blinded by the cites
TRANSCRIPT
WP117.DOC
BLINDED BY THE CITES? HAS THERE BEEN PROGRESS IN
ENTREPRENEURSHIP RESEARCH
by
Howard E. Aldrich and Ted Baker
Department of Sociology, CB 3210
University of North Carolina, Chapel Hill 27599
Published in Donald Sexton and Raymond Smilor, Entrepreneurship 2000
email: [email protected]
Home Page: http://www.unc.edu/~healdric/
Paper prepared for presentation at the Fourth State of the Art in Entrepreneurship
Research Conference, sponsored by the Kauffman Foundation Center for Entrepreneurial
Leadership, Inc. Kansas City, Missouri, May 10-11, 1996.
DRAFT: Ok to cite but do not quote: we may change our minds!
2
Introduction
At the first two state of the art conferences on entrepreneurship, authors were
highly critical of entrepreneurship research, faulting it for sloppy thinking and shoddy
methods. At the third conference, Aldrich (1990) noted that the field had expanded its
repertoire of research designs and analytic techniques, but he concluded his review on an
ambiguous note. Rather than directly answering the question of whether entrepreneurship
research had made progress, Aldrich (1990) argued that the answer depended upon one’s
assumptions about the scientific and normative structure of the field. He posed three
viewpoints: a unitary, normal science view; a multiple paradigm view; and a totally
pragmatic view. In our paper for this year’s conference, we examine the past five years’
research, using his three-part framework.
Our title reflects our feeling that much of the dialogue about entrepreneurship
research methods is grounded in researchers’ unreflective situational applications of only
one of the three viewpoints. Some researchers never waver in their viewpoint, whereas
others vary their position, depending on the situation. Entrepreneurship researchers often
accuse one another of not measuring up to the standards set by more established fields.
Do researchers pay too much attention to norms followed by the top-ranked organizations
and economics journals? Has the gap between entrepreneurship and other fields,
identified by earlier critics, been closed? Answers to these questions require both
information on trends in methods and a perspective from which to interpret the trends.
Three Views
The three possible views of the norms governing entrepreneurship research are
based on different assumptions about what counts for “progress” in the field of
3
entrepreneurship research and how it can be achieved. First, people with a unified or
normal science view hold as their ideal the accumulation of empirically tested hypotheses
and well-grounded generalizations, developed through rigorous research designs,
quantitative data, and the latest statistical techniques. Based on strong theories,
investigators test hypotheses to replicate and confirm previous findings, using negative
findings and disconfirmations of previous work as a signal that theories must be
modified. Continuity with the past is the order of the day (Aldrich et al, 1994). Units of
analysis are carefully chosen to reflect theoretical considerations, rather than by the
pragmatics of data collection. Sampling issues are given high priority, as investigators
search for the bounds to their empirical generalizations (McKelvey and Aldrich, 1983).
Achieving a unified scientific program requires disciplined adherence to accepted
practices, and support may be found in training institutes and the institutional discipline
imposed by senior members in the field. However, the entrepreneurship field lacks an
institutionalized set of programs teaching a consistent paradigm to new scholars. Instead,
recruits are trained by established academics from a variety of disciplines and are exposed
to a wide variety of standards and methods, and they pursue research which frequently
articulates poorly with the contributions of their entrepreneurship colleagues or
predecessors. However, some of the leading institutions in the field, particularly Babson
College and the Journal of Business Venturing, may be providing some pressure towards
coherence through standard setting, dialogue between scholars, and the building of
community.
Second, people with a more eclectic view of entrepreneurship research emphasize
the importance of diversity in theories and methods. Low and MacMillan (1988: 154-
4
155), for example, noted the wide variety of disciplines involved in entrepreneurship
research. Rather than lamenting disagreements between various subgroups of
entrepreneurship researchers, advocates for diversity welcome the invigorating effect of
multiple perspectives and methods on research. If we adopted a multiple paradigms
perspective, we would neither expect nor wish for complete convergence in methods.
Instead, different sub-fields would have different standards against which to judge
research design and execution, and perhaps even their own journals. Encouraging
diversity, however, brings with it some dangers, because subgroups may not only
construct too narrow a slice of the field, but also waste efforts emphasizing how they
differ from one another. If subgroups develop into coherent and self-contained entities,
they may irrevocably fragment the field.
If there are multiple, disparate conferences and journals with widely varying
standards, the scholarly community may become fragmented, and member of different
camps may fail to communicate with one another. Within the field of business strategy --
another young, empirical, multi-disciplinary field -- we have seen the development of
institutions that parallel the role played in the entrepreneurship field by the Kauffman-
Babson Conferences and JBV: the Strategic Management Society Annual meetings and
the Strategic Management Journal. The strategy field may be a useful benchmark for
assessing developments in entrepreneurship research.
Third, if we adopt a more pragmatic approach, focusing on the importance of the
issues addressed, the question of research methods assumes a secondary role. From this
point of view, methods should be chosen to match researchers’ purposes, and these will
change as conditions change. In entrepreneurship research, pragmatism may be oriented
5
toward either policy or practitioner concerns. Current socio-political factors will govern
what issues are researched, and changing conditions, not abstract methodological
concerns, will drive research practice. Positive values will be placed on topicality,
uniqueness, and usefulness, rather than adherence to a confining code of research
practices. Indeed, rewards for peculiarity and topical value can create feedback loops that
destroy researchers commitment to normal science norms, promoting greater pursuit of
uniqueness rather than continuity (Mone and McKinley, 1993).
We’ve suggested that academic journals and conferences can help to either pull
together or fragment a research field. Similarly, in a pragmatically-oriented field, the
relationship between academic and popular journals and the intermingling of authors and
topics may shape the influence that practitioner concerns have on scholarly coherence.
As we summarize previous reviews and present our findings for trends over the
past five years, we will use these three views as benchmarks against which to evaluate the
results.
Previous Reviews
In three previous state of the art conferences, thirteen authors of seven articles
reviewed research methods. At the first conference, Paulin, Coffey, and Spaulding (1982)
found more formal research on entrepreneurship than they had anticipated, and they also
discerned a trend toward more systematic, empirical methods. They were fairly critical of
the state of the art, however, arguing that much of the “knowledge” in the field was based
on untested or narrowly based anecdotal wisdom. They called for more diversity in
research methods; specifically, they advocated more field studies and experiments,
longitudinal studies, and rigorous statistical analyses. Peterson and Horvath (1982)
6
appealed for more cross-national studies, as well as more precise and meaningful
definitions of research questions. In contrast, Perryman (1982) defended the imprecision
and chaos of the emerging field, arguing that entrepreneurship research was still in its
“pre-science” phase.
At the second conference, Wortman (1986) cast a withering glance on the state of
the art, and pronounced it decidedly inferior to the research methods being used in other
social science fields. Not only did he excoriate entrepreneurship researchers for their
weak statistical analyses, but he also advocated a single unifying framework that would
integrate entrepreneurship and small business research. At the same conference,
Churchill and Lewis (1986) took a more encompassing and sympathetic view of the field,
using a systematic review of Babson Conference papers and 10 academic journals to
build empirical generalizations about entrepreneurship research methods. They seconded
Perryman’s (1982) observation that the field was still in its infancy and hence dependent
on exploratory research and simple statistics. Churchill and Lewis introduced a theme
that will also concern us: a practice-oriented discipline has some difficulties that more
theory-oriented disciplines do not have. We will have more to say about their
classification scheme for research methods -- shown in Table 1 -- in the next section.
The other methods paper at the second conference, Carsrud, Olm, and Eddy (1986),
reprised Wortman’s theme: investigators were spending excessive amounts of time on
descriptive statistics and case studies, using unsophisticated methods. They also
proposed that the field adopt a unified research paradigm, employing a comprehensive
categorization scheme for variables, although they shied away from nominating any
particular framework.
7
At the third conference, Aldrich (1990) replicated Churchill and Lewis’s (1986)
analysis, examining all articles in the 1986 and 1989 Babson Conference volumes and in
the same ten academic journals, plus the Journal of Business Venturing. As we review
his results again in our Table 2, we simply mention now that he found strong continuity in
research methods across all three conferences. Entrepreneurship research was still very
much a mono-method field, dependent on mailed surveys and other questionnaire-based
techniques, with fairly unsophisticated data analyses. Field-based research designs were
notable by their absence.
