier Program in 2003 (Entrepreneur).

James N. Morgan is a Research Scientist Emeritus at the Institute for Social Research and Professor of Economics Emeritus at the University of Michigan.

He codirected the Surveys of Consumer Finances, funded by the Board of Governors of the Federal Reserve System from 1949 to 1960, directed a pioneering study of income and wealth and its intergenerational aspects (Morgan, David, Cohen, and Brazer, Income and Welfare in the United States, 1962). From 1968 to 1986, he directed or codirected the Panel Study of Income Dynamics. He taught consumer economics and developed a binary segmentation program called SEARCH (AID). He was elected a Fellow of the American Statistical Association in 1968, a Member of the National Academy of Sciences in 1975, a Fellow of the Gerontological Society of America in 1981, and a Fellow of the American Academy of Arts and Sciences in 1984.

Janet P. Near holds the Coleman Chair of Management in the Kelley School of Business at Indiana University. Her research concerns are whistle-blowing in organizations and the relationship between work and nonwork domains of life, focusing on job and life satisfaction. She teaches in the area of organization theory and design. She received her PhD from the State University of New York at Buffalo and a BA from the University of California at Santa Cruz.

Margaret Owen is a doctoral candidate (political science/economics) at the University of Missouri-Kansas City. She currently is graduate research assistant for University Outreach and Extension's Business Research & Information Development Group (BRIDG). Previously she was a graduate research assistant with the UMKC Entrepreneurial Growth Resource Center.

Martin Ruef is Assistant Professor of Organizational Behavior and (by courtesy) of Sociology at Stanford University. He studies the social context of entrepreneurship from both a contemporary and historical perspective. He has pursued this research at several levels of analysis, including examinations of decision making on the part of individual entrepreneurs, team formation and networking among entrepreneurs, founding processes and risks for new ventures, the emergence of novel organizational forms, the “creative destruction” of existing forms, and institutional entrepreneurship leading to the creation of new organizational governance systems. He has studied entrepreneurial activity through archival research on specific fields, such as U.S. healthcare and postbellum agriculture, as well as through representative sampling of entrepreneurs operating in a variety of sectors.

Joseph C. Rode is Assistant Professor of Management in the Richard T. Farmer School of Business at Miami (Ohio) University. His research concerns the relationship between work and nonwork domains of life. He teaches in the area of organizational behavior. He received his PhD, MBA, and BA, all from Indiana University, Bloomington.

Robert P. Singh is Assistant Professor of Management and Director of the Center for Entrepreneurship and Strategy at Morgan State University in Baltimore, Maryland. He completed his PhD at the University of Illinois at Chicago (UIC) in the nationally ranked Institute for Entrepreneurial Studies, where his research focused on strategy and entrepreneurship issues in information technology startup firms. He recently published his first book, Entrepreneurial OpportunityRecognition Through Social Networks (Garland), and his research has appeared in numerous peer-reviewed journals and national/international conferences. In addition to his academic pursuits, he has founded several businesses including Blade Consulting Corporation, a management and information technology consulting firm (1994), and http://BouncingBaby.com, an Internet-based niche portal business (1998).

Timothy M. Stearns is the holder of the Coleman Foundation Chair in Entrepreneurial Studies and Director of the Lyles Center for Innovation and Entrepreneurship at California State University, Fresno. He received his MBA degree in management and a doctorate in management/sociology from Indiana University. He has taught and lectured on entrepreneurship, venture capital, and strategic positioning to entrepreneurs and executives in Thailand, Poland, Japan, Kazakhstan, Macau, and the People's Republic of China. He is known for his research on entrepreneurial startups and the formation of strategic networks. He has served on the board of the Academy of Management Journal and is a founding member of the National Network for Technology Entrepreneurship and Commercialization, a National Science Foundation-funded project that links universities to enhance and coordinate efforts to commercialize technologies (http://www.n2tec.org). He also serves as President of the Central Valley Business Incubator.

Michael Stouder received his doctorate from Rutgers University in New Jersey. He is a visiting Assistant Professor of Strategy and Entrepreneurship at the University of Michigan-Flint. His research interests include early-stage financing behaviors of entrepreneurs and exchange relations between entrepreneurs and resource suppliers.

Harold P. Welsch, PhD, who holds the Coleman Foundation Endowed Chair in Entrepreneurship at DePaul University, has been active in entrepreneurship development for over 20 years in his role as educator, consultant, researcher, entrepreneur, author, and editor. He is well-known for his expertise in technology commercialization, privatization of centrally planned economies, entrepreneurship career paths, formal and informal strategic planning, information seeking and decision behavior, ethnic entrepreneurship, economic development, and small business problems. His work has appeared in many journals. He was co-Editor of Research at the Marketing/Entrepreneurship Interface and recently published a book, Strategic Entrepreneurial Growth with Thompson-Southwestern.

In his position as founder of the Entrepreneurship Program and Coleman Entrepreneurship Center at DePaul University, he has served as Chairman of the Academy of Management Entrepreneurship Division, President of the International Council for Small Business (ICSB), and President of the U.S. Association for Small Business and Entrepreneurship (USASBE).