THE CRYSTALLIZATION OF ENTREPRENEURSHIP RESEARCH DVS AND METHODS IN MAINSTREAM MANAGEMENT JOURNALS
Denis
Grégoire, University of Colorado at Boulder
G. Dale Meyer, University of Colorado at Boulder
Julio O. De Castro, University of Colorado at Boulder
This study uses co-occurrence analysis to investigate the crystallization of entrepreneurship research variables, levels of analysis, data collection methods and analytical techniques in a sample of 104 empirical articles published in six mainstream management journals between 1985 and 2001. Results indicate that entrepreneurship research published in these journals is characterized by a pattern emphasizing performance variables, increased dominance of firm-level analyses across industries, and greater reliance on secondary data. While we rejoice at the sophistication of analytical techniques, we question whether the crystallization observed isn’t cause for concerns, both theoretically and methodologically.
ENTREPRENEURSHIP RESEARCH AS AN OBJECT OF STUDY
Along with entrepreneurship’s quest for legitimacy among the disciplines of management, a number of scholars have considered the field’s development as an object of study in its own right. For instance, some have provided counts of the most published authors (e.g. Shane, 1997), the most frequently cited sources (e.g. Romano and Ratnatunga, 1996) or the most cited works (e.g. Béchard, 1997; Ratnatunga and Romano, 1997). Others have investigated the networks between authors, institutions, publications and anchoring disciplines (e.g. Déry and Toulouse, 1996), or the field’s thematic axes of convergence (e.g. Grégoire, Déry and Béchard, 2001; Reader and Watkins, 2001).
Still, the emergence of a field like entrepreneurship is not solely limited to thematic or conceptual developments. Accordingly, a few scholars have studied the corpus of entrepreneurship literature from the point of view of research designs and methodologies. Following the pioneering works of Paulin, Coffey and Spaulding (1982), Wortman (1987) and Churchill and Lewis (1986), Aldrich (1992) and Aldrich and Baker (1997) focus on describing the methodological practices of entrepreneurship scholars (e.g. data collection, design and scope, hypothesis testing, sample size, statistical techniques, etc.). Eventually, both the 1992 and 1997 papers attempt to answer the question “has there been progress in entrepreneurship research?” But in the end, Aldrich and Baker admit that if “progress” is taken to mean the convergence of a field towards a certain number of well-defined and accepted “practices” (conceptual and methodological), “entrepreneurship is still in a very early stage. (1997: 398).”
More recently, Chandler and Lyon (2001) used a sample of 416 empirical articles published in entrepreneurship and management journals to systematically analyze some of the methodological trends characterizing the field. While they make some observations regarding the relative use of primary and secondary data (respectively present in 66% and 31% of their sampled articles) and note the increasing use of more sophisticated statistical and analytical techniques, the core of their discussion focuses on the issues of reliability and validity. Indeed, the authors show that if there is a slight increase in the number of articles specifically addressing these questions, “much of the work done in the mainstream entrepreneurship literature . . . remains relatively unsophisticated in its treatment of reliability and validity issues (2001: 110).”
Interestingly, all of the studies above analyze each methodological dimension in isolation, as if these dimensions were independent of one another. We believe that, in practice, choosing to focus on a particular level of analysis, to use specific data sources, or to rely on precise analytical techniques are not independent decisions, but proceed hand-in-hand from the overall design of a study. Moreover, these choices are also related to the choice of a particular focus that is, the choice of one or many dependent variables. To our knowledge however, there has been little research investigating which dependent variables entrepreneurship scholars are focusing on, or the configuration of research variables, design, methods, and techniques that characterizes entrepreneurship research.
OBJECT OF STUDY AND METHOD OF RESEARCH
To investigate the crystallization of entrepreneurship research variables and methods, we adapt the frameworks devised by Aldrich (1992), Aldrich and Baker (1997), and Chandler and Lyon (2001) to the techniques of co-citation / co-occurrence analysis. We then analyze the configuration of dependent variables, research design, and research methods characterizing 104 empirical articles published in six mainstream management journals between 1985 and 2001. The basis for our sample comes from a forthcoming survey of the entrepreneurship research made by Busenitz et al. (2003). While these authors use the sample as a springboard to make conceptual propositions regarding how the field could further its legitimacy and theoretical underpinnings, we focus instead on the methodological characteristics of the corpus.