Beginning in the late 1980s, other voices were added to those of state of the art
reviewers, calling for more explicit attention to research methods in entrepreneurship. In
1986, Churchill and Lewis found only six papers -- all presented at conferences between
1981 and 1984 -- that dealt with methodological improvements in the field. In our review
of the literature, we found that at least 16 journal articles published between 1987 and
1993 that either provided general reviews of methods or advocated a particular remedy to
what were seen as methodological deficiencies. We think the heightened level of self-
consciousness about how research is carried out indicated that some investigators were
trying to push the field toward a more “normal science” posture, whereas others with a
more multiple-paradigms view were resisting attempts at “premature closure.”
Bringing Us Up to Date
To maintain continuity with past reviews, we used the same strategy as our
predecessors in organizing and defining the field of entrepreneurship research. The
Babson College Entrepreneurship Conference is still the premier venue for disseminating
unpublished research on entrepreneurship, and we examined all the articles and
8
summaries in the 1990 and 1994 conference volumes. We found 77 papers in the 1990
Babson volume and 111 in the 1994 volume.
In 1990, Aldrich (1990) followed Churchill and Lewis’s procedures for choosing
the 10 academic journals which published a large number of entrepreneurship articles,
and he added the Journal of Business Venturing (JBV), which began in 1985. These
journals fell naturally into two groups: Group I, which published primarily empirical
articles, and Group II, which published primarily conceptual or “think” pieces. The
journals are listed in Appendix A, classified into the two groups, along with the number
of articles found in each source. Because our focus is on research methods, we decided to
eliminate Group II journals from our analysis, and hence only articles from Group I
journals appear in our tables.
We searched the abstracts of all journals in the ABI/INFORM data base, as was
done for all three previous conference papers. We used the same key words as in Aldrich
(1999): entrepreneur, corporate venturing, and intrapreneur, as well as variations on all
three words. We did not search on “small business.” Our search uncovered 164 articles
in the Group I journals. Our goal was to include all articles published between 1991 and
1995, but Entrepreneurship: Theory and Practice (ETP) did not publish a 1995 issue until
after our search was completed, and the Journal of Business Venturing’s 1995 volume
was not in our library at the time of the search..
Many authors reviewing entrepreneurship methods have made implicit
comparisons to “mainstream” organization studies journals, and so we decided to include
a specific comparison group of journals in our study. The Administrative Science
Quarterly and the Academy of Management Journal are the oldest and most prestigious of
9
the organization and management journals, and so we examined all the empirical articles
they published in 1990 and 1995. All articles were coded, regardless of subject matter.
As they publish almost no non-empirical articles, these journals provide a good standard
of comparison for examining research methods. As luck would have it, there was no
overlap between the articles found in our Group I search and the articles coded from ASQ
and AMJ.
Trends in Methodological Focus
We used the classification scheme of Churchill and Lewis (1986), as modified by
Aldrich (1990), and added several new categories, as shown in Table 1. The added
categories signify several changes in the field. First, in the 1990s ETP began publishing
case studies of entrepreneurial firms as a special section of the journal, evidently as a
service to instructors looking for classroom discussion material. Other journals have also
begun publishing case studies that cannot really be classified as ethnographies, because
they are not based on extensive field work inside the organizations. Second, in addition
to a category called “armchair, or observational and contemplative theory building” in the
last several reviews, we added a new category of literature reviews as an
acknowledgment of a more disciplined type of conceptual article. Literature reviews are
an explicit effort to summarize what other authors have said about a specific topic, rather
than an attempt at model building or commentary. Third, the occasional papers on
research methods in the 1980s apparently spawned a more concentrated effort by authors
in the 1990s to deal with methodological issues, and so we separated such papers out
from other literature reviews. Fourth, the Babson College Conference organizers have
taken note of initiatives by the Kauffman Foundation and others to put entrepreneurship
10
education on a more solid empirical footing, and entrepreneurship journals have also
begun paying more attention to that topic. To maintain continuity with earlier analyses,
we have not included literature reviews and methods/entrepreneurship education topics in
the percents reported in Table 2 and subsequent analyses.
TABLE 1 ABOUT HERE
How much has changed since the last state of the art conference? Not much, as
shown in Table 2. Indeed, research design and sources of data have not changed very
much over the past 15 years, other than a decisive break with journalistic and armchair
methods by the journals after 1985. In contrast to a 69 percent share of journal articles in
1981-84, journalistic and armchair articles made up only a 35 percent share in the most
recent period. In keeping with the Babson Conference’s mandate, almost no journalistic
or armchair papers were accepted in either 1990 or 1994, and both Administrative
Science Quarterly and Academy of Management Journal effectively bar such papers.
TABLE 2 ABOUT HERE
Despite repeated calls in previous state of the art reviews and journal articles on
methods, survey research has maintained an almost constant share of Babson Conference
papers over the past decade, with a 78 percent share in 1986 and a 77 percent share in
1994. Among the entrepreneurship journals, survey methods also maintained their
prominent place, rising only slightly from 43 to 48 percent, after a sharp jump from 1981-
84 (when non-empirical methods predominated). A casual glance at the last two columns
of Table 2 might give one the impression that entrepreneurship journals and ASQ/AMJ
shared the same enthusiasm for survey research, but the nearly similar proportions of
survey-based articles is misleading.
11
In addition to publishing almost no journalistic and armchair papers, the two
general organizations journals included a very large number of empirical papers using
publicly-available data. Whereas only 9 percent of articles published in the
entrepreneurship journals used publicly-available data, 43 percent of the articles
published in ASQ/AMJ did so. With so many papers using data from sources accessible
to all investigators, ASQ/AMJ have built a pool of empirical results that facilitates
replication and disconfirmation research by other scholars. By contrast, most
entrepreneurship researchers still work with proprietary data they’ve collected
themselves, mainly through questionnaires
Ethnographic research has actually declined in significance at Babson
Conferences, and it has fared poorly in the journals, as well. After a modest showing at
the 1986 and 1989 Babson meetings, ethnographers took a holiday in the 1990s, with only
2 reporting at the 1990 and 1994 meetings. Ethnographic studies were also unpopular
with both entrepreneurship journals and authors of ASQ/AMJ articles: only about 3
percent of the papers published in either set featured ethnographic research. The
continued isolation of ethnography reveals the futility of passionate exhortations to
unwilling researchers -- commentators have repeatedly urged that more such research be
conducted, and almost every reviewer at previous state of the art conferences urged
“more, more, more.”
Simulations and experiments are another category of methods that reviewers have
proposed to counter the field’s dependence on surveys, but they have never appeared in
any of the entrepreneurship journal articles covered by reviews. Low and MacMillan
(1988) noted that experiments were extremely rare -- they cited only 2 successful ones
12
from the 1980s. They have also rarely made appearances at Babson conferences. By
contrast, about one in twelve of the papers published in ASQ/AMJ have been simulations
or experiments, mostly testing social-psychological hypotheses.
“Case studies” takes account of non-ethnographic empirical reports on single
firms or industries, typically prepared for teaching purposes. A handful of such studies
have been presented at Babson conferences, and ETP has been including them in a special
appendix to some issues of the journal. The mandates of ASQ/AMJ rule out case studies
in this mold.
Three categories of papers and articles are included in the coding scheme of Table
1 but are not shown in Table 2, because we wanted to make the percentages shown in the
table as comparable as possible to earlier reviews: literature reviews, and methods and
entrepreneurship education reviews. At the 1990 Babson meeting, one methods review
and two entrepreneurship education papers were presented, and in 1994 these numbers
had increased to one literature review, three methods reviews, and 14 entrepreneurship
education papers. Entrepreneurship journals began featuring such articles in the 1990s, as
we found 10 literature reviews, 7 methods reviews, and 10 papers on entrepreneurship
education. If this trend continues, we hope the next state of the art conference will devote
more attention to a detailed examination of these categories. In subsequent sections, we
will have more to say about the methods reviews.
Trends in Research Design
Research design issues involve not only data collection strategies but also
strategies affecting the scope of a research project: what topic to focus on, what
populations to study, how to define a sampling frame, and so forth. Authors of papers for
13
the first and second state of the art conferences discussed such issues, but did not collect
data on them. For the third conference, Aldrich (1990) presented data on nine indicators,
which we have replicated and extended for our analysis: (1) personality traits versus
other foci; (2) inclusion of a non-US nation; (3) inclusion of more than one nation; (4)
reliance on an identifiable sampling frame; (5) identification of a homogeneous versus
non-homogeneous population; (6) longitudinal versus cross-sectional design; (7) sample
response rate; (8) number of cases in the study; and (9) the statistical methods used. We
added a new indicator: (10) probabilistic or random sampling versus other procedures.
