De la bibliometría al emprendimiento: un estudio de estudios ; From Bibliometrics to Entrepreneurship: A Study of Studies

Bibliometric studies of entrepreneurship as a discipline have contributed fundamentally to the creation of a certain order in an apparently chaotic and contradictory literature, examining how the discipline has developed, giving a comprehensive vision of the structure of the field, observing its social networks, detecting trends, discovering knowledge gaps and helping to plan future research lines. The purpose of this article is to explore this special type of research. In terms of methodology, it uses an adaptation of the Systematic Literature Review, and a content analysis using text-mining software in order to look deeper into objectives, conclusions and limitations. Among the main findings, there is some evidence that indicates that the image presented to date about entrepreneurship has not considered the multidisciplinary nature of the field and could, therefore, be distorted. At the same time, a series of inherent problems have been detected, and it has become evident that there is a need to incorporate the latest advances in bibliometrics and to improve collaboration between experts from both fields in order to solve those mentioned issues and move towards future progress.


INTRODUCTION
and Ratnatunga and Romano (1997) brought bibliometrics into the world of entrepreneurship studies towards the end of the 1990s with their works focused on small businesses. However, the first bibliometric analysis dealing with entrepreneurship as a discipline was carried out by Dery and Toulouse (1996) in order to shed light on its social structure. Ever since those first seminal bibliometric research projects on entrepreneurship, there have repeatedly been works dealing with the discipline as a whole, culminating in a special issue published in 2006 by one of the most prestigious journals in the field (Gartner et al. 2006).
All of those studies, and some of the more recent ones (Ferreira et al. 2019;Xu et al. 2018;Landstrom and Harirchi 2018), have contributed fundamentally to the creation of a certain order in an apparently chaotic and contradictory literature with disparate meanings, views, and ways in which entrepreneurship is used and referred (Audretsch et al. 2015), examined how the discipline has evolved, given an overall view of the structure of the field, observed its social networks, spotted tendencies, discovered knowledge gaps and helped plan future lines of research. However, despite being a special, and particularly difficult, type of research that brings together two fields (bibliometrics and entrepreneurship) with peculiar characteristics, and approaches a discipline in its totality, little is known still about it. No review has yet taken up the task of analysing the works that are the result of this kind of research. The study presented below attempts to bridge that gap and is designed as a study of studies, offering a double perspective, looking at the contributions made to entrepreneurship as well as presenting a vision from a point of view of bibliometrics.
The aim of this article is to explore the different bibliometric analyses carried out on entrepreneurship as a discipline by putting together a representative sample selection of documents and subjecting them to a subsequent analysis. In terms of methodology, it employs an adaption of the Systematic Literature Review (SLR) developed by Tranfield et al. (2003) and a content analysis using text mining software in order to detect key features and to respond to the following research questions: How have studies of this type evolved? Have they been able to offer a picture that faithfully reflects the discipline? What have their main objectives and conclusions been? Which specific problems were they confronted with? Have researchers been applying the latest tendencies in bibliometrics? What has bibliometrics been able to contribute to the research in entrepreneurship so far, and what can it still contribute in the future? What is the future for this kind of research?
The rest of the study is divided into five sections. The starting point is a literature review that looks mainly at the origins, evolution, fundamentals, tendencies and limitations of bibliometrics. The next section deals with bibliometric research in entrepreneurship. The third is centred around the most significant methodological considerations. The fourth shows a discussion of the main results obtained. The final section is a presentation of the main conclusions.

