Which variables predict the internationalization type of academic spin-offs?

Abstract

This study aims to find out which variables predict the internationalization type of academic spin-offs. This topic has not yet received any attention in academic spin-offs’ internationalization literature. We use a sample of Spanish spin-offs, with four dependent variables reflecting internationalization types and a large selection of independent variables reflecting various domains considered in the extant literature. Logistic regression is applied to outline marginal effects and prediction accuracies. The results show that academic spin-offs which are oriented towards international markets from the beginning of their activities and constantly pursue new international opportunities are more likely to internationalize fast and remain international for a longer time period than other firms. Being the first to introduce product or process innovations and co-operating with international government institutions, competitors, customers, and/or suppliers is also relatively useful for such internationalizers. The study also demonstrates that predicting if an academic spin-off will become a less active internationalizer results in a higher accuracy than forecasting if it will become a born global. Moreover, predicting if a firm will internationalize later results in a higher accuracy than forecasting if this will happen during the first three years since the spin-off’s foundation.

Highlights

• •We found out which variables predict academic spin-offs’ internationalization type.
• •The study was based on a sample of Spanish spin-offs.
• •We used logistic regression to outline marginal effects and prediction accuracies.
• •Innovative and networked firms with an international mindset internationalize fast.
• •Prediction accuracy is higher for late and less active internationalizers.

1. Introduction

Firms’ internationalization has been actively studied since the 1960s and early internationalization especially since the 1990s (for some recent overviews, see Refs. [1,2,5,34]) but the literature is still fragmented and more quantitative studies on different firms’ internationalization strategies [6] and international performance are needed [7]. As academic spin-offs’ internationalization has received less attention, there are still several under-explored areas in the literature [8,9]: for instance, these firms’ internationalization patterns should be studied more as they do not all enter foreign markets at a similar speed [10].

Studying academic spin-offs is important as they actively contribute to their host countries’ economic development via scientific knowledge transfer [11], promoting innovation [12,13] and internationalization [14], creating jobs [15] and strengthening links between universities and local firms [16]. Moreover, according to some studies, export activities have a positive impact on academic spin-offs’ survival [17–19], and internationalization is the main way for commercializing their research results [20] as due to focusing on narrow market niches, the home market might be too small for them [10].

Despite the abundance of studies on different internationalization-related aspects, so far, only a limited number of authors have tried to predict firms’ internationalization types (for instance, if they will internationalize fast or slowly or become born globals; for an overview, see Ref. [21]), and “there are inconclusive findings in the literature on key predictors of export behaviour” [ [22], p. 33]. Moreover, although according to some authors, academic spin-offs’ internationalization can differ from other firms’ foreign expansion [8,23,24] and it is very important to find out why some academic spin-offs achieve better performance than others [25], export prediction has not received any attention yet in studies on academic spin-offs.

For modeling certain phenomena (i.e., dependent variables), both explanatory and predictive approaches have been used actively. The central focus in predictive studies is on achieving the maximum possible forecasting accuracy [26]. In some areas of business research – e.g., for finding out which financial factors associate with firm bankruptcy – predictive modeling has flourished during the past decades (see, e.g., Ref. [27]) and hundreds of papers have been published by now. In some other contexts, including international business domains, such articles are scarce [21], which might be caused by difficulties in identifying necessary (high accuracy) predictors.

Based on the literature, three research questions/gaps can be identified. First, which variables predict the internationalization of academic spin-offs? Second, are all variables related to academic spin-offs equally predictive? Third, which pattern of academic spin-offs’ internationalization can be predicted with a higher accuracy? To address these questions/gaps, the aim of the paper is to find out which variables predict the internationalization type of Spanish academic spin-offs. Spain’s exports have not achieved full potential yet due to firms’ resource constraints and relatively low innovativeness [17,28]. Moreover, although since 2001, more efforts have been made in Spain to create academic spin-offs [11], the share of born globals and other fast internationalizers is still low among these firms [18] as Spanish academic networks’ capacity to support spin-offs’ internationalization is relatively limited [29]. Thus, finding out how to predict and increase Spanish academic spin-offs’ exports will help to develop some useful theoretical, managerial and policy implications that can be also applied to countries with similar conditions.