First, we collected all the articles identified by Busenitz et al. (2003) in their survey. Given that their analysis only covers articles published between 1985 and 1999, we included publications made in the years 2000 and 2001 by following the same criteria that they put forth to build their sample, that is:
Following Buesnitz et al.’s procedures, all editor notes, book reviews, and replies to published papers were omitted so that the data would contain only articles and research notes that were non-invited and peer reviewed. However, and since we focus on research designs and methods, we excluded from the sample all articles that had no empirical data. This meant discarding 22 texts originally inventoried by Busenitz et al., including all 10 articles from the Academy of Management Review. We also followed the procedure laid out by Chandler and Lyon (2001) in excluding two articles from Management Science whose observations were based on mathematical modeling. Finally, we made sure to include all the empirical articles published in two special issues, i.e. that on International Entrepreneurship in AMJ (2000) and that on Strategic Entrepreneurship in SMJ (2001). In total, the resulting sample comprises 104 empirical articles. The full list of articles is available upon request to the first author.
In a second step, all the articles were coded along the four dimensions of dependent variables, levels of analysis, data collection methods, and data analysis methods. The last three categorizations are based in large part on initial reviews made by Aldrich (1992), Aldrich and Baker (1997), and Chandler and Lyon (2001). We derived the first categorization of dependent variables by compiling the most commonly used dependent variables in all 104 articles. Table 1 lists the schemes employed to code each article. Allowances were made to account for multiple coding, i.e. the possibility that an article might make use of more than one dependent variable, level, data collection method or analytical technique.
Once all articles in the database were coded, a third step focused on the analysis of the corpus itself. To do so, we use the principles of co-occurrence analysis underpinning many works in the sociology of science. In entrepreneurship proper, Déry and Toulouse (1996) used the technique to offer a detailed account of the social structuration of the field—as evidenced by the Journal of Business Venturing between 1986 and 1993 (Déry & Toulouse, 1996). Grégoire, Déry and Béchard (2001) also used the technique in a co-citation analysis of the long-form articles published in Frontiers of Entrepreneurship Research to reveal the converging axes of research characterizing the field of over the last 20 years.
Two methodological considerations are worth noting before moving to the actual mapping of research designs. First, and in order to present a clearly understandable picture of the sample, we focus mainly on those dependent variables, levels of analysis, data collection methods and analytical techniques that are present in more than 5% of the articles sampled (occurrence frequency threshold). Second, we posit that to represent a significant pattern of research design characterizing the sample of entrepreneurship papers, the relationship between its constituting elements has to be at least 25%. This decision rule means, for example, that a “meaningful” relationship will be said to exist between a dependent variable and an analytical technique if the latter is used in more than 25% of all the articles that focus on that particular dependent variable.
Lastly, we investigate whether the results we observe when analyzing the overall sample have evolved in time. To minimize the effect that special issues might have in “skewing” the results towards certain sources, we compare two unequal periods (1986–1995 vs 1996–2001) that have roughly the same number of articles (50 vs 54) AND a similar distribution of articles among the journals considered.
Table 2 presents some basic characteristics of the 104 papers sampled. As can be seen, papers published in SMJ account for roughly 36% of entrepreneurship articles published in mainstream management journals. AMJ and ASQ come in second and third place, slightly under 17% (but with the order reversed between them for the two periods). OS and MS come in third and fourth place at about 11%, with JOM coming in last position. As can be expected, this distribution mirrors the observations of Busenitz et al. (2003) and Chandler and Lyon (2001).
For reasons of spatial efficiency, Figure 1 presents side-by-side the co-occurrence relationships between dependent variables and level of analysis, level of analysis and data collection methods, and data collection methods and analytical techniques (from left to right respectively). Figure 2a) and Figure 2b) present the co-occurrence relationships between dependent variables and data collection methods, and between dependent variables and analytical techniques. The relative occurrence of each variable, level, method or technique can be read on the vertical axis: the higher an element is on a figure, the more frequent its occurrence. Significant omnibus chi-square tests of the cross-tables underlying our analyses give credence to our proposition that the choices of variables, levels, methods and techniques are not independent of one another.