For the two premier entrepreneurship journals -- ETP and JBV -- we also
collected information on four new indicators that directly address the extent to which
investigators are following a normal science strategy: (11) extent to which the study
replicates previous research; (12) presentation of explicit hypotheses; (13) tests for the
reliability of measures; and (14) reporting of negative findings. We also collected the
same new information for the top-rated organization and management journal, ASQ.
We only present results from the Group I journals identified in Appendix A, as
well as Babson, because they published almost all the empirical papers in
entrepreneurship. The coding scheme is shown in full in Appendix B.
TABLE 3 ABOUT HERE
Changes in some aspects of research design have occurred, as shown in Table 3,
but we are more impressed with the continuity between our results and those of the last
round.
We chose only one topic for detailed investigation -- the extent to which articles
focused on the traits of entrepreneurs versus other themes. Debate over the usefulness of
14
studying entrepreneurial traits has been raging for over a decade, and a small committed
community of researchers continues to publish traits research. At the Babson meetings,
personality traits research has fluctuated between 13 and 20 percent of papers presented,
with the most recent conference (1994) at 14 percent. Only 15 percent of the
entrepreneurship journal articles between 1985-90 emphasized an entrepreneur’s
personality traits, but between 1991-1994, 31 percent of journal articles focused on this
theme. ETP had the highest percent, 44, but JBV was not far behind, at 32 percent, and
the JSB was close by, at 26 percent. In sharp contrast to the entrepreneurship journals,
AMJ/ASQ published almost no traits articles -- only 5 percent fell into this category.
We are puzzled by the rise in traits articles in entrepreneurship journals, especially
because so many articles critical of traits research have been published over the past
decade (Gartner, 1988, 1989; Aldrich and Wiedenmayer, 1993). Moreover, traits articles
have not increased as a proportion of Babson conference papers. All three
entrepreneurship journals for which the case base is sufficient to allow separate
computation of percents have at least one-quarter of their papers in the traits category,
and so we must look to some cross-journal factor for an explanation. Perhaps traits
articles, because they fall into the psychological paradigm, are easier to review and more
likely to win favor with reviewers than non-traits articles.
Less parochial than before?
At the first state of the art conference, calls were issued for more cross-national
research, and since then, international participation at Babson conference meetings has
steadily increased. Several initiatives over the past decade have expanded international
involvement, such as Babson College’s holding two of its conferences overseas. Sue
15
Birley and Ian MacMillan’s founding of the Global Entrepreneurship Conference, with its
associated edited volumes (Birley and MacMillan, 1992, 1995), was explicitly designed
to involve more non-US scholars and their students in entrepreneurship research. The
fruits of these efforts are apparent in the steady increase in the proportion of papers and
articles reporting research on a non-US nation. For Babson, the percent began at 33 in
1986, dropped to 28 in 1989, but then rose to 31 in 1990 , and climbed to 43 by 1994.
Similarly, 18 percent of the journal articles in the 1985-90 period included a non-US
nation, growing to 40 percent in the 1991-95 period. By contrast, AMJ and ASQ are
much more parochial journals, with about 82 percent of their articles focusing only on US
cases.
As in previous years, a much lower percentage of articles are truly comparative --
including more than one nation -- than they are international (using a non-US nation). In
1986, only 5 percent of the Babson papers were comparative, but by 1995 this percentage
had more than tripled, to 16 percent. In the 1985-90 period, only 4 percent of the journal
articles were comparative, and again this percentage more than tripled by 1991-95, to 14
percent. Comparative studies made up only about 8 percent of AMJ and ASQ’s articles.
The gap between the high percentage of international contributions and the relatively low
percentage of comparative contributions can be attributed to the slow growth of
collaborative relations and alliances between scholars from different nations. As Aldrich
(1990) noted, enthusiasm for entrepreneurship research has become global, but
researchers have lagged in responding to Peterson and Horvath’s (1982) observation that
cultural values are critical in accounting for national differences in rates of
entrepreneurship. Perhaps now that about two out of five entrepreneurship papers use
16
non-US data sets, the pace of international collaboration will quicken. AMJ/ASQ might
benefit from adopting the more international spirit of the entrepreneurship journals.
Sampling
Sampling procedures are central in a normal science view of entrepreneurship.
The normal science paradigm emphasizes building empirical generalizations by
identifying the scope conditions under which a principle holds (McKelvey and Aldrich,
1983) and collecting data from populations that are homogeneous enough to allow
generalization to apply to most of their units. Two interrelated decisions are involved:
what population is an appropriate context in which to test an hypothesis, and what
sampling frame will generate a representative sample of that population or populations.
These decisions clearly affect a third decision, as well: what statistical techniques to use.
Most modern statistics are based on the assumption that a representative sample has been
drawn from a larger, limited population. Another research design decision that affects
statistical decisions concerns whether to use a cross-sectional or longitudinal design, as
dynamic analyses are usually more complicated and trouble-prone than static ones.
State of the art reviews at the first two conferences did not explicitly address
sampling and population definition issues, but authors implied that researchers had not
been as systematic as the should have been. Many of the studies reviewed were based on
convenience or quota samples, rather than being drawn from identifiable sampling
frames. In 1990, Aldrich presented evidence showing that 62 percent of the journal
articles in 1985-90 used an explicit sampling frame, compared to only 35 percent of the
1989 Babson papers. As shown in Table 3, there has been no change in the sampling
17
frames of journal articles since then, whereas Babson conference papers have been
inconsistent, rising and falling over the years.
For 1990-95, we expanded the classification of sampling frames to include four
sub-categories: use of a sub-national frame, and use of a national or international frame,
use of a random sample or complete census of the identified population, and choice of a
cross-sectional or dynamic data collection strategy.
We were interested in whether investigators were attempting to generalize their
results to entire nations, rather than limiting their claims to local or regional populations.
For Babson, no trend is apparent between 1990 and 1994, as the percent of papers with an
identifiable sampling frame dropped from 63 to 47 percent, and the percent with national
or international frames dropped from 41 to 16 percent.
For entrepreneurship journals, we can make two comparisons -- to the Babson
conferences, and to AMJ/ASQ articles. First, as we noted, the journals have been very
consistent in the percent of articles reporting an explicit sampling frame: 62 percent in
each of the two study periods. Within the 1991-95 group, the percent of
national/international frames in journals was lower than the 1990 Babson Conference but
higher than the 1994 meeting. Second, AMJ/ASQ clearly enforce a different set of
standards on their submissions than do the entrepreneurship journals, as not only did 87
percent of their papers report an identifiable sampling frame, but also almost half -- 49
percent -- used a national or international frame. This striking difference means that
readers of AMJ/ASQ articles can more easily envision generalizing the results of what
they read to a national level than can the readers of entrepreneurship journals. Whether it
is reasonable to do so, however, depends on two other sampling decisions: did the
18
investigators identify a reasonably homogeneous population or set of populations, and
was the sample truly representative of the population or populations?
Population definitions are also critical to investigators’ abilities to generalize from
their results. Investigators who draw illustrations of their ideas indiscriminately from
diverse industries and those who draw samples of firms from convenient lists are
operating on the principle that all organizations are pretty much alike (McKelvey and
Aldrich, 1983). Investigators who devote entire articles to a case study of a single
organization and who disdain explicit comparisons with others are operating on the
principle that each organization is unique. By contrast, some investigators explicitly limit
the scope of their generalizations to specific types of organizations, such as venture
capital firms, semiconductor manufacturers, or California wineries.
AMJ and ASQ again stand out in comparison to entrepreneurship journals and
Babson Conference papers: 46 percent of their articles identified a homogeneous
population, with about 80 percent of such articles focusing on a single population and the
rest on multiple homogeneous populations. Using an explicit sampling frame
substantially increases the likelihood that a homogeneous population will be identified by
AMJ/ASQ authors: with no frame, 24 percent of the articles involved a homogeneous
population; with a sub-national frame, the percent increased to 47 percent; and with a
national or international frame, the percent increased further to 61 percent. Within the
entrepreneurship journals, there was no association between the type of frame used and
the resulting population. Taken together with the strong emphasis in AMJ/ASQ on using
an identifiable sampling frame, these results are another manifestation of the strong
normal science norms underlying AMJ/ASQ’s editorial strategy.