The paper starts from a review of the literature on academic spin-offs’ internationalization and, thereafter, about variables various authors have used for predicting non-academic firms’ internationalization types or statuses. After the methodology section, survey data of 128 Spanish academic spin-offs will be analyzed to determine the accuracy of different variables for predicting which firms will achieve at least some internationalization in three years or later and which firms will become born globals in three years or achieve some born global criteria (export share, number and location of foreign markets) later. The paper ends with conclusions and managerial, policy and research implications.

2. Literature review

2.1. Literature on academic spin-offs’ internationalization

Many academic spin-offs remain small and home-country focused due to lacking entrepreneurial growth intentions: this can be caused by risk aversion and excessive focus on technology development or on academic tasks [30]. In addition to facing resource constraints [31] and high uncertainty [32], such firms often lack international business experience [33] and entrepreneurial skills [29].

Despite some firms’ lack of interest, resources and/or capabilities for entering foreign markets, several authors have studied academic spin-offs’ internationalization as such firms’ founders often have international network relationships due to academic ties abroad, and such contacts can be useful for initial and subsequent foreign market entries [8,12,34]. Expanding to foreign markets helps them to grow faster [35], and it also sometimes gives an impulse for developing new products [9]. Moreover, internationalization can be necessary for spin-offs with substantial research and development costs, especially if they focus on niche customers [10] or if their home market is very small [36]. An overall framework of academic spin-offs’ internationalization is presented on Fig. 1. It is based on the literature presented in Appendixes 1 and 2.

Fig. 1.

A framework of academic spin-offs’ internationalization.

An overview of 32 studies on academic spin-offs’ internationalization is provided in Appendix 1 and they are summarized in Table 1. It is evident that authors have used various quantitative and qualitative methods and focused on different topics – for instance, the role of academic entrepreneurs’ international or local academic or business experience, attitudes and network relationships in internationalization or the impact of internationalization on achieving high turnover growth. However, they have not tried to find out if it is possible to predict these firms’ internationalization types. Only [37] analyzed academics’ intentions to become entrepreneurs and internationalize, but they did not find out if this happened or not.

Table 1.

An overview of methodological and theme variety of the literature on academic spin-offs’ internationalization.

Characteristic	Number of studies
Method
quantitative	18
qualitative	12
conceptual	2
Theme
factors fostering academic spin-offs’ (fast and/or successful) internationalization	19
academic spin-offs’ internationalization processes	10
scholars’ involvement in or plans to become involved in internationalized academic spin-offs	6
internationalized (versus local) academic spin-offs’ (financial) performance	6
differences and similarities between internationalized (and local) academic spin-offs	4
impediments to academic spin-offs’ (fast) internationalization and ways for overcoming them	3
universities’ role in academic spin-offs’ internationalization	3
reasons for academic spin-offs’ (fast) internationalization	3
under-researched topics in the literature on academic spin-offs’ internationalization	1

Source: based on Appendix 1.

There is evidence that some academic spin-offs internationalize very fast, while some experience slower and/or more gradual internationalization [13,38]. Moreover, for some firms, internationalization is only a sporadic event [34] and for some, initial fast growth does not continue [36,39]. Thus, it is important to study whether any specific predictors help to find out which academic spin-offs will internationalize fast or slowly and which remain highly international during a longer time period.

2.2. Literature on predicting firms’ internationalization types or statuses

As the literature on academic spin-offs’ internationalization does not focus on predicting firms’ internationalization types or statuses, it was necessary to investigate which variables have been used in the literature on non-academic firms’ internationalization. Thus, Table 2 gives an overview of the variables different authors have used for predicting if firms will become exporters or foreign investors or will not internationalize at all, if they fill internationalize fast or not or if they will achieve considerable or modest international involvement. A more detailed overview of these variables is provided in Appendix 2.

Table 2.

An overview of variables used for predicting firms’ internationalization types or statuses.

Variables	Number of studies
Firm characteristics
attitudes	16
decision-makers’ or the firm’s prior export or import knowledge	14
availability and efficiency of resources	14
firm size	12
products’ competitiveness	10
other production characteristics	10
other skills and/or abilities	9
innovativeness	8
Other variables
home market characteristics	12
foreign markets’ characteristics	10
(perceived) export (market) risk and/or export barriers	10
foreign market information availability and diversity	7
technological or other co-operation	6
network’s international characteristics	5

Source: based on Appendix 2.

