Abstract
This article develops a co-citation analysis covering the period 1991–2014 outlining the collective logic of authors conducting studies of entrepreneurship education. The core themes, major contributions and topological features characterising their relationships are identified. The findings indicate that the field has a polycentric structure with five core themes of which entrepreneurial intentions emerges as the most influential while the entrepreneurial learning and evaluation themes have emerged in recent years. These two core themes appear to foster changes within the current structure with the introduction of new theoretical inputs and the formation of new theoretical hubs. We argue that future research should attempt to organise focal themes into a theoretical framework that allows a comprehensive picture of entrepreneurship education.
Keywords
Introduction
Entrepreneurship education has attracted considerable attention in recent years with numerous special issues of academic journals devoted to this subject (Matlay, 2006, 2010, 2011, 2012) and new calls issued for future work (Fayolle et al., 2014; Pittaway and Penaluna, 2013). Extant literature reviews have enhanced the understanding of the major themes addressed by scholars and have identified important criticisms. Pittaway and Cope (2007a) recognised four core topics: first, the general policy climate for entrepreneurship education; second, the university context; third, programmes and their connection with the entrepreneurship propensities of students and fourth, entrepreneurship education outputs. They note a scarcity of empirical works in some areas and a fragmentation among contributions which are generally developed without drawing theoretical support from adult and management learning, higher education and labour market policy.
More recently, Byrne et al. (2014) examined the literature using five core themes as lenses; this analysis covered studies focused on first, the current state of the topic; second, the targets and needs of entrepreneurship education; third, evaluations; fourth, entrepreneurial learning and fifth, teaching methods. This review, together with other sources (e.g. Fayolle and Liñán, 2014; Nabi et al., in preparation), has confirmed that the topic of entrepreneurship education is fragmented, deficient in terms of theoretically well-grounded research and is lacking a critical approach (Fayolle, 2013). With respect to these issues, Fayolle (2013) has encouraged scholars to question ‘the main research streams, theories, methods, epistemology, assumptions and beliefs dominating the field, and the educational practices of entrepreneurship education’ (p. 700).
This article identifies the intellectual structure of the field by studying the behaviour of the citing scholars recorded in the ISI Web of Science. The following questions are addressed: Which topics have citing scholars targeted the most? Which links have been conceptually designed among these topics? What might be the central contributions and streams of thought in the future?
We adopted a bibliometric lens to achieve this goal and used co-citations to highlight the most frequent connections between pairs of articles (Cobo et al., 2011a, 2011b; Culnan, 1986, 1987; Di Guardo et al., 2012; Garfield, 1972, 1979; McCain, 1990; Ramos-Rodríguez and Ruíz-Navarro, 2004; Rowlands, 1999; Small, 1973, 1977; Small and Griffith, 1974; White and Griffith, 1981). These connections reveal the prevailing logic that scholars followed to organise the produced knowledge and provide inferences about the mental models that guide analyses of the topic. The investigation of the topological features of these mental models allows us to represent the collective behaviour of researchers interested in entrepreneurship education and to draw a scientific map of the field (Acedo and Casillas, 2005; Acedo et al., 2006; Di Guardo and Harrigan, 2012; Di Stefano et al., 2012; Small, 2003, 2010).
In undertaking this study, we followed methods adopted by previous research (Busenitz et al., 2014; Meyer et al., 2014) and used the ISI Web of Science database as a bibliometric source to retrieve publications and citing scholars. The publications constituting our unit of analysis were selected by following the most recent methodological indications, and various multivariate analyses were performed to design the scientific map (Di Stefano et al., 2012). In particular, a factor analysis was performed to verify the existence of different factors constituting the theoretical foundations of entrepreneurship education (Di Guardo and Harrigan, 2012; Nerur et al., 2008; Rowlands, 1999). Multidimensional scaling (MDS) enabled us to arrange the articles on a Cartesian plane with relative distances/proximities depending on the co-citation frequency for each pair (Di Stefano et al., 2012; White and McCain, 1998; Wilkinson, 2002). Finally, a network analysis was implemented to show the relationships between the articles. This analysis provides a picture of the knowledge structure of the topic and an indication of the high-impact contributions of each core theme (Chen and Carr, 1999; Quirin et al., 2008; White, 2003).
The article is structured as follows: we commence by elucidating the methodological details for the creation of our unit of analysis. The results section then summarises findings drawn from bibliometrics, MDS, and factor and network analyses. Finally, the main findings are discussed along with the limitations and theoretical implications.
Methodology
Overview of the method
Various citation-based approaches to science mapping are available: direct citation, bibliographic coupling and co-citation analysis (Boyack and Klavans, 2010). Direct citation, deemed to be the simplest kind of publication citation, corresponds with a direct link between two articles (Huang and Chang, 2011). Bibliographic coupling occurs when two articles have at least one reference in common in their bibliographies (Kessler, 1963). Co-citation occurs between articles ‘A’ and ‘B’ when an article ‘C’ cites both articles ‘A’ and ‘B’ in its references (Crane, 1972; Small, 1973).