LITERATURE REVIEW. BIBLIOMETRIC RESEARCH IN ENTREPRENEURSHIP
Entrepreneurship is an extraordinary phenomenon. It is a field that is able to bring together the interests of institutions, scientists and society as a whole. It is so special that there is almost unanimous agreement on its significance. This has led to a situation in the last decades where the number of institutions that offer their support to new entrepreneurs has not stopped growing, and a document corpus has evolved, which aims to decipher its key characteristics. The question remains, however: Is it a phenomenon, a field or a discipline?
Despite its undeniable social recognition and 'popularity', a series of questions have always provoked a profound debate about it in academic circles, as shown in Kushkowski (2012). The debate includes questions related to the way in which research is conducted (Venkataraman 1997), to methodology (Busenitz et al. 2003;Low and Macmillan 1988), or to the fierce debate over whether it is an independent and legitimate discipline (Shane and Venkataraman 2000) or an interdisciplinary field based on the study of empirical phenomena (Sorenson and Stuart 2008), and even those tackling its essence, the actual definition of entrepreneur (Carlsson et al. 2013), and the lack of consensus hereon.
This definition remains elusive, heterogeneous and complex like each of the entrepreneurs it represents. On the political-institutional level there is an elevated consensus traditionally based on the important economic and social benefits entrepreneurship generates, which has almost without exception been the main reason for its study (van Praag and Versloot 2007). These widely studied benefits can be translated into more and greater economic growth, increasing productivity and competitiveness, the discovery of opportunities, emergence of innovation and dynamic generation of employment (Shane and Venkataraman 2000;Audretsch and Thurik 2001;Audretsch et al. 2006;Acs et al. 2009).
Characterised by an apparently chaotic and contradictory literature (Audretsch et al. 2015), it positions itself as an ideal candidate to take advantage of the potential of bibliometrics. Zupic and Cater (2015) point out that in the context of management and organisation, bibliometric methods contribute, among other effects, to the synthesis of past research findings, to the advancement and discovery of new lines of investigation, to the introduction of systematic, transparent and reproducible review processes, to the improvement of the quality of reviews, to the mapping of different specialties, to the introduction of objective measures for literature evaluation leading to increased rigor and reduced bias, and to the detection of formal as well as informal networks (invisible colleges). These tools when applied to entrepreneurship multiply its possibilities. Their use allows researchers the opportunity to make headway in their theoretical understanding of it and to analyse such relevant questions as the ones mentioned, in greater depth. Research in this area is still young and its representation is weakened as it blurs into other categories in the two main scientific reference bases (Web of Science and Scopus).
The generalised use of databases such as Web of Science (WoS), Scopus and others as a crucial resource which allows scientists to access an elevated number of documents and all the bibliometric information they index (references, citations, etc.) in their research area, as well as the development of software that provides better handling of the resulting data and makes the tasks involved in their analysis more efficient, has led to a more widespread use of bibliometric tools: Sitkis (Schildt 2002), one of the first; Bibexcel in combination with Pajek (Persson et al. 2009), and others such as SciMAT, HistCite, CiteSpace, VOSviewer.
The documents that contain the knowledge accumulated in entrepreneurship are scattered over a variety of, at times very different, categories in the two mentioned databases. As Landstrom et al. (2012) indicate, the phenomenon is multidisciplinary in nature, and registers mainly under Management, Business, Economics (WoS) or Business Management and Accounting, Economics, Econometrics and Finance (Scopus), as well as in other categories like Psychology, Sociology, History, etc., but to a lesser degree.

METHODOLOGY
An adaption of the procedures developed by Tranfield et al. (2003) has been followed. Systematic literature reviews (SLR) differ from traditional ones in that the process is reported openly in the same way empirical research would be, and that they are governed by transparency, clarity, equality and accessibility (Pittaway and Cope 2007). The methodology, applied and developed in detail in the context of Management and Social Sciences in works such as Liñan and Fayolle (2015), Pittaway et al. (2004), Thorpe et al. (2006), has been modified to adapt itself to the objectives and requirements of this study. The procedure which has been followed is illustrated in the Figure 1. Step 1 Search Start: The objective of the review is to respond to different research questions: How has bibliometric research on entrepreneurship as a discipline been conducted? What have been the main objectives and conclusions? What specific challenges had to be faced and how were they overcome? Has a reliable image of the field emerged? Are the results up to date? What can bibliometrics offer to research into entrepreneurship and what has it contributed so far? Have advances in bibliometrics been used to keep research up to date? What does the future of this type of research look like?
In order to comply with such a variety of objectives, a non-restrictive strategy has been opted for, performing the document search in the two reference databases of the scientific community, WoS and Scopus, and using Google Scholar as an auxiliary tool to find more documents and download those whose access is restricted in the former.
In terms of the formula applied in the search, after testing different combinations and performing a number of tests runs it was decided to carry out accumulative searches for word pairs without applying any type of filters such as year or document type. Keywords associated with bibliometrics detected in previous works on information sciences were used (Chang et al. 2015), in total 23 terms and their possible lexical variations of occurrence were collected (for example: co-citation analysis / cocitation analyses), combined with the root "entrepr *", strategy previously used in bibliometric analysis on entrepreneurship (Cornelius et al. 2006, Schildt et al. 2006. A combination of terms that could include the different bibliometric works were used for document selection. The complete sequence is detailed as supplementary material in Annex 1. Following this strategy, 260 documents were compiled from WoS and 257 from Scopus. The inclusion/exclusion criteria were established in accordance with the previously mentioned objectives • Documents with a global focus on entrepreneurship (as a field or discipline), using bibliometric tools, indicators or analyses form part of the review.
• Documents focusing only on fragments or specific areas in entrepreneurship research (social entrepreneurship, female entrepreneurship, family firms, small enterprises, etc.) and those that do not carry out a whole analysis have been discarded.
Step 2 Search Start: All titles and abstracts of every document compiled according to the previously described criteria were read. Applying the first filter, the result already provides valuable information for the study. Among those documents discarded for their lack of global focus on entrepreneurship, an increasing amount of literature can be found where bibliometric analyses are used as a main tool or as a complement to explore specific lines or sub-fields of research that are closely linked with entrepreneurship (family business, female entrepreneurship, social entrepreneurship, born global firms, informal entrepreneurship, international entrepreneurship), or simply to answer very specific questions in certain lines of investigation (Caputo et al. 2018;Galvao et al. 2018).
The result was a first list of 30 documents:

Figure 2. Common documents
The wide search sequence produces documents that do not tie in with the proposed objectives. It does, however, offer more results to process and picks up some as Yu and Tang (2014) or Qian (2014) that would not have been included otherwise. Furthermore, the results are enriched by allowing any type of document to be included. Auxiliary searches were carried out in Google Scholar, using the same strategy, which resulted in 6 additional documents that had previously not been detected in Scopus or WoS.
This section and the next one will inevitably introduce a certain level of subjectivity, as the inclusion or exclusion, for example, of those documents which generate doubts, such as Bhupatiraju et al. (2012) or Schmitz et al. (2017), was decided in accordance with the objectives of this review, although it is subject to different interpretations. In the end, the mentioned documents were not included.
Step 3 Characterization: In the third phase, the definitive sample emerged. To achieve this, all documents produced in the previous phase were read, discarding those where access has been impossible (some documents in the sample are proceedings that weren't available in their complete form). Additionally, conference presentations were substituted by the articles that later reported on them.
The list contains two works that compiled a list of bibliometric studies in entrepreneurship as part of their research: Landstrom and Persson (2010) and Teixeira and Ferreira (2013). They were used in order to be contrasted with the results obtained and to add documents that had not been found but which comply with the criteria for the review. The final result consisted of 40 documents 1 , subsequently, the sample was characterised, and all the relevant information required to respond to the proposed research questions was identified.
Step 4 Content Analysis: Text mining software was used to detect key characteristics and to increase the objectivity of the study. The process followed mainly consisted of: • New reading and independent extraction of objectives, conclusions and limitations of the document sample. This resulted in 120 text files (40 for each item) • Pre and post-processing tasks carried out in Wordsat 8.0.7 and Qda Miner 5.0.23 by Provalis Research: mainly consisting of the exclusion of terms not required for the analysis (a, about, an, another, etc.) as well as word substitutions (develop, developed, develops = development, etc.), and the definition of the frequency threshold of words to be included in the analysis (add words with frequency = higher than 4).
• Topic extraction using the WordStat function.
Application of a combination of natural language and statistical analysis; mainly factor analysis. Topic extraction is achieved by calculating the frequency matrix of documents and words. Clustering and co-occurrences. The following configuration was used: occurrence (same document), index (Jaccard's coefficient); type (word co-occurrence first order).
This phase and the next one (step 5: conclusions) are developed together with the results and conclusions of the article.

Characterisation of the Document Sample Set
According to various authors like Landstrom and Persson (2010), Sassmannshausen and Volkmann (2018) the seminal bibliometric research in entrepreneurship is to be found in the articles by Romano and Ratnatunga (1996), Ratnatunga and Romano (1997), both of which are centred on small enterprises. According to the list, the first research of this type focusing on entrepreneurship as a field was carried out by Dery and Toulouse (1996) in an attempt to shed light on its social structure.
Two works appear next: Shane (1997), Busenitz et al. (2003. They do not mention the use of bibliometric tools specifically in their methodology. They have, however, been considered to be studies of this type by Landstrom and Persson (2010), which is the reason why they have been included in the list, as they respond to the definition of what a bibliometric study is ("Bibliometric studies, in which a given field is studied by means of quantitative analysis and statistics to describe publication patterns"), and comply with the proposed objectives for this review. Both were published in Journal of Management and show how certain areas, categories or even journals are more likely to be cited as references in later works. Busenitz et al. (2003) were the most cited from the Google Scholar, Web of Science and Scopus sample. Figure 3 shows how global bibliometric studies have gradually gained importance and have shown a stronger and more consistent presence in a number of publications. As Sassmannshausen and Volkmann (2018) point out, the publication in 2006 of a special issue of "Entrepreneurship Theory And Practice" (Gartner et al. 2006) that brings together some of the most valued articles in number of citations, can be seen as the starting point of a growing reputation. Since then, other studies have been carried out which show that the community of researchers in entrepreneurship has felt the need to regularly compile the acquired knowledge in the subject by using these types of tools as a way to help lead to new advances. Especially the article by Landstrom et al. (2012) stands out among the previously mentioned documents as the one that required the shortest exposure period to accumulate the citations necessary to position itself as a reference. The sample concludes in 2018; the year for which 5 documents have emerged so far. Together with 2006 (4), 2014 (5) and 2015 (5) it is one of the years which records the greatest number of works.

Characterisation of Content
This has been divided into two parts. The first analyses technical aspects of the documents: time frame, data retrieval, unit of analysis, search terms, sample, software and main bibliometric analysis, and the second examines the main objectives, conclusions and limitations encountered in the different bibliometric studies of entrepreneurship as a discipline.

Technical Aspects
A summary with the main technical aspects is available as supplementary material (Annex 3). Showing up next, the most significant results are extracted. i.e. analyses of shorter intervals between 5 and 10 years, which would present the most recent picture of the current state of research. This can be explained because the majority are based on studies of citations, and documents need at least three years of exposure in order to accumulate them. Moreover, they tend to subdivide those ampler periods into smaller intervals in order to better observe their evolution.
Data Retrieval: When it comes to compiling the information required to elaborate different analyses, many of the samples establish a search sequence, and extract it directly from ISI-WoS. Another large group does it by choosing several journals that are representative of entrepreneurship research and then extracting the information. That group too, however, regularly uses ISI-WoS once the appropriate sources have been established, except in rare cases like Teixeira (2011), which uses Scopus for retrieval.
There have been few works that have required the prior creation of a specific database to be studied subsequently, or that have used alternative sources such as books Landstrom et al. (2012) to retrieve articles and references.
The choice of database for information retrieval exposes one of the biggest problems any study of this type must face: trying to find a collection of documents that represents the discipline as a whole. Different authors in the sample favour various strategies, and mostly tend to justify their choice by referring to the coverage offered by the chosen databases or journals. Dery and Toulouse (1996) and Gregoire et al. (2006) choose their exclusive source in this way, Journal of Business Venturing (JBV) and Frontiers in Entrepreneurship Research (FER), respectively, and note this as a major limitation. However, selecting more than one source like da Costa Ferreira (2009) or Teixeira (2011), for example, does not solve the problem either, as not all the documents contained in the chosen journals deal exclusively with entrepreneurship.
It can also be observed that with the exception of Cabeza-Ramírez et al. (2017), which use frequency to unify data in one single index and thus manage to work with two databases, the possibility of using different ones in order to complement each other has not been explored. This is because the majority of works are based on citations, and different citation patterns cannot be mixed.