19
The contrast with entrepreneurship journals and Babson papers is clear-cut: the
percent of entrepreneurship journal articles reporting on homogeneous populations
actually fell from 19 percent in 1985-90 to 16 percent in 1991-95, whereas Babson’s
papers have fluctuated between 15 and 29 percent over the past decade. In contrast to
AMJ/ASQ papers, Babson papers were about equally divided between single and
multiple homogeneous populations, and the small fraction of entrepreneurship journal
articles with a homogeneous population was slanted toward multiple populations.
On the third sampling issue -- the representatives of the sample -- the distinct
differences noted above between the different journals and Babson disappeared. We
ranked sampling decisions from (1) not a probabilistic, random, or systematic sample; (2)
a probabilistic, random, or systematic sample; and (3) a complete census of the identified
population or populations. As shown in Table 3, almost half of the articles in all sources
did not draw a sample from the population or populations identified by their frame.
Indeed, truly probabilistic or random samples made up only between 10 and 24 percent of
the studies between 1990 and 1995.
The fourth sampling issue involves deciding between cross-sectional and
longitudinal data collection, taking account of resource constraints and problems of
access to data from more than one time point. In previous methodological reviews,
almost every author decried the scarcity of longitudinal designs and dynamic statistical
analyses. Aldrich (1990) thought he detected some hopeful signs in his review of
research between 1985 and 1990, but a closer look at his coding scheme revealed that he
had been too generous in crediting investigators with longitudinal designs. Accordingly,
20
we started over again with our coding of the 1990-95 period, as shown in Table 3, where
we have entered asterisks for columns from previous years.
Very few entrepreneurship conference papers or articles used an explicit dynamic
design: 9 percent in 1990 and 16 percent in 1994 for Babson, and only 5 percent for
entrepreneurship journals in 1991-95. AMJ/ASQ articles, by contrast, were much more
likely to use dynamic designs, with one-quarter of all papers making use of longitudinal
data sets. The considerable gap between AMJ/ASQ and the entrepreneurship journals
cannot be fully explained by resource constraints on entrepreneurship investigators, we
believe, because entrepreneurship journal articles are now using samples that are
equivalent in size to those in AMJ/ASQ, as we note below. Moreover, entrepreneurship
researchers are more likely than AMJ/ASQ authors to use cross-national designs, which
undoubtedly involve more logistical and organizational problems than one-nation
designs. Instead, the difference in use of longitudinal data may stem from differences in
how problems are conceptualized and how much attention is given to the fit between
theories of action and the data needed to test such theories.
Replication and hypothesis testing
Normal science norms concerning “progress” strongly support four other
practices: replicating previous studies before placing much confidence in their results,
formulating explicit a priori hypotheses to avoid data-dredging and dust-bowl empiricism,
checking the reliability of one’s measures whenever possible, and publishing negative or
disconfirming results rather than only supportive ones. Woo, Cooper, and Dunkelberg
(1991) demonstrated the value of replication when they found that a common typology –
craftsmen versus opportunists – did not stand up to closer scrutiny. These practices are
21
held up as a standard against which to judge research, but note that they also have the
effect of slowing down the introduction of innovative topics. Authors who adopt a more
pragmatic, topical orientation may dismiss these practices as stultifying and antithetical to
creativity in entrepreneurship research.
As an experiment, we attempted to code articles in ETP, JBV, and ASQ along
these four dimensions. (Because the process was time-consuming and we felt ASQ was
the most stringent benchmark, we did not code AMJ articles on these dimensions.) The
coding process itself was quite revealing, because we discovered how the norms
regarding professional discourse differ across fields. ASQ editors and reviewers
obviously expect to see precedents clearly listed and hypotheses formally stated, whereas
we often had to wade through paragraphs of prose for such information in ETP and JBV.
Most articles in all three journals made some mention of theoretical precedents for
their research -- 86 percent of ETP/JBV and 88 percent of ASQ papers -- but ASQ
authors were more likely to claim a direct replication of previous work than were
ETP/JBV authors. We found two types of direct replications. First, a handful of papers
specifically replicated a prior finding by choosing the same population, the same
concepts, the same indicators, and examining the same relationships. Second, and more
typical, authors attempted to confirm the form of a relationship that was previously
observed, such as the pattern of density dependence in founding rates, or a finding that
early adopters of an innovation are more central in a network than late adopters. A key
criterion was that an author not only cited general precedents for examining a problem but
also cited specific empirical generalizations as justification for testing a proposition
22
(again). We found that 32 percent of the ASQ and 12 percent of the ETP/JBV articles
fulfilled this criterion.
More common were authors who used earlier papers as justification for applying
an older idea to a new domain, arguing that a concept or principle was important, without
citing previous empirical generalizations. The new study may have created hypotheses
that don’t specifically refer to a past study, using concepts that cover a wide range of
previous work, with indicators unique to the new study. About 56 percent of the ASQ
and 74 percent of the ETP/JBV papers fell into this category.
Even the relatively small percent of papers we classified as non-replications (14
percent of ETP/JBV and 12 percent of ASQ) made at least passing reference to earlier
work, but in a general way and without regard to any empirical generalizations. Most
papers went beyond such superficial acknowledgment of the past, however. Authors of
ASQ papers were especially advantaged, as they had more freedom to write long
introduction than authors in the other journals, giving them more space to develop their
concepts and models.
Hypothesis testing practices differed dramatically between the two sets of
journals: 83 percent of the ASQ papers contained explicit hypotheses, compared to 48
percent of ETP/JBV papers. In the roughly one-half of the ETP/JBV articles without
explicit hypotheses, we found that authors posed questions that were descriptive and did
not predict the form or sign of an expected relation. The researchers provided rather
diffuse theoretical discussions which did not lend themselves to hypothesis development.
Among papers that were not replications, only 14 percent spelled out hypotheses. Among
23
papers that were indirect replications, 61 percent reported explicit hypotheses, and 100
percent of the papers that were direct replications developed formal hypotheses.
Without clearly specifying hypotheses, authors cannot really know if their results
confirm or disconfirm previous findings. In any case, as we show in Table 3, very few
published papers reported findings that conflicted with the expectations authors laid out
in their theoretical introductions. Our criterion for negative findings was that 50 percent
or more of the measures of association reported had to be non-significant, and only 17
percent of the articles in both sets of journals contained such a pattern. Among articles
with explicit hypotheses, 23 percent reported negative findings, compared to only 6
percent of authors who’d failed to set out formal expectations (2 out of 35 articles). This
pattern did not differ significantly between ASQ and ETP/JBV.
Determining whether a study actually obtained negative findings was difficult, a
sign that norms surrounding research practice are still in flux, even in the most
prestigious journal in organization studies. Even if authors explicitly stated their
hypotheses, their results and discussion sections often did not refer back to which
hypotheses were confirmed and which were not. Often, authors concluded their papers by
ignoring what they had said in their theory review section. Even in ASQ, authors who
presented formal, numbered hypotheses sometimes ignored them when discussing their
results, e.g. Greve (1995)!
Our next assessment of normal science practices was whether authors tested the
reliability of at least some of their measures. Reliability was easiest to code when
personality traits or other individual characteristics were studied, for two reasons. First,
authors tended to adopt measures with previously documented reliability. Second,
24
authors studying traits or using individuals as units of analysis and data collection tended
to use multiple item questionnaires, allowing the collection of multiple indicators, and
thus their raw data lent itself to easy manipulation. When organizations were the unit of
analysis, authors apparently either assumed that their data were reliable, or ignored this
issue. We noticed that when organizational records were used, for individuals or
organizations, investigators paid almost no attention to their reliability, or to the use of
multiple indicators.
Two other problems came to our attention. First, authors sometimes extensively
discussed data reduction techniques, such as factor analysis or cluster analysis, but then
did not examine the reliability of the resulting new measures. Second, even studies that
reported reliability tests for some variables omitted it for many others. For example,
researchers performed reliability checks on individual-level traits but not for group or
organizational level characteristics.
About 31 percent of the ETP/JBV papers reported reliability checks on at least
some of their measures, compared to 42 percent of the ASQ papers. We found a strong
association between authors’ attempting to replicate prior research and doing at least
some testing for data reliability. Reliability testing was performed in 7 percent of the
non-replications, 37 percent of the indirect replications, and 50 percent of the direct
replications.