In bibliometrics, many studies have been published which aim to understand the accuracy of these techniques in mapping a scientific field (Boyack and Klavans, 2010; Huang and Chang, 2011; Persson, 2010). Despite the debate around the advantages characterising each technique is ongoing, direct citation appears to be the least accurate approach compared with bibliographic coupling and co-citations (Boyack and Klavans, 2010). Given our objective, we focused our attention on co-citations in light of recent works (Zhao and Strotmann, 2008a, 2008b, 2014) arguing that co-citations are a useful means of unfolding the theoretical core of a topic; bibliographic coupling may be adequate for investigating recent trends.
With co-citations, the associations between articles can be represented by a frequency matrix of co-occurrences in which a unit of analysis consisting of n articles is paired by citing articles (Small, 1973). The more often two articles are cited together, the stronger the relationship between them; consequently, the greater the number of co-citations between articles, the more likely they are affiliated with the same school of thought, thereby revealing an ‘invisible college’ among the associated articles or topics (Braam et al., 1991; Small, 1977; Small and Griffith, 1974; White and McCain, 1998). For this reason, co-citation analysis is an advanced citation metric able to map a specific domain’s collective cognitive patterns over a particular period of time (Crane, 1972; Nerur et al., 2008; Small, 1973, 2003, 2010; Smith, 1981).
Two focal points coexist in co-citations: one representing articles that are ‘cited’, namely, the unit of analysis, and the other one representing the ‘citing’, which refers to those that are citing the aforementioned unit of analysis. Through the cited, we are able to outline the theoretical core of a discipline. Through the citing, we are able to investigate the connections drawn among the articles and to explore the knowledge structure of a topic.
Crane (1972) and Small (1973) first introduced co-citations as a bibliometric technique with the aim of developing a new and complementary ‘way of modelling the intellectual structure of scientific specialities’ (Small, 1973: 28), with respect to the traditional methods for reviewing literature. This technique simultaneously allows researchers to explore how a field is articulated theoretically and to infer the mutual relationships among contributions and core themes by revealing distances and proximities, as seen in a scholarly community (Cobo et al., 2011a, 2011b; White and McCain, 1998). These quantitative approaches are continuously improved methodologically, given their increasing application in a wide range of disciplines and the increasing production of scientific knowledge (Zhao and Strotmann, 2014). In the next section we present the methodological details we followed to retrieve the unit of analysis and the citation list.
Data and selection of the unit of analysis
Following the McCain prescriptions (Di Guardo and Galvagno, 2010; McCain, 1990; Nerur et al., 2008), our analysis consists of eight steps: selecting the units of analysis using keywords, selecting the citing contributions, filtering the unit of analysis according to the citation count, retrieving co-citation frequencies, compiling the raw co-citation matrix, converting the raw co-citation matrix into a correlation matrix, conducting multivariate analysis and interpreting the findings.
Regarding the selection of the unit of analysis, any study based on co-citations must identify sources that represent the core of a discipline over a specified time period (Callon et al., 1993). In particular, the unit of analysis can be defined in terms of articles, authors, institutions or words (Cobo et al., 2012), depending on the aims of the analysis (Culnan, 1986; White and Griffith, 1981). We adopted articles to identify the connections among the most influential contributions in the entrepreneurship education literature because we were interested in exploring the intellectual base of the topic rather than the social structure, which relies on authors or institutions (Acedo et al., 2006; Cobo et al., 2011b; Glänzel, 2001).
The co-citation analysis covers the 1991–2014 period. From a methodological stance, this timeframe is suitable for selecting a unit of analysis that balances recent and more dated publications (Chen, 2006; Pilkington and Meredith, 2009; Small, 1973; White and McCain, 1998). A longer time period might overestimate the presence of older articles that are more likely to have recorded higher citations. In contrast, a shorter and more recent time period prevents the unit of analysis from having sufficient citations to perform a co-citation analysis (Rowlands, 1999; White and Griffith, 1981). Furthermore, previous bibliometric studies on entrepreneurship identified a time span of 15–25 years as appropriate for investigating the field (Busenitz et al., 2014; Di Guardo and Harrigan, 2012; Grégoire et al., 2006; Meyer et al., 2014; Nerur et al., 2008; Ramos-Rodríguez and Ruíz-Navarro, 2004). In particular, Meyer et al. (2014), who considered the 1991–2009 period, have shown that entrepreneurship grew steadily during the 1990s and was only validated as an academic discipline in the 2000s. The chosen time span thus allows us to juxtapose the observed publication history of entrepreneurship education with that of previous entrepreneurship studies.