Dependent Variables and Level of Analysis
As can be seen on Figure 1, performance is the most studied dependent variable in our sample. At 31.7% however, this variable is by no means as dominating as one might hypothesize: manifestly, other variables also have been used to investigate entrepreneurship, including growth (14.4%), strategic change (13.5%), strategy content (12.5%), the start-up decision (8.7%) and survival (5.8%). In addition, 25.0% of the articles inventoried focus on other dependent variables (e.g. power, image, reputation, managerial succession, network characteristics, ownership, perception, invention, new product development, speed to market, wealth creation, knowledge acquisition, etc.).
One thing remains however. As can be seen by the many links that all variables share with the level of firms across industries, and to a lesser degree with the level of firms within an industry, a majority of entrepreneurship studies published in mainstream management journals focus on the manifestations of entrepreneurship at the firm level—and this is apparently true regardless of the dependent variable of interest. By contrast, none of the principal dependent variables seem to be associated with other levels of analysis. Moreover, there are relatively few entrepreneurship investigations conducted at levels other than the firm, at least among those studies appearing in mainstream management journals. For instance, studies conducted at the level of individuals across industries or at “other” levels (e.g. state, business group, joint ventures) appear in only 5.8% of the corpus, while studies of one firm only have a 4.8% frequency.
Level of Analysis and Data Collection Methods
As can be seen in the middle of Figure 1, a majority of entrepreneurship research published in mainstream management journals is conducted through the analysis of archival data (52.9%). Indeed, this seems particularly true for studies done at the firm level (whether within or between industries), as well as studies conducted at “other” levels. As indicated by the solid lines, 50% or more of the studies conducted at either of these levels are making use of archival data. Mailed surveys have also been used extensively (40.4%), albeit they associate primarily with studies anchored at the firm level (between or within industries), and not with other levels. Interestingly, studies done at the level of individuals across industries exhibit a different pattern, as they make little use of archival data, but rely instead on both surveys and interviews—arguably more “personal” methods. Lastly, it is worth observing that while studies focusing on one firm associate primarily with case study methods, 40% of these studies also make use of archival records to substantiate some of their observations. That being said, it remains that the dominant pattern is to conduct research at the firm level through archival data.
Data Collection Methods and Analytical Techniques
In line with previous research, we note that at a frequency of 32.7%, regression is the most commonly used analytical technique in our sample, and that is true regardless of the data collection methods being used. Longitudinal studies are also present, as is testified by the number of studies making use of various techniques associated with econometrics and time-series regression (21.2%). One noteworthy pattern is that while more personal data collection methods, such as interviews and surveys, are associated with regression techniques (and analysis of variance), the use of archival sources correlates with both regression and time-series methods. This suggests that longitudinal investigations have mainly been anchored by archival data collection methods, and not by interviews and surveys.
Dependent Variables and Data Collection Methods
Figure 2a) shows the patterns of co-occurrence between dependent variables and data collection methods. Analysis of the pattern of links points to three main observations. First, each dependent variable seems to have been studied through data obtained from archival records. Indeed, the pattern seems especially strong for variables like performance, the creation decision, and survival (all showing relationships above the 0.50 level). Second, most dependent variables have also been studied through mailed surveys, with the exception of the creation decision and survival. Interestingly, the latter variable also appears to have been studied through interviews—in about a third of the articles surveyed. As was seen before however, archival data remain the data collection method of predilection.
Dependent Variables and Analytical Techniques
Figure 2b) shows the patterns of co-occurrence between dependent variables and analytical techniques. As we have seen above, regression is the dominant technique, and it appears to have been used in studies focusing on performance, growth, strategic change, strategic content and “other” variables. Interestingly, the use of time-series regression techniques denoting longitudinal concerns is primarily associated with the dependent variables of start-up decision, survival, and performance—suggesting that it is those variables that have been studied in a longitudinal manner. Indeed, closer analyses of the raw data reveal that other variables have rarely been studied with longitudinal techniques.