Unit of Analysis:
One aspect that tends to go unnoticed is that it is necessary to observe units of analysis for data retrieval. The most commonly used units in bibliometric studies, network building and science mapping are documents (including any indexed typology and information: articles, books, notes, proceedings, papers, reviews, letters, etc.), articles (with indexed information: authors, cited references, journals, etc.) authors (including affiliations) and words or terms of description. According to our sample, more than half of the documents exclusively use articles. Although 25% of the sample reach into other typologies, not including books might represent a major bias in a field like entrepreneurship, where the elaboration of textbooks is common.
It is noteworthy that content studies using words, for example, are hardly represented at all. On the other hand, there are no studies focused specifically on references either, as these have been analysed like any other element in those studies that use the article as a unit of analysis, and it is difficult to find works that specifically deal with references.

Search Terms:
The absence of specific categories for documents on entrepreneurship in the main databases together with the difficulty of defining entrepreneur or entrepreneurship means that the search strategies used to find documents that are representative of the discipline become more complicated. In most cases of our sample, the root "entrepr*" or the combination of different terms has been the chosen option. Selecting one or another option can lead to a significant change in the results. Although there is no literature on the topic, taking a definition of entrepreneur/ entrepreneurship that is in accordance with the proposed objectives can make the task easier. It can be used to create different search terms, the most adequate of which will then be used to filter the documents. On the other hand, including all the results obtained from a specific source, one or several journals, for example, or compiling documents randomly without first applying a filter, would mean that documents that are less likely to be classified as dealing with entrepreneurship would end up being included.

Sample:
When it comes to the sample documents selected for the different bibliometric studies, there is no connection between the number of documents chosen and the number of years under study. In the group of studies that used the article as a unit of analysis, we can find Busenitz et al. All of this brings us back to the problem researchers face, which is to find an adequate and representative sample of documents. When observing the 40 documents, it seems that most of them have chosen to select a wide-ranging sample in order to use the greatest number of articles and documents possible. This might, however, not be the perfect strategy, since even if we manage to compile all accumulated knowledge, not all of it has had real repercussions and led to an advance in understanding. In the sample, Cabeza-Ramirez et al. (2018) use this idea to look for possible solutions from a bibliometric approach, using citation thresholds. A method suggested by Martinez et al. (2014) was used here to identify the classics of a scientific area applying the H-Classic approach and the H-Index.

Software:
In recent years, significant advances have been made in bibliometric software, tools specifically designed to aid with complete workflows as well as with science mapping (Gutierrez-Salcedo et al. 2018;Cobo et al. 2011), which have had an impact on research (Pan et al. 2018). It is noteworthy to see though, that more than half of the sample documents (Table I) do not make use of them or specify them.

8
Main Bibliometric Analysis: In the sample, citation and co-citation analyses including authors, documents or co-cited journals stand out first ( Table  II). The second-most relevant type (35%) are works where evaluation, performance or scientific production are analysed, establishing different rankings of authors, articles, countries, universities, journals or impact. Co-word content and bibliographic coupling studies have hardly been used.
The only document that uses co-word analyses (Lopez-Fernandez et al. 2016) was meant as a complement to an author co-citation analysis (ACA) "to trace the connections between researchers and fields". The sample documents as a whole display a clear interest in getting to know the authors and the most representative works, as well as in understanding the relationships that they have established between them. Aspects related to the actual content of those works are of secondary importance. This seems to present a major gap in the representation of the discipline and an opportunity for future research.
Another noteworthy aspect is that practically all the works are based on citations as an indirect measure of quality. Two problems emerge which have hardly been dealt with: the time citations need to accumulate and the multidisciplinary nature of entrepreneurship. This means that articles with a shorter period of exposure to citation or belonging to another discipline with different exposure and citation patterns have  2014) have performed approximations to counteract these disadvantages. The former developed the J index in order to let works with low citations rates, but a more recent publication date, move up in the ranking, and the latter used the mean observed citation rate (MOCR) as an indicator for impact. In this last article, one of the authors (W. Glänzel) is an expert in bibliometrics. It is the only one in the sample that uses bibliographic coupling analyses as an alternative to citation studies. This methodology has been proven to be effective in identifying changes in research topics (Chang et al. 2015).
It is also noteworthy that, even though the H-Index has been a major milestone in the world of bibliometric indicators it has hardly found application in the study of entrepreneurship (i.e Cabeza-Ramírez et al. (2017) and Cabeza-Ramirez et al. (2018) used it not only to determine the citation threshold but also for sample selection). Observing a single citation pattern, the one used in ISI-WoS, is the norm. Experimenting with other ones like those used in Google Scholar or Scopus or to make comparisons would, no doubt, be enriching for the results. New metrics linked to the social development or the use of science are also not used, although their application could contribute to a better understanding of the discipline.