We used the ISI Web of Science as a database for retrieving our data. This choice is favoured in previous co-citations in entrepreneurship with Busenitz et al. (2014), Meyer et al. (2014), Schildt et al. (2006), Cornelius et al. (2006) and Reader and Watkins (2006) using this database as a data source. This selection provided the opportunity to adopt a well-consolidated procedure, thereby allowing us to ensure methodological rigour and to juxtapose our results with those from previous entrepreneurship studies. It should be stressed that every database has its peculiarities, which may influence the resulting scientific map, so results should be interpreted within their specific context (Glänzel, 2012; Meyer et al., 2014) and bibliometric studies should generally be considered complementary to each other (Levine-Clark and Gil, 2009). Furthermore, data were collected from all categories of sciences (SCI), social sciences (SSCI) and arts and humanities (A&HCI) citation indices and all publication types (articles, conference papers, reviews and other materials) to obtain an interdisciplinary and multi-source perspective.
Using the keyword entrepreneur* education*, a preliminary list of 597 entries was generated. A subsequent content analysis of the 100 most-cited articles (Culnan, 1987) produced an additional list of 20 keywords (Table 1), which were used to identify a more comprehensive list of entries, composed of 1956 sources. As shown in Table 1, the keywords capturing the majority of publications (73%) are business education (N=712), entrepreneur* education (N=597), student* entrepreneur* (N=140) and entrepreneur* learning (N=115).
Keywords.
We retrieved entrepreneurship education articles using truncated strings, obtained with the wildcard asterisk (*), which represents any group of characters, including no characters, according to the searching rules of ISI Web. This means that the string entrepreneur* includes the keywords entrepreneur, entrepreneurs, entrepreneurial and entrepreneurship. We used the field tag ‘TS’, which stands for topics, with the aim of searching for keywords in the following field of a record: title, abstract, author keywords, database keywords (keywords Plus).
By shifting the attention to the filtering procedure, we adopted an ad hoc selection criterion to include more recent work and to balance our interest in defining the core of the discipline (Acedo et al., 2006; Béchard and Grégoire, 2005; Small, 1977; Stern, 2014). Specifically, for articles issued from 1991 to 2009, 30 citations need to be included in the filtered unit of analysis. For work published after 2009, the criteria stipulated at least 10 citations, and 2 citations on average per year, which correspond to the median of citations for the 2009–2014 period. Overall, 79 articles met these criteria.
To standardise data and avoid possible scale effects, we converted the raw co-citation matrix into a correlation matrix and used the Statistical Package for the Social Sciences (version 20) to calculate Pearson’s correlation coefficient for each cell in the matrix (Rowlands, 1999). To reduce the probability of including an extraneous school of thought, we eliminated all articles having more than 66% of zeros or those with a co-citation score lower than 40 per row (Acedo et al., 2006; Rowlands, 1999). All these filters produced a final list of 44 articles (Table 2), which was used to perform the multivariate analyses that are explained below (Culnan, 1986; Di Stefano et al., 2012; Nerur et al., 2008).
The set of 44 top co-cited articles.
Multivariate analysis
First, non-metric MDS was employed to display the intellectual distance between scientific contributions (White and McCain, 1998) and to generate a map using the similarities between objects (Nerur et al., 2008; Wilkinson, 2002). By computing the Euclidean distance between each pair of papers constituting our unit of analysis, the map reveals each article’s position and shows the areas in which few, or no studies, have been performed. This mapping exercise enables scholars to spot gaps in the literature that emerge by linking schools of thought that have not been connected previously (Di Stefano et al., 2012; White and McCain, 1998; Wilkinson, 2002).
We then employed correspondence factor analysis, the relevance of which in this context is based on the notion that articles related to one another are likely to be cited together in earlier contributions, while studies infrequently cited together will not be cited together in earlier contributions (Di Guardo and Harrigan, 2012; Rowlands, 1999). Following previous research, this analysis was performed using a promax rotation (Culnan, 1986).
Finally, we conducted a Pathfinder analysis (Di Stefano et al., 2012; Nerur et al., 2008; Schvaneveldt et al., 1988, 1989) with the aim of mapping links among papers and elucidating their centrality in the network. The Pathfinder is an algorithm that prunes the matrix of co-occurrences by providing the most important links (Schvaneveldt et al., 1989). The resulting map constitutes a simplified network (PFNet) in which articles represent the nodes while the edge between nodes represents the frequency with which they have been co-cited. In this case, the PFNet provides scholars with the opportunity of highlighting dominant contributions and links among them (Peteraf et al., 2013).
Results
Descriptive analysis of entrepreneurship education factors
Table 2 lists the primary publishing journals, conference proceedings and books in which the articles regarding our unit of analysis appeared. The majority of contributions were published in the fields of Business, Management, Education and Learning. The 44 articles selected have 2774 citations, meaning that our unit of analysis received 35% of the citations with respect to the initial list of 1956 articles, which received 7952 citations, as shown in Table 3. This table also shows that the Journal of Business Venturing and Entrepreneurship Theory and Practice are the most influential journals as they published approximately half of the articles included in our unit of analysis.