Without data from prior years, we cannot tell whether the patterns uncovered here
are a continuation of past practices or points in a long-term trend. We have results that
will encourage some normal science advocates but disappoint others. Almost all authors
made some pretense of linking their work to the past, but only a minority of
25
entrepreneurship researchers situated themselves squarely in a stream of empirical
generalizations that they intended to replicate. About half of the entrepreneurship papers
developed formal expectations or hypotheses, an encouraging sign but one that pales by
comparison to ASQ’s standard. Few studies disappointed their authors, as over 80
percent reported confirming findings, thus providing few clues as to what to discard for
the next round of research projects.
Samples: response rates and sample size
Regardless of whether investigators chose cases from an identifiable sampling
frame, many nonetheless treated the resulting target population of respondents or cases as
a “sample.” They then reported how many cases they obtained information from, out of
the original target population, enabling them to report a response rate for their efforts. In
Table 3, we report response rates for all studies using surveys or public data bases. Some
studies did not report response rates, possibly biasing our figures upwards.
As Aldrich (1990) noted, response rates to entrepreneurship surveys have never
been very high, and that tradition continued in 1990-95. Response rates for studies
presented at Babson conferences have fluctuated somewhat over the past decade, and may
have become worse. In 1986, 60 percent of the studies reported response rates below 50
percent; in 1989, it was 67 percent; in 1990, 70 percent, and in 1994, 66 percent. For the
journals, the latest period was a step backward, as 74 percent reported response rates
below 50 percent, compared to 50 percent in 1985-90. At these levels of response from
the target population, serious questions arise about possible sample bias. Only a small
percent of the studies reported tests for possible sample selection bias.
26
Response rates for ASQ/AMJ were higher than for the entrepreneurship journals
in 1990-95, as 60 percent of their papers reported rates of 50 percent or better. Because
we have no data for AMJ/ASQ from the earlier period, we don’t know whether this is an
improvement. We do note that only 8 percent of the articles reported response rates
below 25 percent, compared to about one-third of papers published in the
entrepreneurship journals or presented at Babson in recent years.
In calculating the number of cases on which a study was based, we included not
only surveys but also ethnographies and case studies, to maintain continuity with reviews
at previous conferences. At the last state of the art conference, sample sizes were larger
in Babson conference papers than in the entrepreneurship journals, and Aldrich (1990)
thought he detected a trend toward increasing sample sizes. His predictions were
premature, however, at least insofar as Babson is concerned. In 1990, 60 percent of the
Babson papers worked with samples under 100, and in 1994, the percent had actually
risen to 67.
In contrast, the number of cases analyzed in journal articles increased substantially
between 1985-90 and 1991-95. In the earlier period, about 60 percent of the studies were
based on fewer than 100 cases, and the figure dropped to only 38 percent in the latest
period. Indeed, the study size distributions for entrepreneurship journals and AMJ/ASQ
are statistically indistinguishable, especially in their tails -- both have about the same
percentage of very small and very large scale studies.
Statistical methods
At the first state of the art conference, Paulin, Coffey, and Spaulding (1982) were
critical of entrepreneurship researchers’ lack of statistical sophistication, but Perryman
27
(1982) adopted a more restrained tone, arguing that the field was still in an immature
phase. At the second conference, Wortman (1986) wrote an extremely critical paper,
scolding researchers who settled for crude descriptive methods and challenging the field
to borrow methods already in use in other social science fields. Carsrud, Olm, and Eddy
(1986) also criticized the descriptive orientation of many papers they reviewed, whereas
Churchill and Lewis (1986) reprised Perryman’s (1982) more benevolent tone, citing the
field’s immaturity. At the third conference, Aldrich (1990) noted that investigators
continued to rely heavily on simple descriptive statistics, but he detected a trend toward
greater use of significance tests and models assessing the strength of associations between
variables.
In preparing our report for this conference, we took account of the evolving nature
of analytic techniques and added two new categories to our classification scheme: “other
regression and discriminant analysis” and “non-recursive, event history, or formal
network models.” These categories will ultimately prove their value when they are used
again at the next conference.
If we use articles in AMJ/ASQ as a standard against which to judge the field, then
the statistical-sophistication gap noted in earlier critiques is still enormous: 81 percent of
the papers in AMJ/ASQ use ordinary least squares or more sophisticated methods to
analyze their data, compared to only 34 percent in entrepreneurship journals and 39
percent at the 1994 Babson conference. At the other extreme, only 6 percent of the
articles in AMJ/ASQ rely on simple percents or raw numbers, compared to 35 percent in
the entrepreneurship journals and 28 percent at the 1994 Babson conference.
28
If we use changes from an earlier period as a standard against which to judge the
present state of the field, then the statistical-sophistication gap may be closing, although
progress has been slow. Simple percents, raw numbers, or no numbers at all dropped
from 53 percent at the 1986 Babson conference to 28 percent in 1994. In
entrepreneurship journals, use of methods such as regression analysis and other methods
that explore the form of associations between variables increased from 28 percent in
1985-90 to 34 percent in 1991-95. More sophisticated forms of analysis -- beyond
ordinary least squares -- have also appeared, including event history analysis.
Babson as an Incubator
Have the Babson College conferences been playing a role as an incubator for
papers that are revised and then published in entrepreneurship journals? Aldrich’s (1990)
observed that the level of statistical sophistication at the most recent Babson conference
was nearly the same as in the empirically-oriented entrepreneurship journals. If Babson
conferences have served as incubator, then they may have contributed to the
standardization of research practices in entrepreneurship, as well as creating a core
community of researchers who can play gatekeeper roles in the profession and enforce its
emerging standards.
We investigated this question by asking the 20 researchers most active at Babson
from 1990 to 1994 whether they had published their Babson papers. The 20 authors’
names appeared 151 times in the Babson Conference Proceedings between 1990 and
1994. We sent them letters, asking whether their Babson papers had been published, and
where. Fifteen people replied, representing 117 Babson papers. Forty percent of the
papers had been published, 6 percent were under review, 6 percent were still in process,
29
and 3 percent were in limbo, with the authors having lost track of them (co-authors may
have been working on them). The rest were unpublished and no longer being worked on.
Of the 47 papers that had been published, 14 were in JBV, 10 in books or
monographs, 3 in Entrepreneurship and Regional Development, 3 in the Journal of
Business Research, and the remainder were scattered over 13 other journals. We have no
baseline from which to judge the relative success of these authors, but three points stand
out. First, over half of the papers presented have either been published or were still under
active revision. These authors have treated the Babson conferences as an opportunity for
the first presentation of their ideas, rather than as an end it itself. Second, since its
founding in 1985, JBV has become the premier outlet for entrepreneurship research, and
it has been the journal these Babson participants have considered first when preparing
their work for publication. Third, the diversity of outlets for entrepreneurship research is
indicated by the range of 16 journals in which the 47 articles were published, and there
are many others that occasionally publish entrepreneurship papers. Diversity enhances
the opportunities for specialized or unusual research to find an outlet, but it also
complicates the process of developing agreed-upon norms of research practice. JBV’s
rising prominence in the entrepreneurship journal market may allow it to act as a core
arbiter of standards for the field.
Because we think JBV plays an important role in providing some boundaries and
coherence to the entrepreneurship research field, we examined its citation patterns during
the period from 1989 to 1994. As a comparison from a similar multi-disciplinary field,
we examined citation patterns for Strategic Management Journal (SMJ) over the same
period. We found that JBV has consistently published research which depends heavily (at
30
least in its citation patterns) on prior entrepreneurship research. Since 1989, the median
annual percentage of citations each year from articles published in JBV to other articles
published in JBV has been 21 percent. The median annual percentage of citations from
JBV to articles from entrepreneurship outlets in general has been 46 percent (based on
individually listed publications in Journal Citation Reports). In contrast, the analogous
percentage for cites from SMJ articles to other SMJ articles was only 10 percent. This
suggests that by some combination of scholarly behavior and editorial policy, JBV is
playing an important role in keeping members of the entrepreneurship research
community attuned to work done by other members of this community. If it is difficult to
publish in JBV without coming to terms with prior entrepreneurship research, and if
scholars remain committed to JBV as an outlet, then some level of coherence and
progress will emerge over time.