Objectives, Conclusions and Limitations
A summary with the three main content items is available as supplementary material (Annex 4). The results obtained with the word processing software (Qda Miner and Wordstat) and their qualitative analyses are presented below.

OBJECTIVES
The analysis of the objectives of the 40 studies of the sample was carried out after the individual reading of each document. The objectives were isolated in an individual text document for each element of the sample and the 40 resulting files were introduced in the text mining software. According to the WordStat User Guide, the Topic Extraction function attempts to uncover the hidden thematic structure of a text collection through natural language processing and statistical analysis. This function is used to increase objectivity and facilitate interpretation of content.
The objectives of the articles are usually found in the introduction section. The number of words that shape the text of the objectives is usually reduced; therefore, it was decided to extract only the 7 most representative thematic nuclei at the level of lexical coherence and statistical figures, as shown in Figure 4.
The thematic study of the items obtained showed a high thematic coherence, this metric is based on measures of how frequently individual words occur and pairs of distinct words co-occur (Kuhn, 2018). Values close to 0 indicate optimal figures and consequently increase cohesion in the topics (Mimno et al. 2011). Table III shows the 7 topics detected. WordStat uses an algorithm to automatically assign a label to each group, as well as the main keywords associated with that topic in descending order according to the cutoff criteria (in this case, minimum frequency, 4); the total frequency of the main keywords of the thematic core, the number of cases or documents that contain at least one of the keywords and their percentage.
As can be seen, the thematic core Field of Entrepreneurship, how could it be otherwise, is the most prominent, has a coherence of 0.388; a total frequency of 98; appears in the objectives of 39 of the 40 documents, that is, in 97.5% of cases. The rest of the thematic cores also show very positive values, for example Based Citation, Entrepreneurship Research and Evolution Studies. It is noteworthy that two particular thematic nuclei, Region Similarities and Convergence Cohesion, are appearing in 10 and 6 documents respectively.

Figure 4. Topic Groups in Objectives
The analysis of the main motivations for carrying out this type of research by looking at proposed objectives reveals that 1996 was the year when the first global work (Dery and Toulouse 1996) of this type was elaborated to "reveal the social structuration of knowledge in entrepreneurship". The objectives have changed over time and show unique characteristics that are not typical of bibliometric research in general; e.g. namely to prove the legitimacy of the discipline. Observing the frequency with which authors employ words, and the topical groupings by means of factor analyses carried out using text mining software, a group of significant documents appears which are based on a solid theoretical foundation and recur to bibliometrics in order to expose cohesive and converging features in the discipline (Busenitz et al. 2003;Campos et al. 2012;Cornelius et al. 2006;Gregoire et al. 2006;Reader and Watkins 2006;Schildt et al. 2006).
The rest of the topic groups that emerge are more common and tie in with the need described by Low and Macmillan (1988), "a body of literature develops, it is useful to stop occasionally, take inventory of the work that has been done, and identify new directions and challenges for the future". They use bibliometrics to compile the most fundamental works and authors, and to show their evolution as well as their social structure to improve understanding of them, and to make advances in their theoretical construct.

CONCLUSIONS
The procedure followed with the limitations and conclusions is similar to that described in the previous section. Only the thematic nuclei have been extended to 8, since the texts that include them are usually more extensive at the end of the documents. The analysis of the conclusions of the sample documents shows different topic groups: Category Management, Program top, Significant article, Concepts Strong, Appears Identified, Innovation Related, Entrepreneurship Research and Core Themes ( Figure 5). Table IV shows the main statistics related to the conclusions and the main thematic associations. Three of them appear in a greater number of documents: "Appears Identified": It is linked to obtaining and identifying main trends within the field of entrepreneurship, presents high frequencies of the keywords contained, appears in 36 documents and shows high cohesion. "Entrepreneurship Research": It is a group related to the objectives of the documents, it also appears in 36 articles and keywords emerge related to the increasing disciplinary cohesion and converging nuclei.
"Category Management": reflects the idea that most of the bibliometric research coincides in signalling that entrepreneurship is a discipline with a markedly multidisciplinary character whose essence lies in other main fields or categories (Management, Business and Economics). The words that configure it are present in 32 of the 40 documents in the sample with high frequency and cohesion.
The rest of thematic associations (program top, concepts strong, innovation related, core themes, significant article), although they decrease in the number of cases and frequency (even if they are high) deepen conclusions related to greater internal theoretical strength related to innovation and the emergence of nuclei of recognizable authors linked to the strong growth of the field of entrepreneurship.