The top publishing journals/sources of 44 most co-cited articles.
Journals and publication types varied considerably between the 1956 articles extracted without filtering and the filtered 44 articles (Table 4; Appendices 1 and 2, respectively). In the latter, publication type was considerably more homogeneous and included mostly articles (37), followed by reviews (6), editorial materials (1), books (0) and proceedings (0). Therefore, scholars are generally more inclined to cite articles than other publication types that are published in leading journals.
The top 20 publishing journals of 1956 articles extracted without filtering.
Results found: 1956 articles without filtering which received 7952 citations; average citations per articles: 4.99. Note that there are cases in which the database ISI Web does not cover the entire cycle of publications of a journal (e.g. International Small Business Journal, British Journal of Management). This implies that publications which appeared before this date are not included in this count.
Content overview of the core themes of entrepreneurship education
Through factor analysis, five factors emerge, as shown in Table 5. According to the main topics addressed by each factor, these factors have been labelled as follows: (1) introspection, (2) entrepreneurial intentions, (3) pedagogy, (4) entrepreneurial learning and (5) evaluation.
Factor analysis.
Variance explained: 81.516%; extraction method: principal component analysis. Rotation method: Promax with Kaiser normalisation. Rotation converged in six iterations.
Table 6 reports a set of descriptive characteristics for each factor. It shows that Factor 2 (entrepreneurial intentions) and Factor 4 (entrepreneurial learning) show more articles than the other factors and they have the highest number of total citations. Factor 5 (evaluation) has the most recent articles, published mostly in 2010. Factor 3 (pedagogy) has the lowest average citation growth rate in the 24-year time span. Factor 1 (introspection) and Factor 2 (entrepreneurial intentions) have the highest average citation rate in the last seven years when compared with the rate of the entire period, confirming that these two factors have the highest positive growth trend in terms of impact. Factor 4 (entrepreneurial learning) and Factor 5 (evaluation) have the lowest range of variation, which is measured by analysing the difference between the oldest and the most recent articles. The fact that both factors represent recent streams of thought that are still growing perhaps explains this trend. Factor 3 (pedagogy) has the highest level of journal homogeneity (10 articles in the group have been published by three impact-factor journals), while all contributions come from different academic sources in Factor 5 (evaluation). A description of their principal content characteristics can be found in Table 6.
Descriptive statistic for factors.
The Journal homogeneity index (div/tot) is the ratio between the journals and the total articles (N) grouped in the factor. Values near 0 indicate the maximum rate of homogeneity, and values near 1 indicate the highest level of heterogeneity among sources. The Range of variation (years) is the difference between the oldest and the most recent articles in a factor.
Introspection – Factor 1
Overall, six contributions constituted Factor 1. These articles trace the state of entrepreneurship education within university contexts. The common content linking these contributions is represented by the need to respond to the challenges associated with enhancing the quality of entrepreneurship education to attain two important objectives: (1) improving the quality of entrepreneurship studies, for instance, via doctoral programmes (Brush et al., 2003) and (2) in an effort to respond to the new mission of universities to contribute directly to the development of territories (Rasmussen and Sørheim, 2006), trying to better understand the impact of university entrepreneurship courses on individuals and society (Gartner and Vesper, 1994; Vesper and Gartner, 1997). Some authors have also highlighted the challenges in entrepreneurship education by sharing a vision of entrepreneurship as a dynamic process that requires students to learn about more than an ideal procedure to manage the venture creation and growth (Honig, 2004; Kuratko, 2005).
Entrepreneurial intentions – Factor 2
Factor 2 comprises 12 articles focusing on understanding the antecedents of entrepreneurial intentions (e.g. Hmieleski and Corbett, 2006) or opportunity recognition (Baron, 2006). This factor’s basic assumption involves defining entrepreneurship as an intentional process (Krueger et al., 2000). Accordingly, entrepreneurial intentions constitute the central concept of this core theme, which has been investigated from different perspectives. Some articles have addressed the construct as a dependent variable (Chen et al., 1998; Crant, 1996; Hmieleski and Corbett, 2006; Kourilsky and Walstad, 1998; Wilson et al., 2007) and have highlighted the implications for entrepreneurship education (Krueger et al., 2000; Liñán et al., 2011). McGee et al. (2009) have developed a new measure to collect entrepreneurial intentions and have discussed the implications for entrepreneurship education. Finally, some articles have started considering the intentional models as a theoretical framework able to guide researchers in understanding the impact of entrepreneurship education (Peterman and Kennedy, 2003; Souitaris et al., 2007).