Ethnography
Ethnographic or direct field observations allow researchers to uncover the
meaning of patterns in social processes and to detect subtle processes and interpret
interactions which participants may be unable to articulate in interviews. Unlike mailed
surveys or archival records, ethnographic methods allow researchers to pursue intriguing
lines of thought they had not considered, prior to entering the field. Themes in the
anthropological literature mesh very well with the concerns of entrepreneurship
researchers, including the processes of accumulation knowledge and skills and gaining
access to resources. Ethnographic work is extremely time-consuming, however, and
requires a significant “methodological investment” (Stewart, 1991).
31
At the third state of the art conference, only 13 true field studies were discovered:
11 from the Babson conferences and 2 more in the HBR. None formally tested
hypotheses, about half were explicitly longitudinal, and the number of cases observed was
quite small in all but one case. In our search, we found only 6 true field studies on
entrepreneurship topics: 2 from Babson and 4 in JBV. (We found 5 non-
entrepreneurship ethnographic articles in AMJ (3) and ASQ (2) while preparing the data
for our comparative analysis.) Five of the six reported no statistics at all, one was
explicitly longitudinal, one was cross-national, and two were conducted outside of the
United States.
The Spring 1995 issue of ETP – which arrived too late to be included in the data
presented in Table 2 -- was a special issue on field research on firm-level
entrepreneurship, containing 10 papers. Five of the papers were conceptual overviews of
other people’s work, three were based on interviews and surveys, one was based on
interviews and records, and only one was based on actual non-participant observation.
Thus, only one of the papers would have been classified as “ethnographic” in our scheme.
Unlike many of the other interview-based studies we reviewed, the interviews and
surveys in the special issue were mostly in-depth, semi- or unstructured, and conducted
face-to-face with the respondents. Four of the five studies were of established firms,
mostly large ones, and focused on management issues.
Reviewing the empirical papers in the special issue, Guth (1995: 170) argued that
the field methods used have a paradoxical characteristic: “they are useful in making
groundable and creative contributions to corporate entrepreneurship theory, yet produce
results of low inherent credibility and breadth of applicability.” He argued for a normal
32
science approach, in which such studies would be used to develop good theory, which
could then be tested by more rigorous research designs. Stewart et al. (1995) demurred,
arguing that field researchers could archive their data and make it available to others, thus
creating a pool of results that could lead to cumulative generalizations. However, they
also noted the serious impediments to using other people’s field data.
Calls for more field research have been as ubiquitous as the clamor for more
longitudinal studies, and with the same consequences: still more calls for action but
almost no actual research heeding the calls! Despite repeated affirmations of the value of
field work, the number of ethnographic studies found through our search strategy dropped
by about half from the last conference. The lack of ethnographic research cannot be
attributed to ignorance of its value or the methods for carrying it out, as a large and
growing literature exists on such methods. Applications are also visible in other
management and organization fields, such as among organizational culture researchers
(Frost et al, 1991). Perhaps the declining significance of ethnographic studies is a result
of a professional reward system that places a higher value on projects that have a faster
payoff than fieldwork-based projects.
Community-building
The coherence of an academic field depends on the socialization of scholars to its
current research values. If a paradigm exists, it needs to be accepted and reinforced as a
research world-view. We examined two aspects of socialization: training of new cohorts
of scholars, and the consequences of research conferences and leading journals.
From a normal science viewpoint, generations play a key role in the consistency of
scientific progress: consistency as one generation trains the next, and progress as the new
33
cohort learns lessons, established generalizations, and methods which were not available
to their teachers. We therefore identified the 16 researchers whose names appeared most
frequently across our entire database of entrepreneurship articles. We then split the group
into the 8 scholars receiving their terminal degrees prior to 1985, and 8 completing their
graduate work more recently. We then compared the papers written by members of the
two groups (n=86) on all of the characteristics for which we had previously coded each
article. We expected to see generational progress. To our surprise, there were very few
substantial differences between the work of the junior and senior cohorts. The only
difference worth mentioning was that the older cohort was a little more likely to use
relatively sophisticated statistical techniques than was the junior cohort. Our finding
directly contradicted our expectation of inter-generational progress and reinforced our
sense of infrequent normal science victories in entrepreneurship research.
At several points in our analysis, we have cited the importance of the Babson
College Conferences in providing opportunities for professional socialization and
standard setting that may affect which of the three views – normal science, multiple
paradigms, or pragmatism – gains supremacy in the field. Thus, we were curious about
pattern of participation at Babson: do the same people attend year after year, forming a
closed community; do most people attend only once, severing any possible continuity
between years; or, is the picture more complicated?
We compared scholars writing in entrepreneurship journals with those writing
papers for Babson. Over 60 percent of the people whose journal articles we discovered in
our review also wrote papers for Babson during 1990-1994. Of the 448 people whose
names ever appeared on the Babson roster as an author, 323 attended (or co-authored
34
papers) once, 77 attended twice, 30 attended three times, 12 attended four times, and 6
people attended all 5 meetings. Our data are right censored in that the closer we get to
1994, the fewer the opportunities for people who began attending recently to have
attended more than one time. However, for our purposes, we ignore this subtle but
important methodological point. Of the 119 people who attended in 1990, 34 (29
percent) were in attendance again 4 years later, in 1994. Of the 201 people in attendance
in 1994, 93 (46 percent) were “new” in the sense of not having participated in any of the
preceding 4 years.
Babson participants form a loosely-knit community of scholars that contains a
small core of people who attend almost every meeting, thus maintaining continuity across
the generations of attendees. Many people attend one year and skip the next, but the
long-term hold of Babson is apparent in the 29 percent of 1990 attendees who were
present again in 1994, even though most had missed at least two meetings somewhere in
between. Babson may have its greatest impact on the newcomers who show up each year
and participate -- as they must -- in the rituals involving giving a paper and receiving
constructive criticism on it.
We have already noted that the most active Babson participants have turned nearly
half of their papers into published articles or book chapters, and if we extrapolate from
their experiences, a very high proportion of the articles in entrepreneurship journals must
have begun their lives at the Babson meetings. Thus, the norms and standards made
manifest in presentations and discussions at Babson may account, to some extent, for the
patterns we have reported. About 39 percent of the authors whose papers we found in our
keyword search for this research did not attend any of the Babson conferences between
35
1990 and 1994. They comprise a sizable minority whose non-participation may reflect
disagreement with the premises of the Babson conferences -- fostering a sense of shared
community among entrepreneurship scholars -- or perhaps a stronger attachment to other
research communities. Further research is required -- perhaps of an ethnographic nature -
- to discover if there is one voice or many represented at the Babson meetings.
Summary and Conclusions
We have suggested that reasonable people might adopt one of at least three
perspectives - normal science, multiple paradigms, or pragmatic - in considering whether
entrepreneurship research has made much progress over the past several years. The
perspective we adopt both suggests what features are important in defining a coherent
field of research, and focuses our attention on whether progress is being made in the
features which define coherence. In this section, we discuss features which suggest
coherence from each of the three perspectives, and review the evidence in favor of
progress for people adopting each of the three.
Normal Science
Previous reviewers have mainly adopted a normal science view when they
examined gains and shortcomings of entrepreneurship research. Because we wished to
build cumulatively on the past, we have also spoken predominantly in the voice of normal
science. To recapitulate, coherence, from the normal science perspective, consists in
pursuing research that results in the cumulation of findings over time, both within and
across individual scholars' work. Cumulation is generated when a group of scholars
comes to accept as a world-view the discipline of a research paradigm that defines both
36
what are legitimate and interesting questions and what are legitimate methods for
attempting to answer them.
Progress from this perspective has been limited, in comparison both with past
work and with more general, high-quality organizational research. The entrepreneurship
literature has become somewhat more comparative and international, a little more
longitudinal, and less likely to engage in purely speculative or purely descriptive work.
Statistical sophistication has gotten slightly better -- at least at the low end. The work we
reviewed is still dominated by survey research, and response rates have not improved.
Empirical articles in entrepreneurship journals are using bigger samples, often as large as
those found in the top organizations journals. Entrepreneurship articles are more likely
than work in ASQ and AMJ to utilize non-US and cross-national data. However, papers
in ASQ are more likely to test hypotheses explicitly, and to directly replicate prior work.
The top organizations journals are more likely than the top outlets in entrepreneurship
research to publish articles which are statistically sophisticated, use explicit and national
or international sampling frames, to identify homogenous populations, and to provide
dynamic analyses of longitudinal data.