Limitations
As for the limitations offered by the authors of the sample set, a significant number of documents do not indicate them expressly. In 12 of them no limitations are mentioned. That represents 30% of the total and is due to the fact that some of these works were preliminary presentations at conferences.
There are 8 interconnected topical nuclei as can be seen in Figure 6. The one that displays the greatest cohesion and frequency (Evolving SSCI) has to do with limitations with respect to the coverage of the sources and databases used in the analyses, as well as the inclusion or lack of it of certain document typologies such as books or proceeding papers. Table V shows the statistics of the following thematic groups, their evolution related to the words used by authors, and reflects problems associated with the limitations of bibliometrics as a methodology and, in second place, those inherent to research in entrepreneurship, e.g., the multidisciplinary essence of the discipline.
Two topic groups (limitation contribution; nature subjective) show how difficult it is to decipher the results obtained and how subjective they are. This illustrates the need to possess prior understanding of entrepreneurship as well as bibliometrics in order to be able to interpret them. Most works were elaborated by authors, who come from a background in entrepreneurship research. Exploring the union of the two knowledge areas through the collaboration of authors from both fields might contribute to minimising possible biases and offer a more realistic image of the discipline by minimising errors in interpretation.
Other limitations are linked to the static nature of results in contrast with the dynamic structure of a field in constant expansion, or to the measurements used; in this case the number of citations as an exclusive measure, disregarding a complementary analysis of content.
The analysis of the main technical aspects revealed: • Most of the bibliometric studies on entrepreneurship analyse long periods of time, without any relationship between the number of documents analysed and the selected period of time.
• The main technical problem of this type of analysis is to find a set of documents representative of the discipline. The favourite options have been to recover the ISIWoS data or choose a set of journals as representative of the area. There are no optimal search strategies, or sets of representative terms to perform them, beyond the use of the root entrepr * or the arbitrary combination of keywords.
• The favourite unit of analysis has been the article, leaving aside too many other types of important typologies in the discipline such as books or manuals.
• The main bibliometric analyses carried out are based on the citation as the only quality measure. The time needed for the citations to accumulate or the possible disciplinary differences between documents from different areas of study have not been taken into account. Bibliographic coupling focused on references and content cowords analysis have hardly been used.
• Regarding the use of bibliometric software, a large number of studies do not indicate whether they use it.
The analysis of the objectives, conclusions and limitations of this type of research showed: • Some characteristic objectives such as the search for cohesion and convergence patterns to strengthen the legitimacy of the discipline.
• Conclusions that expose the marked multidisciplinary nature of entrepreneurship.
• Limitations linked precisely to the multidisciplinary nature of entrepreneurship and associated with the bibliometric methodology such as the static nature of most analyses or the difficulty in interpreting the results.
The results obtained in the analysis of the selected documents offer a solid base for a better understanding of a research field that presents enormous difficulties on account of its multidisciplinary nature. The application of bibliometric methods is showing great potential for a quantitative confirmation of pre-supposed ideas associated with its structure and growth. The literature review that has been carried out shows only the tip of an iceberg when it comes to the possibilities that bibliometrics offer for analysis. Researchers who published some of the most influential works in entrepreneurship (Busenitz et al. 2003;Cornelius et al. 2006;Gregoire et al. 2006;Schildt et al. 2006) in a quest to find an answer to the question of legitimacy have defined the search for patterns of cohesion and convergence as a key objective. The review also confirms that studies of this type have contributed significantly  Moreover, a series of problems and gaps have been identified which need to be addressed in the future. Some of the most significant ones are: • The need to incorporate a bibliometric focus in this type of analysis, taking into consideration the recommendations made in the Declaration on Research Assessment (DORA) and the Leiden Manifesto (Hicks et al. 2015); especially those that make reference to the differences in publication and citation practices between scientific fields. It seems imperative to take into account the time citations require to accumulate and to consider the multidisciplinary nature of the discipline, looking at the normalisation of citations as a possible solution (Bornmann and Wohlrabe 2017; Waltman and van Eck 2013).
• The challenge of selecting significant document samples to carry out the different analyses must be explored in greater depth. Arbitrary criteria and strategies aimed at producing the greatest number of documents possible are generally used even though not all works have contributed equally to the discipline. There is also no relationship between the number of sample documents and the time period under study.
• There is an almost exclusive dependence on citations as the only representative or qualitative reference to a document that belongs to the area or discipline. It would be necessary to explore the possibility of applying other types of indicators, or even to work on elaborating indicators that are specific to the field. The H-Index, one of biggest milestones in bibliometrics, has hardly found application. The sample does not contain any documents (not even among the most recent) using metrics linked to the social development of science and the new information platforms like user metrics and altmetrics.
• The references the main documents contain and are used in the different analyses have hardly been studied, and the article has almost always been the main unit of analysis. On too many occasions, other document categories such as books or textbooks, which are of great importance to the discipline, as explained in Landstrom et al. (2012), have been left out.
• The bibliometric methods that use a quantitative approach have the potential to improve systematic review processes. They aim to provide transparency and offer reproducible and replicable results. However, a significant number of documents among the sample do not indicate whether a bibliometric software was used or what limitations they had to deal with.
This article largely confirms some of the conclusions presented by Zupic and Cater (2015) in Management and Organization, like the need to use new bibliometric methods which are based more on content, so as to obtain more accurate groups as defined by semantic similarities between documents, for instance. It is also necessary to employ less exploited types of analysis such as bibliographic coupling, co-word analyses and hybrid methods as well as the combination and comparison of results obtained when using different methodologies. Despite all of these problems, the review provides indications of collaboration between the two fields of knowledge aimed at resolving them, as in the case of Meyer et al. (2014). A certain degree of specialisation can also be appreciated. Hans Landstrom is the most outstanding example of this.
The article is not without its limitations, which are mostly due to aspects of methodology. Firstly, the search for the sample literature might have failed to pick up and include every relevant document in existence despite including three different databases in order to widen coverage. Secondly, the review includes subjective components, which could lead to a bias in the results. They are the result of inclusion/exclusion criteria used on the documents which make up the final sample. Thirdly, these subjective components extend into the parameters used to perform the content analysis of the documents. A different configuration might have led to a different interpretation. However, we believe that the findings presented are sufficiently significant to help obtain a better understanding of the discipline and, more importantly, they could be helpful in the quest to introduce more rigour to future bibliometric analyses.
Finally, a promising future can be foreseen for the relationship between bibliometrics and entrepreneurship. It is a special type of research, which needs to incorporate the latest theories and advances emerging in both fields in order to stay up to date. Two suggestions can be made regarding future lines of investigation: the scope of this review ought to be widened by including the remaining bibliometric studies in entrepreneurship in order to verify the results obtained, and efforts should be made to better understand whether the use of this type of study merely serves to provide new bibliometric research, or if it is actually instrumental in obtaining a greater understanding of the discipline.