Pedagogy – Factor 3
Factor 3 includes 10 contributions that have reflected extensively on the methods and approaches for teaching entrepreneurship. One of the main assumptions is that the complex processes surrounding the entrepreneurial experience should be the basis for creating entrepreneurship courses. The basic foundation is that entrepreneurship is a process of opportunity identification that requires the development of a wide range of skills to ensure that the trainee is able to manage his or her own business. Emotions (Shepherd, 2004), creativity (DeTienne and Chandler, 2004; Hood and Young, 1993) and the ability to manage uncertainty and unpredictable events (Neck and Greene, 2011) are presented as key constructs of an entrepreneurial training path. Many implications for entrepreneurship education have been outlined including the recently identified importance of considering social entrepreneurship as a phenomenon that requires the integration of managerial skills to cope with social and commercial goals (Tracey and Phillips, 2007). Another issue concerns the refinement of an effective connection between theory and practice. In this sense, some authors have traced a new pedagogy for teaching entrepreneurship (Fiet, 2001) and have tried to understand the contact points between start-up experiences and what a student can learn from textbooks (Edelman et al., 2008). Finally, some studies have also illustrated the state of entrepreneurship as a discipline (Béchard and Grégoire, 2005) within academic contexts (Katz, 2003, 2008).
Entrepreneurial learning – Factor 4
Factor 4 includes 11 articles addressing the topic of entrepreneurial learning, which is viewed as a key construct that enables scholars to better understand the entrepreneurial phenomenon (Cope, 2005; Harrison and Leitch, 2005; Politis, 2005; Rae, 2006). This phenomenon should be addressed using various interpretative lenses (Gibb, 2002). Cognitive processes represent the theoretical keystones, providing an insightful means for interpreting how entrepreneurs make decisions in uncertain conditions (Holcomb et al., 2009), how they succeed (Ravasi and Turati, 2005), how they discover opportunities (Politis, 2005), how they learn from failures (Cope, 2011) or critical events (Cope, 2003) and how previous knowledge might affect future expectations for firm growth (Parker, 2006). Knowledge structures are considered as a central construct allowing us to understand the entrepreneurial behaviour (Holcomb et al., 2009). A different perspective is provided by Pittaway and Cope (2007b), who analysed the practical implications for entrepreneurship education, as derived from the entrepreneurial learning perspective.
Evaluation–Factor 5
Factor 5 is composed of five articles investigating with evaluations. With the exception of the contributions of Martin et al. (2013) and Pittaway and Cope (2007a), which are quantitative meta-analyses on the impact of entrepreneurship courses and a systematic literature review focused on entrepreneurship education, the remaining publications are empirical studies aiming to investigate entrepreneurship education outcomes (Athayde, 2009; Oosterbeek et al., 2010; Von Graevenitz et al., 2010).
Disentangling the content structure of entrepreneurship education
The distance and proximity of the core themes of entrepreneurship education
Figure 1 shows the MDS results. By combining the position of such articles in the plan with its content, two dimensions emerge that allow us to provide a broader description of the foci theoretically characterising the topic: (1) the learning process and (2) the level of analysis.
Multidimensional scaling and factor analysis.
The horizontal axis can stand for the learning process, which has two different foci according to its extremes: one in which the learning process is observed within a real environment and one in which it is observed within a training context. On the left, Factor 4 (entrepreneurial learning) refers to articles that investigate how entrepreneurs learn from their experiences in an attempt to explain the mechanisms and dynamics surrounding entrepreneur changes. Moving towards the right side, the rest of the factors explore the learning process within a training context. These articles focus upon understanding which contents, pedagogies and outcomes deserve to be considered effective in the delivery and evaluation of entrepreneurship courses.
On the vertical axis, the top is distinguished by the levels of analysis that have been examined. At the top, Factor 2 (entrepreneurial intentions) and Factor 5 (evaluation) have empirically investigated student entrepreneurial intentions in an effort to understand antecedents (Hmieleski and Corbett, 2006; Liñán et al., 2011) or to reveal the training effects of entrepreneurship courses (Oosterbeek et al., 2010; Peterman and Kennedy, 2003). Moving down from the top of the map, the focus of interest has switched from the individual level to training characteristics. Specifically, at the bottom of the map, the vertical axis juxtaposes Factor 1 (introspection) and Factor 3 (pedagogy), both of which have focused on training content and approaches to delivering entrepreneurship education (DeTienne and Chandler, 2004; Fiet, 2001); these factors have also examined challenges and trends in entrepreneurship education (Vesper and Gartner, 1997). The map shows that these two factors appear to be scattered and not clearly clustered, as is the case for the factors in the upper side of the map. The scholars who co-cited these references have considered them complementary, which may explain the map’s composition. Some studies furnish a theoretical foundation with respect to the need to adopt new methods and procedures to deliver entrepreneurship education (Honig, 2004; Kuratko, 2005), while some offer new insights and empirical examples in that direction (DeTienne and Chandler, 2004; Shepherd, 2004).