Multiplicity and Diversity
We have argued that there is no well-developed paradigm creating coherence in
entrepreneurship research. Multiple voices, expressing multiple points of view and
methods, contribute to the body of work reaching the outlets we examined. Coherence, in
a field supporting multiple paradigms, suggests the existence of multiple communities of
scholars accepting and committed to research world-views that dictate which questions
are interesting and what methods are legitimate. In a normal science view, multiple
37
paradigms might appear to exist either in a somewhat chaotic “pre-paradigmatic” state of
development, or when a new paradigm is positioned to replace an old, but is waiting out
the retirement of scholars wedded to the old world-view.
Multiple paradigms might be expected in entrepreneurship research, because
scholars might come to the entrepreneurship field with commitments to their home
disciplines, such as strategy, economics, sociology, psychology, and others. The closest
evidence we found of multiple paradigms was reflected in research which we have
labeled "traits" or "not traits." Research focusing on psychological traits of entrepreneurs
has been reviewed and sharply criticized in several prominent articles. The authors of
these critiques have called for an end to traits research, and have tried to cleanly
distinguish traits research from other scholarship. However, as we have seen, "traits "
topics are continuing to generate significant numbers of papers, and several authors have
published rejections of the earlier critiques. If one measure of a paradigm is the topics it
defines as interesting, then we might infer that traits research is generated from a
paradigm which is evolving separately from the paradigm which underlies other
behavioral and "rates" research. We have suggested as well that traits research is more
likely to discuss and evaluate the reliability of its measures, and to utilize established
psychological measures and scales. If paradigms define what methods are legitimate, we
might also infer that a traits paradigm continues to emerge.
However, the evidence for multiple coherent paradigms is weak, at best. Once we
identify the potential for a "traits" paradigm, for example, we are unable to convincingly
label its competition any more suggestively than as "non-traits." Entrepreneurship
research continues to support a diversity (not to say a cacophony) of voices, and to
38
produce a great deal of research which often does not articulate easily with other research
defined in any paradigm.
Within the chorus of multiple voices, nevertheless, there are some signs of
coherence, at the level of the research community. A substantial core of scholars write
papers for Babson year after year, and continue to develop those papers over time. Ideas
introduced at Babson find later outlet as journal articles far more frequently than we had
imagined. Coherence and progress are also evident in the emergence and improvement of
leading entrepreneurship journals. In particular, JBV provides a favored forum for a wide
variety of scholars, including one group which is committed to Babson as an incubator,
and another group which seems to eschew participation at Babson. ETP has taken on an
important role as an outlet for literature reviews and self-reflection by entrepreneurship
researchers.
Pragmatism
Two main expressions of pragmatism among entrepreneurship scholars are
apparent in orientations toward either policy issues practitioner issues. A great deal of the
research we reviewed was energized by the belief -- debated in some circles, but largely
accepted among entrepreneurship researchers -- that entrepreneurial firms generate an
inordinate proportion of job growth in this country, and that they provide an powerful
outlet for innovation (see the special issue of Small Business Economics in which
Harrison (1995) took on his critics). An important group of researchers has tried to focus
on separating what are called truly "entrepreneurial" businesses from "small business,"
"lifestyle businesses" or simple "self-employment" and "income replacement"
motivations to found new firms. In general, we find that a great deal of entrepreneurship
39
research shares an underlying concern with the macroeconomic effects of entrepreneurial
activities. This is often implicit, but is sometimes quite blunt as when Low and
MacMillan (1988) suggested that entrepreneurship research ought to focus on explaining
and facilitating the role of new enterprise in furthering economic progress.
In normal science, a paradigm helps to structure research by defining what
questions are interesting. The implicit orientation to the macro effects of entrepreneurial
activities similarly serves to define what is and what is not interesting, and may thereby
help to lend some coherence. A number of questions one might imagine as important
research topics are not of much interest in the major entrepreneurship research outlets; for
example: entrepreneurship as a temporary career adaptation, non-profit entrepreneurship,
entrepreneurship and American ideals of freedom versus bureaucracy, entrepreneurship as
coping with restricted opportunities, the relationship between entrepreneurship and social
mobility, the study of nascent entrepreneurs and organizations, and business disbandings.
By defining such topics as uninteresting, the loose macro-economic impact orientation
helps to support greater focus and coherence around the topics which are seen as
important.
The second form of pragmatism is an orientation toward practitioner concerns.
Many entrepreneurship research articles conclude with commentary on how the results of
the research might inform entrepreneurial practice. The proliferation of teaching case
studies, particularly in ETP, along with more papers on entrepreneurship education, is an
indication that academic researchers are making progress in trying to be relevant to
practitioners. Readers of journals such as INC. and Nation's Business will find that some
articles written for practitioners are indeed informed by academic research. But the vast
40
majority of "how to" articles, and of topics assumed interesting to practicing
entrepreneurs bear little relationship to the more common topics of academic
entrepreneurship research. Some researchers have examined the lack of fit between topics
of academic concern and the expressed needs and problems of practicing entrepreneurs
(Banks and Taylor, 1991). There is little documented evidence that practicing
entrepreneurs find useful -- or are even aware of -- much of the research which claims to
hold strong implications for practice.
We find it ironic, but not surprising, that a great deal of entrepreneurship research
is of little short term value to practitioners. In particular, the macro focus which
energizes so much recent research may work against research of value to individual
entrepreneurs. Few entrepreneurs can afford the luxury of ruminating on the powerful role
of entrepreneurship in the greater economy. In contrast to academic researchers, most
entrepreneurs need to be closely oriented to the relatively micro concerns of their own
businesses.
The development of good theoretical frameworks and explanations, which is
difficult enough as it is, might not survive the additional burden of needing to be
continually topically useful to practitioners buffeted by the waves of fashion and short
term shocks of the economy. Good theory probably requires the freedom of researchers
to focus on questions which are interesting mainly in theoretical terms. The ability to
generalize findings over time and extend explanations to new topics requires the
development of good theory, as well as some degree of separation from insistent demands
for constant practical relevance.
41
Conclusions: A Managed or Evolutionary Process?
We have argued that from each of three perspectives: normal science, multiple
paradigms, and pragmatism, a case can be made for researchers' contributions to the
coherence of entrepreneurship research and progress over time. It seems clear to us that
attention to scientific standards, an openness to competing interests and approaches, and a
concern with practical relevance all continue to contribute to entrepreneurship research.
But we conclude that, from all three perspectives, progress has been quite limited.
Judging from normal science standards, entrepreneurship research is still in a very early
stage. If there is no single powerful paradigm, then there is even less evidence for
multiple coherent points of view. Finally, entrepreneurship research is of questionable
current topical concern and value to most practicing entrepreneurs.
Taking one step back, the state of the art appears to be a multiplicity of
perspectives, little structured by multiple paradigms, but finding some progress through a
variety of relatively uncoordinated sources. We reviewed some limited methodological
progress from a normal science perspective. We argued that ETP and especially JBV
have helped to bring coherence, as have Babson and the committed community of
scholars who participate in its forums. An orientation to macro issues has limited the
potentially boundless diversity of research topics. The academic community has taken
several concrete steps toward reaching out to practicing entrepreneurs and those who
would join the entrepreneurial ranks.
Despite some marks of progress, the unstructured multiplicity of perspectives and
approaches has left a number of thoughtful commentators frustrated. Part of the "methods
talk" -- the increased number of articles reflecting on entrepreneurship research methods -
42
- is plainly oriented to finding better ways to deal with this multiplicity. Two competing
arguments have emerged. One group of commentators seeks coherence through the
friendly imposition of core definitions (Bygrave, 1989) whereas the other group argues
that we should worry less about definitions and focus more on accurate descriptions of
the populations we study (Katz et al., 1993; Vanderwerf and Brush, 1989). Motivating
both groups is a belief that following either set of prescriptions will help build a common
ground of coherent research in which progress will become easier and quicker. Common
definitions make it harder to talk past one another or to define away inconsistencies.
Common sets of data descriptions make findings more open to replication, or to
empirically-sound reinterpretation through alternative perspectives.
Neither group of commentators has had much of an influence on research practice.