NOTES
1. In a bibliometric work it is convenient to separate those articles or documents that have been useful for the writing of the research, from those others that make up the sample and that are available. The whole list is included in supplementary material Annex 2 (although some of the documents appears both in the sample and in the references).  Articles "family business*" or "family firm*" or "family own*" or "family control*" AND "entrepreneur*" or "venture*" "Editorship of journals, the publication of scholarly books, the sponsorship of research conferences, and the training of doctoral students or other activities. These contributions are excluded from this analysis. A second limitation is that this study measured contribution as the quantity and quality of articles, rather than the content of those publications.

A third limitation of this study is that it is static. This paper measured the impact of scholars and institutions on entrepreneurship research at one moment in time. A fourth limitation is that the results of this study may not be predictive. Shifts in institutional affiliations of scholars can alter institutional rankings quickly."
3 "Trends and Growth Points in the field" "The use of textual analysis software does allow clustering which, by and large, seems to accord with the expectations of those in the field" Not indicated.

4
Legitimacy: "How is entrepreneurship emerging? Are entrepreneurship scholars obtaining increased legitimacy? Where should research be directed to build the field?" "We find that the boundaries of the entrepreneurship field continue to be highly permeable. Accumulated fragmentalism Evidence of a growing internal culture and knowledge base, and thus a growing level of exchange internal to the entrepreneurship community" Not indicated.

5
"Get an overview of research in Entrepreneurship" "Identify and analyze the relationships between the documents that have had the greatest impact for the construction of the knowledge base of the discipline" "Axes of convergence: 1) the study of entrepreneurial behavior in existing organizations and their relation to the performance of the organization, also known in the academic field as" corporate entrepreneurship "; 2) the sociocultural or institutional approach and, predominantly, within this one, the study of the influence of belonging to certain ethnic groups on the creation of companies known under the theory of marginalization, 3) the psychological traits approach or identification of the psychological factors of successful entrepreneurs and 4) the economic approach to explain the entrepreneur's role in economic growth and development. " "Number of cites, it is impossible to distinguish the intention with which they were made.

The interpretation of the factors and graphs obtained is subjective"
The criterion of selection of the citing sample and the division of the time horizon of analysis in three subperiods.

Article Main Research items Main Conclusions Main Limitations
6 "There is no widely accepted categorization of different streams of entrepreneurship research, and it is not even clear if distinct streams exist." "In addition, a considerable diversity in the field across countries has been noted, but there is little systematic knowledge regarding country or continent specific differences in entrepreneurship research" "We identify and describe the 15 most cited dense groups representing the most central theoretical streams." "Our findings reveal that collaboration across universities tends to be relatively modest, although the level of co-operation varies greatly." "15 Groups of references that were most commonly cited by entrepreneurship articles. There is considerable amount of more recent literature that is making a significant impact on the field. Is difficult to define the group of articles constituting "entrepreneurship" and it could be questioned whether they all belong to the field of entrepreneurship. People cite articles with varying purposes, and therefore the popularity of the groups does not necessarily represent their importance to theoretical argumentation or empirical."