With regard to Factor 4 (entrepreneurial learning), the map shows the articles to be clustered around the horizontal axes. Most of the contributions offer a theoretical perspective by introducing propositions to be verified in future research (Politis, 2005) and adopting narrative approaches to investigate the learning processes (Cope, 2011; Rae, 2006), except for Parker (2006) who proposed an econometric framework to gauge the extent to which entrepreneurs consider new information at the expense of what they already know when making decisions. Overall, the focus of their analysis is on the learning process, and the individual dimension represents a source of data that furnishes an empirical basis for inferring learning dynamics.
The figure also shows that most of the articles are concentrated on the left side of the map, except for those falling under Factor 4 (entrepreneurial learning), which appears to be quite isolated from the other factors. A greater distance, in particular, emerges between Factor 4 (entrepreneurial learning) and Factor 2 (entrepreneurial intentions), while proximity can be detected among Factor 4 (entrepreneurial learning), Factor 1 (introspection) and Factor 3 (pedagogy); the contributions of Pittaway and Cope (2007b) and Gibb (2002) appear to act as connectors. As mentioned in the ‘Methodology’ section, these aspects are useful in identifying gaps and opportunities for future research.
A network overview of co-citations
The network in Figure 2 highlights five central nodes of the knowledge structures, based on the number of links connecting them to other nodes. This representation particularly coheres with the factor analysis as it reveals the existence of different hubs, each referring to a specific core theme.
Pathfinder of the field of entrepreneurship education.
Cope (2005) is connected with studies investigating entrepreneurial learning and emerges as a reference point for this cluster. Krueger et al. (2000) are the basis for the core themes associated with entrepreneurial intentions (Factor 2) and evaluation (Factor 5). Kuratko (2005) is linked with the works addressing new pedagogies and topics for entrepreneurship education (Fiet, 2001; Tracey and Phillips, 2007) and joins the hub dealing with introspection (Factor 1), in which Katz (2003) is the central point. Honig (2004) represents a less dense hub when compared with the nodes discussed previously. However, it is worth noting that this work connects the contributions addressing new training topics and approaches. This finding agrees with the idea proposed by Honig (2004), who laid the foundation for the adoption of an experiential approach in delivering entrepreneurship courses.
All these hubs are connected by three works that seem to function as bridges. One of these is a systematic review of the literature written by Pittaway and Cope (2007a). The authors depicted a broader representation of the literature by showing the core areas of entrepreneurship education and outlining the important gaps. In this respect, they have suggested an increase in the cross-fertilisation among the different areas, indicating adult and management learning as important theoretical sources to be carefully reviewed. This work, in particular, connects the entrepreneurial learning core theme with those that have addressed new pedagogy and trends in entrepreneurship education.
Another bridge is represented by the meta-analysis of Martin et al. (2013) who reviewed empirical studies in an attempt to verify the effects of entrepreneurship education on entrepreneurial intentions and behaviours. This work links both the core themes of entrepreneurial intentions (Factor 2) and evaluation (Factor 5) with the remaining hubs. Peterman and Kennedy (2003) play a similar role. Their work connects the most recent contributions focusing on the evaluation theme (Factor 5) with those contributions addressing the pedagogy theme (Factor 3).
This network also illustrates the straightness of the links; with regard to the thicker links, a path connecting studies on entrepreneurship intentions with studies reviewing entrepreneurship trends emerges. The most important connection, based on the frequency with which two nodes have been co-cited, is represented by Chen et al. (1998) and Krueger et al. (2000), whose studies are also closely connected with the study by Souitaris et al. (2007). These authors contribute to building the basis of entrepreneurship education by emphasising the key role played by entrepreneurial self-efficacy, which is both a key variable within intentional models and an important outcome of training effectiveness programmes. Another important link emerges between Kuratko (2005) and Katz (2003); both authors have reflected on the maturity and legitimacy of the entrepreneurship division and have contributed to the global portrait of entrepreneurship education as a discipline.
Discussion
This article shows the intellectual structure of entrepreneurship education resulting from a co-citation analysis covering the period 1991–2014. These results provide a complementary perspective to the existing literature as we focused on citing behaviours to investigate the theoretical foundation of the construct. Through co-citations, this article illustrates the most frequently targeted core themes, unveils their content and shows their interconnections. According to the scope ascribed to co-citations, the emerging factors can be highlighted as determining the theoretical core themes of each topic (Zhao and Strotmann, 2014).
The first step of the retrieval process aiming to consider any contributions related to entrepreneurship education produced 1956 publications. By applying a set of filtering criteria developed to identify co-citations, 44 articles were included in our final unit of analysis, which received 2774 citations. Through multivariate analyses, we have investigated how these citations shaped the field.