For example, some researchers are still prone to invent their own definitions of key
concepts, and sometimes even to argue that their definitions (even of basic terms like
"entrepreneur") are superior to those offered and used by all previous researchers. Our
evidence also suggests that there has been limited progress toward improving the
description of the samples and populations studied. The sampling frames investigators
have chosen are often obscure, sampling methods are ad hoc, and there has been no
significant improvement in the identification of homogenous populations. Further,
entrepreneurship scholars have continued, relative to other organizational analysts, to
make little use of public data sets, which are by definition available for replication and re-
analysis.
Admonitions to change practice in entrepreneurship research seem to have little
effect, regardless of the status of their source. The field lacks the sort of overall
43
coherence which might allow a limited group of leading figures to exert strong intentional
direction on the field. Rather, what coherence exists depends on a mixture of legitimate
perspectives. Progress comes in scattered bits from a variety of sources, and is much
closer to being an evolutionary process than one which is managed, or manageable by
leading scholars.
Brush and VanderWerf (1989), in a review of the emergence of other successful
scientific fields, noted that progress tends to be achieved without consensus on
definitions. Progress has tended to come after groups of researchers converge on study of
simple, well defined “entities” or populations. But for us, their most important point is
that convergence on what is studied comes because researchers are attracted by the initial
progress made by early investigators. The convergence is spontaneous rather than forced.
What lesson do we learn from history? Influence comes from exemplary research,
not from the propagation of rules or admonitions. The field will be shaped by those who
produce research that interests and attracts others to build on their work. This may also
provide an antidote to the frustratingly repetitive and ineffective calls we hear to do
"more ethnographic work" and more "longitudinal studies." Those who believe they
know the path forward need to do such work themselves, and provide exemplars which
attract others to follow.
44
References
Aldrich, Howard E. and Gabriele Wiedenmayer. 1993. “From Traits to Rates: An
Ecological Perspective on Organizational Foundings.” Pp. 145-195 in Jerome
Katz and Robert H. Brockhaus (eds.), Advances in Entrepreneurship, Firm
Emergence, and Growth, Vol. 1. Greenwich, CT: JAI Press.
Aldrich, Howard E., Sally W. Fowler, Nina Liou, and Sara J. Marsh. 1994. "Other
People's Concepts: “Why and How We Sustain Historical Continuity in Our
Field." Organization, 1, 1: 65-80. London: Sage.
Banks, McRae C. and Stephen Taylor. 1991. “Developing an Entrepreneur and Small
Business Owner-Defined Research Agenda.” Journal of Small Business
Management, 29, 2 (April): 10-18.
Birley, Sue and Ian C. MacMillan (eds.). 1992. International Perspectives on
Entrepreneurship Research. Amsterdam, The Netherlands: North-Holland
(Elsevier).
Birley, Sue and Ian C. MacMillan (eds.). 1995. International Entrepreneurship. London:
Routledge.
Bygrave, William D. Bygrave. 1989. “The Entrepreneurship Paradigm (I): A
Philosophical Look at Its Research Methodologies.” Entrepreneurship: Theory &
Practice, 14, 1 (Fall): 7-26.
Carsrud, Alan L., Kenneth W. Olm, and G.E. Eddy. 1986. “Entrepreneurship: Research
in Quest of a Paradigm.” Pp. 367-378 in Donald L. Sexton and Ray Smilor, eds.,
The Art and Science of Entrepreneurship. Cambridge, MA: Ballinger.
45
Churchill, Neil C. and Virginia L. Lewis. 1986. “Entrepreneurship Research: Directions
and Methods.” Pp. 333-365 in Donald L. Sexton and Ray Smilor, eds., The Art
and Science of Entrepreneurship. Cambridge, MA: Ballinger.
Frost, Peter J., Larry F. Moore, Meryl Reis Louis, Craig C. Lundberg, and Joanne Martin.
Eds. 1991. Reframing Organizational Culture. Newbury Park, CA: Sage.
Gartner, William B. 1988. "`Who Is an Entrepreneur?' Is the Wrong Question."
American Journal of Small Business, 12, 4 (Spring): 11-32.
Gartner, William B. 1989. “Some Suggestions for Research on Entrepreneurial Traits and
Characteristics.” Entrepreneurship: Theory & Practice, 14, 1 (Fall): 27-38.
Greve, Henrich R. 1995. “Jumping Ship: The Diffusion of Strategy Abandonment.”
Administrative Science Quarterly, 40, 3 (September): 444-473.
Guth, William D. 1995. “Theory from Field Research on Firm-Level Entrepreneurship:
A Normal Science Overview.” Entrepreneurship: Theory and Practice, 19, 3
(Spring): 169-173.
Harrison, Bennett. 1995. “Symposium on Harrison’s ‘Lean and Mean’: What Are the
Questions?” Small Business Economics, 7, 5 (October): 25-31.
Hofer, Charles W. and William D. Bygrave. 1992. “Researching Entrepreneurship.”
Entrepreneurship: Theory & Practice, 16, 3 (Spring): 91-100.
Katz, Jerome A., Robert H. Brockhaus, and Gerald E. Hills. 1993. “Demographic
Variables in Entrepreneurship Research.” Pp. 197-236 in Advances in
Entrepreneurship, Firm Emergence, and Growth. Greenwich, CT: JAI.
46
Katz, Jerome A. 1992. “Secondary Analysis in Entrepreneurship: An Introduction to
Databases and Data Management.” Journal of Small Business Management, 30, 2
(April): 74-86.
Low, Murray, and Ian C. MacMillan. 1988. “Entrepreneurship: Past Research and
Future Challenges.” Journal of Management, 14, 2: 139-161.
McKelvey, Bill and Howard E. Aldrich. 1983. “Applied Population Science.”
Administrative Science Quarterly, 28, 1 (March): 101-128.
MacMillan, Ian C., and Jerome A. Katz. 1992. “Idiosyncratic Milieus of Entrepreneurial
Research: The Need for Comprehensive Theories.” Journal of Business
Venturing, 7, 1 (January): 1-8.
Mone, Mark A. And William McKinley. 1993. “The Uniqueness Value and Its
Consequences for Organization Studies.” Journal of Management Inquiry, 2, 3
(September): 284-296.
Paulin, W.L., R.E. Coffey, and M.E. Spaulding. 1982. Pp. 353-373 in C.A. Kent,
Donald L. Sexton, and Karl H. Vesper, eds., Encyclopedia of Entrepreneurship.
Englewood Cliffs, NJ: Prentice-Hall.
Perryman, M.R. 1982. “Commentary on Research in the Field of Entrepreneurship.”
Pp. 377-378 in C.A. Kent, Donald L. Sexton, and Karl H.Vesper, eds.,
Encyclopedia of Entrepreneurship. Englewood Cliffs, NJ: Prentice-Hall.
Peterson, R. and D. Horvath. 1982. “Commentary on Research in the Field of
Entrepreneurship.” Pp. 374-376 in C.A. Kent, Donald L. Sexton, and Karl H.
Vesper, eds., Encyclopedia of Entrepreneurship. Englewood Cliffs, NJ: Prentice-
Hall.
47
Smith, Ken G., Martin J. Gannon, and Harry J. Sapienza. 1989. “Selecting
Methodologies for Entrepreneurial Research: Trade-Offs and Guidelines.”
Entrepreneurship: Theory & Practice, 14, 1 (Fall): 39-50.
Stewart, Alex. 1991. “A Prospectus on the Anthropology of Entrepreneurship.”
Entrepreneurship: Theory & Practice, 16, 2 (Winter): 71-92.
Steward, Alex, Kevin E. Learned, Stanley W. Mandel, and Kristin M. Peterson. 1995.
“Using Field Research on Firm-Level Entrepreneurship: A Coda.”
Entrepreneurship: Theory and Practice, 19, 3 (Spring): 175-184.
VanderWerf, Peter A. and Candida G. Brush. 1989. “Achieving Empirical Progress in
an Undefined Field.” Entrepreneurship: Theory & Practice, 14, 2 (Winter): 45-
58.
Woo, Carolyn Y., Arnold C. Cooper, and William C. Dunkelberg. 1991. “The
Development and Interpretation of Entrepreneurial Typologies.” Journal of
Business Venturing, 6, 2 (March): 93-114.
Wortman, Max S. Jr. 1986. “A Unified Framework, Research Typologies, and Research
Prospectuses for the Interface between Entrepreneurship and Small Business.”
Pp. 272-332 in Donald L. Sexton and Ray Smilor, eds., The Art and Science of
Entrepreneurship. Cambridge, MA: Ballinger.