7
"In order to determine the stage of maturation of the field of entrepreneurship" "Determined whether researchers have provided the foundation for systematic disciplinary advance" "Entrepreneurship research has been increasingly self-reflective" "The increasing complexity of the research in entrepreneurship alone indicates a greater maturity in the discipline. "Only one research area was imputed to each top cited author, which narrows down the academic scope of the researchers. Which provides a static report of entrepreneurship. Also, the subjective nature of the key element, "informal communication relations", that underlies the concept of the invisible colleges raises some concern." 13 "The aim of this research is to gain insights into our research behavior. The paper follows the argument by Low and MacMillan (1988)  "A group of core knowledge producers seem to emerge over time.
Still the field relies on old theoretical frameworks imported from mainstream disciplines. However, over the last decade sign could be seen of a stronger knowledgebase of its own in entrepreneurship research is emerging. Our analysis of the knowledge users in entrepreneurship research shows that the field is heavily anchored in "business" and "management". On the other hand, the core works in entrepreneurship are included in a large number of studies within many different fields of researchcreating a "long tail" of users…" "We have to bear in mind that bibliometric analysis is based on the assumption that research is essentially cumulative -new research is built on and cites earlier high quality foundations -i.e. a "normal science approach" (Kuhn, 1970), but we know that this is not the only way to communicate and organize research, particularly in new and evolving fields, for example, fields that are organized and communicated through "negotiations" between actors (Knorr Cetina, 1999;Åström and Sándor, 2009

26
"This paper seeks to map out the emergence and evolution of entrepreneurship as an independent field in the social science literatura." "Our analysis indicates that entrepreneurship has grown steadily during the 1990's but has truly emerged as a legitimate academic discipline in the latter part of the 00's. The field has been dominated by researchers from Anglo-Saxon countries over the past twenty years, with particularly strong representations from the US, UK, and Canada. The results from our structural analysis, which is based on a core document approach, point to five large knowledge clusters and further 16 sub-clusters. We characterize the clusters from their cognitive structure and assess the strength of the relationships between these clusters." Not indicated.

27
"An effort to gauge trends in and contributions to the broad field of ''entrepreneur/ entrepreneurship,'' "We offer evidence that in the process of internationalization of entrepreneurship field, knowledge diffusion has contributed substantially to homogeneity in all the examined regions as common interests in certain research topics can be identified. It should be pointed out that most of these common focuses tend to be theoretical-driven topics Differences in contexts slowed the move towards convergence and enriched entrepreneurship knowledge." "However, bibliometric analysis is not without limitations. For example, we have to bear in mind that bibliometric analysis is based on the assumption that research is essentially cumulative -new research is built on and cites earlier high quality foundations -i.e. a "normal science approach" (Kuhn, 1970). However, we know that this is not the only way to communicate and organize research, particularly in new and evolving fields (Knorr Cetina, 1999). In addition, there are concerns about the databases typically used for bibliometric analysis (Watkins, 2005). Although the SSCI database is a wonderful resource for citation analysis, it has some limitations with regards, for example, the database consists primarily of scholarly journals (less of books and conference papers), and the coverage of journals varies greatly depending on the research field, the language and origin of the publication, and the age of the journals. Thus, citation databases such as SSCI have limitations when it comes to relatively new and evolving research fields such as entrepreneurship."

29
"This study aim at entrepreneurship research dynamics in 1992-2013." "The results conclude four issues and nineteen sub-themes these issues included as entrepreneur, innovative, corporative, and business operations these is core issues of entrepreneurship." Not indicated.

30
"The increasing internationalization of the field also raises three major questions: How has the field of entrepreneurship developed in different regions such as the USA, Europe and not least China? What are the similarities and differences in the development process in different regions? And what are the reasons for these similarities and differences?" "It appears that the development of entrepreneurship as a research field in China has followed a different path compared to the USA and Europe, where "contextual force" was the main driver in the early stage, but during the development process the external influence became weaker and that of "internal force" becomes stronger. In China, the main driver of entrepreneurship research is "internal force" while the "contextual force" has been downplayed. Similarities and differences in the development process across regions have also been identified." "Results provide evidence of the increasing interest in entrepreneurship as a field of study, but also of its interdisciplinary nature, with infusions of concepts and theories from a wide array of management disciplines." "ISI Web of Knowledge but while ISI is a good resource, it comprises only a small subset of all existing journals and leaves out other source documents such as books and dissertations. We only included a subset of all journals in ISI, which further limits the scope of the analysis especially in an emerging field such as entrepreneurship. An additional limitation is that ISI includes almost exclusively articles written in English which may generate some bias. Other limitation pertains to the use of citation and co-citation data. Relying on citation and co-citation data is well established in bibliometric studies to scrutinize the intellectual structure and knowledge base of a field, but it may tend to favor older, more established, works over new contributions. Some older works have gained the status of "mandatory" references and may be cited for ceremonial reasons. Co-citation metrics are used to infer conceptual proximity but analyzing the ties says little about the context."

33
"Carries out a comprehensive and systematic review of academic research on entrepreneurship in family firms applying bibliometric indicators.Review the literature published…" "Is a relatively new area of study We have identified two periods: the first (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002) with low output and a second (2003present) of clear growth, coinciding with the start of the corporate entrepreneurship cluster in the field of entrepreneurship. The analysis verifies compliance with Lotka's Law, which means that there is a higher concentration of items in few productive authors compared with other disciplines. The most productive authors and journals do not necessarily coincide with those most cited. The most notable result in this sense is the fact that this field is highly interconnected with high co-citation between authors. The field is structured around widely developed themes-Risk Taking and Entrepreneurship-and underdeveloped peripheral themes-Gender, Governance and Family Firm-without clusters in either peripheral or emerging quadrants." Not indicated.