The key outcomes are as follows: first, the field has a polycentric structure with five core themes addressing the topic from different perspectives: first, introspection, outlining the manner in which the topic has evolved as an academic discipline; second, entrepreneurial intentions, investigating the antecedents and implications of the topic; third, pedagogy, trying to determine the content and teaching methods for entrepreneurship; fourth, entrepreneurial learning, studying how entrepreneurs learn from their experiences; and fifth, evaluation, exploring the effectiveness of training.
Second, our findings reveal entrepreneurial intentions (Factor 2) as one of the most important core themes in entrepreneurship education, in accordance with previous observations (Fayolle and Liñán, 2014). This factor appears to have both a longer publication history (1994–2011) and a higher citation rate (13 citations for articles during the 2007–2014 period) than other factors. Models of entrepreneurial intentions drive the empirical analyses seeking to understand the processes that lead people to become entrepreneurs with implications for entrepreneurship education. Studies exploring entrepreneurial intention provide an explanation of the surrounding mechanisms and the leveraging dimensions that influence behavioural intentions; they offer many ideas about how to deliver training to influence, for example, perceived control, social norms or entrepreneurial attitudes – considered to be closer antecedents of intentions (Ajzen and Fishbein, 2005); finally, these studies establish clear outcomes by measuring the impact of entrepreneurship education at the individual level. This factor offers an important theoretical framework offering a solid basis for empirical research.
Third, this analysis shows that the core theme of entrepreneurial learning (Factor 4) has attracted a great deal of scholarly interest indicated by the average citation rate − 5.79 citations for articles during the period 2007–2014 (Table 4). Furthermore, according to the MDS solution, this factor emerges as a potential instigator of change. The map reveals proximity between this factor and pedagogy (Factor 3). The Gibb (2002) study represents a key contribution acting as a bridge between the two themes. Gibb claimed that entrepreneurship is a complex phenomenon that should be analysed by adopting various interpretative lenses to extend the concept beyond the business sphere. By following this path, scholars working on training methods and teaching approaches have emphasised the idea of entrepreneurship as an unpredictable phenomenon that requires a mindset to cope with uncertainty (Neck and Greene, 2011). Entrepreneurial learning, which aims to understand the learning mechanisms with which entrepreneurs generate new knowledge when coping with routine or critical events (Cope, 2003; Pittaway and Cope, 2007b), represents a theoretical source for the creation of a training environment that might enable students to develop the skills required to manage this complexity. It is plausible that these studies will prompt new research and empirical approaches to evaluate training outcomes (Pittaway et al., 2010).
Fourth, this work shows that understanding the real impact of entrepreneurship education has been increasingly emphasised in recent years. A result worth noting, which has been made available through network analysis, is the loop linking the most recent contributions examining entrepreneurship training effectiveness. This loop links the works of Souitaris et al. (2007), Oosterbeek et al. (2010) and Von Graevenitz et al. (2010). These analyses are theoretically grounded in entrepreneurial intentions (Factor 2), with a link between Souitaris et al. (2007) with Krueger et al. (2000). This empirical work has primarily assessed the effect of entrepreneurship education by highlighting entrepreneurial attitudes, intentions and self-assessment skills as training outcomes. This loop will probably become stronger in light of recent calls (Fayolle, 2013) to improve the quality of entrepreneurship education and approaches to evaluate training outcomes (De Clercq and Arenius, 2006; Fayolle and Gailly, 2015; Fayolle and Liñán, 2014).
Finally, by examining proximities, distances and connections among contributions, this article highlights cross-fertilisation among the core themes. In particular, the role of literature reviews has been revealed as a bridge between separate core themes. This issue is particularly noticeable when observing the network that shows the Pittaway and Cope (2007a) systematic literature review and the Martin et al. (2013) meta-analysis as connecting nodes in the entrepreneurship education topology. Such analyses can help build a theoretical framework through which a comprehensive view of the entrepreneurship education can be created.
Limitations
This study should be examined in light of its limitations, which are inherent to the co-citation method and the use of the ISI Web of Science as a database. Regarding the limitations inherent to the method, immediacy and keyword generation represent the most relevant sources of bias in co-citation analysis (Hicks, 1987). In relation to immediacy, the concern is that this method requires a long time span to obtain an adequate number of citations. Therefore, most recent papers, which are likely to have a lower rate of citations compared with less recent publications, might be underestimated. To minimise this risk, in line with previous research (Di Guardo et al., 2012; Small, 1977), we decided to reduce the minimum threshold from 40 to 10 individual citations. Furthermore, we adopted a yearly average citation rate as a criterion, enabling us to include newer high-impact papers, such as Martin et al. (2013). This choice is justified by a recent study (Stern, 2014) that showed that the early year citation trend of a paper remains constant over time.