Evolution in Time: 1986–1995 vs 1996–2001
By and large, analyses of the co-occurrence relationships between research variables and designs showed very little differences between the two periods under consideration, at least as far as the overall patterns are concerned. In fact, most of the differences between the periods occur within each dimension, at the level of the relative frequency of particular variables, levels, methods or techniques. In other words, it is not the patterns of association that change between the periods as much as the relative occurrence of particular elements. Accordingly, we present only the evolution of individual elements within each dimension—and not the co-occurrence patterns.
As can be seen on Figure 3, entrepreneurship studies published in mainstream management journals are increasingly focusing on performance. However, the study of the creation decision and variables in the “other” category is also on the upswing. By contrast, variables like growth, strategic change, strategic content, personal characteristics and individual decisions other than creation all appear to generate decreasing interest.
Apart from showing the dominance of firm-level studies, the interesting pattern displayed on Figure 4 concerns the sharp decrease of studies taking place at the levels of firms within an industry, and the corresponding increase of studies taking place at the level of firms across industries. One also notes the virtual disappearance of studies taking place at the level of individuals within a single firm, or of case studies of individual firms.
While Figure 5 shows a decreasing reliance on surveys and case studies, one notes relatively sharp increases in the use of archival methods. More specifically, this increase is manifest in both the analysis of public or private databases, as well as in those cases when a database is constructed from public sources. The figure also shows the marginal use of methods like ethnography and experimentation—an observation made by Aldrich as early as in 1992.
Finally, Figure 6 demonstrates that while regression continues to be the dominant analytical technique, the use of time-series regression and “other” techniques (such as Structural Equation Modeling) is increasing. By contrast, there appears to be decreasing use of descriptive statistics, categorical data analysis, analysis of variance, as well as exploratory and confirmatory data reduction techniques.
What characterizes the empirical research on entrepreneurship appearing in mainstream management journals? Four main elements emerge from our analysis.
Overall, the analysis of those patterns and their evolution in time suggests two paradoxical observations. On the one hand, there is evidence that the field is converging on some identifiable practices, that is, the study of some objective variables, anchored at specific levels of analysis, studied through reliable and valid data and analyzed with increasingly sophisticated techniques. On the other hand, it is not sure whether such crystallization is a sign of progress. Indeed, a number of issues can be raised as entrepreneurship research appears to be moving towards:
Manifestly, the evolution of entrepreneurship research evidenced in our analysis raises some important theoretical and methodological issues. But what accounts for such evolution? One could speculate that in the absence of a strong paradigm within the emerging field itself, entrepreneurship scholars have adopted the canons of other research areas. Concretely, such legitimization dynamics could be especially strong as entrepreneurship research struggles to gain its lettres de noblesse in respected management journals. Indeed, we note that by contrast to Chandler and Lyon (2001) who had included ETP and JBV in their sample, the use of secondary data is much higher in our sample (51.9% as opposed to their 31.0%). Such difference suggests that the crystallization of research variables and methods is somewhat different in mainstream management journals than it may be in predominantly entrepreneurship journals.
At the individual level, it is possible that the use of archival data and the focus on firm-level phenomenon appear as a logical solution to “the academic reward system in place at many universities and the corresponding pressures for publication”—an argument used by Chandler and Lyon with respect to the un-frequent use of longitudinal designs (2001: 112). Indeed, building a database from which to publish a series of papers seems to be the strategy of choice for many Ph.D. students and tenured-track Faculty.
Regardless of the specific dynamics that preside over it, the observed crystallization of research variables and methods calls for entrepreneurship scholars to question whether it is a sign that the field is making true advances, both theoretically and methodologically. On one hand, the increased sophistication of analytical methods long desired by Wortman (1987) and evidenced by Chandler and Lyon (2001) is one positive element, even as we second Chandler and Lyon’s call for increased attention to reliability and validity. More importantly, we believe that the increasing interest generated by variables like the decision to create a new venture and a series of other variables (e.g. new product development, speed to market, wealth creation, knowledge acquisition, to name but a few) is cause for hope with respect to the distinctive contribution that entrepreneurship can make.