Regarding keyword generation, a recurrent bias involves losing relevant papers or including non-relevant papers (King, 1987; Rowlands, 1999). To reduce this type of risk, we adopted a strict method for generating keywords and eliminated the papers without co-citations. These choices allowed us to balance the need to select the most high-impact papers examining entrepreneurship education by minimising the risk of including extraneous contributions.
Another limitation is that the results are strictly dependent on the database adopted which, in this case, was the ISI Web of Science (Reader and Watkins, 2006). The major concern of using any specific database is that the results are restricted to the contributions indexed internally and are dependent on the operation rules that drive its functioning (e.g. the criteria that contributions must meet to be included in the database, which can rely on quantitative, qualitative or hybrid evaluation approaches). The core themes and connections highlighted in this study present only a partial view of the topic, as reflected by the chosen database. However, it should be noted that the adoption of the ISI Web of Science as a data source provides this study with a methodologically corroborated path for performing bibliometric studies, as it has a longer history compared with other databases (e.g. Scopus and Scholar).
Implications and future research
This article conveys a complex picture of entrepreneurship education; many topics crowd the field, each of them deserving further empirical exploration. The practical implication is that this complexity requires both creativity and rigorous experimentations by professionals and scholars. Many of the studies reviewed call for a paradigm shift in terms of how the entrepreneurship phenomenon is conceptualised. This call implies that research and practice should be able to accept the challenge of continuous experimentation with methods, teaching approaches and evaluation frameworks to generate new knowledge and new theoretical inputs.
From a theoretical point of view, the study has shown that each core theme furnishes insightful inputs to guide practice and new studies. However, a special effort should be made to theoretically coalesce these different foci. Many challenges, for instance, should be seized by scholars working on entrepreneurship effectiveness to combine the methodological rigour required in training evaluations with the complexity inherent in entrepreneurial concepts (Block et al., 2011; Díaz-Casero et al., 2011). In particular, the network analysis has shown the connection between Honig (2004) and Peterman and Kennedy (2003), suggesting a route towards merging the teaching perspective studies with those aimed at evaluating training effectiveness. This link advocates incorporating new skills acquisition in evaluation programmes. This idea supports arguments by Pittaway and Cope (2007a) regarding the extent to which entrepreneurship education enables students to become more effective entrepreneurs.
Much effort is also required to understand how and to what extent entrepreneurship education can help people to successfully manage their emotions, creativity and ability to seize entrepreneurial opportunities and entrepreneurial intentions. Entrepreneurial intentions and the mechanisms surrounding entrepreneurial learning deserve to be theoretically merged to provide entrepreneurship education scholars with an extensive framework, allowing them to better explain the relationships between intentions and skills in managing entrepreneurial complexity. An insightful example is the study by Pittaway and Cope (2007b) that simulated some elements of entrepreneurial learning, such as the emotional exposure, within an entrepreneurship education course.
Many additional insights might emerge through a deeper exploitation of bibliometric approaches. In that sense, future research should consider the adoption of bibliographic coupling to focus on the most recent trends characterising the topic (Zhao and Strotmann, 2014). Such research would allow the theoretical core highlighted in this study to complement recent trends. Furthermore, bibliometrics show that mapping a scientific topic reveals the database’s peculiarities (Levine-Clark and Gil, 2009). Every study suffers for having a partial view of a field that can be overcome by carrying out complementary works (Glänzel, 2012). Therefore, future research in entrepreneurship education should use different databases (e.g. Scopus, Scholar), which would make complementing the topology outlined in this work possible.
The field of entrepreneurship education appears to be intrinsically dependent on multi-theoretical sources and progresses during their development. Accordingly, our investigation suggests that future research should advance along two interconnected paths: one devoted to empirically exploring each core theme and one aimed at finding new theoretical connections among these themes and future core topics, nurturing a continuous process of internal meta-cognition and introspection (Fayolle, 2013).
Conclusion
All the analyses performed converge to represent the topic of entrepreneurship education as composed of various foci. The theoretical core is represented by the entrepreneurial intentional construct, which offers a guide both to define training objectives and to establish the basis for evaluating training effectiveness. The entrepreneurial learning factor appears to be a promising theoretical source that will nurture training approaches and contents. Training effectiveness is the most recent theme that might play a key role in suggesting new theoretical inputs for the other factors. The most crucial challenge that scholars should address is providing a more comprehensive framework that is able to explain the theoretical path connecting the core themes that have emerged.
Footnotes
Appendix
Types of publication for citing articles.
| Source | N = 7952 (%) | N = 2774 (%) | |
|---|---|---|---|
| 1 | Article | 6027 (76) | 2042 (74) |
| 2 | Proceedings paper | 988 (12) | 480 (17) |
| 3 | Book/book chapter | 63 (1) | 6 (0) |
| 4 | Editorial material | 342 (4) | 33 (1) |
| 5 | Review | 529 (7) | 213 (8) |
Funding
The author(s) received no financial support for the research, authorship and/or publication of this article.