But in the quest for legitimacy among the established disciplines of management, the increasing focus on performance, the anchoring of entrepreneurship as a firm-level construct and the reliance on archival data is cause for concern, just as is the dearth of research using methods like ethnography, simulation and experimentation and the virtual absence of investigations at levels other than the firm. As many scholars continue to lay the theoretical foundations of entrepreneurship’s distinctive contribution to the management sciences, shouldn’t we also question whether we are meeting all the methodological challenges that these conceptual advances may need?
CONTACT: Denis Grégoire, Leeds School of Business, University of Colorado, Campus Box 419, Boulder, CO 80309-0419; denis.gregoire@colorado.edu
Aldrich, H. E. (1992). Methods in our madness? Trends in entrepreneurship research. In D. A. Sexton & J. D. Kasarda (Eds.), The State of the Art of Entrepreneurship (pp. 191–213). Boston, MA: PWS-Kent Publishing Company.
Aldrich, H. E., & Baker, T. (1997). Blinded by the cites? Has there been progress in entrepreneurship research? In D. L. Sexton & R. W. Smilor (Eds.), Entrepreneurship 2000 (pp. 377–401). Chicago, Ill: Upstart.
Béchard, J.-P. (1997). Understanding the field of entrepreneurship: a synthesis of the most often quoted contributions. Paper presented at the Frontiers of Entrepreneurship Research, Babson College.
Busenitz, L., West, G. P. I., Shepherd, D. A., Nelson, T., Chandler, G. N., & Zacharakis, A. (2003). Entrepreneurship in emergence: fourteen years of entrepreneurship research in management journals. Journal of Management.
Chandler, G. N., & Lyon, D. W. (2001). Issues of research design and construct measurement in entrepreneurship research: the past decade. Entrepreneurship Theory and Practice, 25(4), 101–113.
Churchill, N. C. (1992). Research issues in entrepreneurship. In D. L. Sexton & J. D. Kasarda (Eds.), The State of the Art of Entrepreneurship. Boston, MA: PWS-Kent Publishing Company.
Churchill, N. C., & Lewis, V. L. (1986). Entrepreneurship research: Directions and methods. In D. A. Sexton & R. W. Smilor (Eds.), The art and science of entrepreneurship (pp. 333–365). Cambridge, MA: Ballinger.
Déry, R., & Toulouse, J.-M. (1996). Social structuration of the field of entrepreneurship: a case study. Canadian Journal of Administrative Sciences, 13(4), 285–305.
Grégoire, D., Déry, R., & Béchard, J.-P. (2001). Evolving Conversations: A Look at the Convergence in Entrepreneurship Research. Paper presented at the Babson Kaufman Entrepreneurship Research Conference, Jönköping, Sweden.
Paulin, W. L., Coffey, R. E., & Spaulding, M. E. (1982). Entrepreneurship research: methods and directions. In C. A. Kent & D. L. Sexton & K. H. Vesper (Eds.), The encyclopedia of entrepreneurship (pp. 353–374). Englewood Cliffs, NJ: Prentice-Hall.
Ratnatunga, J., & Romano, C. (1997). A “citation classics” analysis of articles in contemporary small enterprise research. Journal of Business Venturing, 12, 197–212.
Reader, D., & Watkins, D. (2001, November 22–23, 2001). The intellectual structure of entrepreneurship: an author co-citation analysis. Paper presented at the RENT XV, Turku, Finland.
Romano, C., & Ratnatunga, J. (1996). A citation analysis of the impact of journals on contemporary small enterprise research. Entrepreneurship Theory and Practice, 20(3), 7–22.
Shane, S. A. (1997). Who is publishing the entrepreneurship research? Journal of Management, 23(1), 83–95.
Shane, S. A., & Venkataraman, S. (2000). The promise of entrepreneurship as a field of research. Academy of Management Review, 25(1), 217–226.
Wortman, M. S. (1987). Entrepreneurship: an integrating typology and evaluation of the empirical research in the field. Journal of Management, 13(2), 259–279.
2002 Babson College. All Rights Reserved. Last Updated March 2003.