Table 2 and Appendix 2 show that most authors have used some firm characteristics – for instance, size, resources, innovativeness, competitiveness, experience, skills and/or attitudes – while some have focused on network characteristics or the business environment of their home or host markets. Some authors have used only a few variables while some have selected a large number of them; however, none have covered all dimensions. Thus, based on the above, it can be concluded similarly to [22, p. 34] that “export propensity does not depend on single attributes but instead, on specific combinations of attributes”. However, there is no consensus regarding which predictors should be used or which methods should be chosen [21].

The literature review presented in sections 2 and 2.2 shows that compared to other firms’ international activities, academic spin-offs’ internationalization has still received relatively scant attention, but, at the same time, there is a considerable methodological and theme variety. As firms’ internationalization types or statuses have not been predicted in the academic spin-offs’ internationalization literature, an overview of wider internationalization literature was provided.

Table 2 and Appendix 2 demonstrated that the range of applied predictors is extensive, although all of them are never applied simultaneously: while most authors focused on firm-related characteristics, several others also used network-, home- or host market related variables. Still, consensus regarding which predictors should be used for certain research topics is missing. Thus, the following analysis tried to take both spin-off-related and internationalization-related aspects into account. The selection of variables is explained in the Methodology section.

3. Methodology

3.1. General methodological considerations

It was evident from Table 1 and Appendix 1 that academic spin-offs’ internationalization has been studied both quantitatively (for instance, by Refs. [8,9,12,31,39]) and qualitatively (for example, by Refs. [33,36,[40], [41], [42]]). While the former authors used survey data or data from various databases, the latter conducted case studies. This paper is based on survey data as this provides larger generalizability of the results than case study data [43].

Based on an extensive literature review, key variables were identified and a preliminary questionnaire was designed. In 2018, it was pre-tested through three in-depth interviews with ASOs’ managers and academic entrepreneurs. Based on their recommendations, the final version of the questionnaire was developed.

3.2. Data collection and sample

The population included all academic spin-offs (ASOs) established by Spanish public and private universities in 2003–2018. To create the ASOs database, during the first quarter of 2019, a formal request for collaboration was sent to Spanish Universities’ Research Results Transfer Offices Network (RedOTRI). They contacted 70 managers of technology transfer offices (TTOs) included in this network. TTO managers provided the following information of ASOs created with their support: company name, foundation year, industry, phone number, email address, website. Also, they provided information about the founders: name and research area.

Based on this information, a database of 628 ASOs from 51 Spanish universities was created. In April–May 2019, a questionnaire was sent by e-mail to the founders of all ASOs in the database. Non-respondents were contacted by phone. In May–December 2019, responses from 173 ASOs (response rate: 27.6%) were obtained. Non-response bias was analyzed: a t-test was applied on respondent and non-respondent ASOs’ age and size (number of employees). There were no mean differences between these firms (p = 0.139 and p = 0.310, respectively). Therefore, the sample can be considered to represent the total population of Spanish ASOs well. For the analysis, the data from 128 firms were used as these respondents had provided information about all dependent and independent variables. The sampling error of studying 128 ASOs out of the initial 628 is 7.74% with a statistical confidence level of 95%.

The mean and median firm age was eight years at the time of answering the questions. Therefore, the majority of the sample reflects adolescent companies. The latter is important as in case of studying old entities, several of the questions might not have effectively portrayed the early stages in firms’ lifecycle. Therefore, the probability that firms’ answers to questions about strategies would substantially vary throughout their lifecycle is low.

3.3. Dependent variables

The dependent variables of this study reflect each firm’s internationalization status during the first three years of activities and at the time of responding to the survey (on an average, eight years later). For the first time span, two different internationalization definitions were applied. First, just being active on at least one foreign market without setting additional restrictions on the market’s location and export share was used. Thus, we could call such firms “early internationalizers” as a three-year time span has been often used for distinguishing between early/faster and late/slower internationalizers [6,44,45]. Second, following [21,46], we adhered to the criteria characteristic to the born global definition, requiring for three foreign markets (including at least one outside Europe) and export share of at least 25%. For the second time span, we applied similar criteria: just being active on at least one foreign market for the first group, and, for the second group, exporting to at least three markets (including at least one outside Europe) and having an export share of 25% or more. Thus, we distinguished between less and more active/committed internationalizers [6,45,47].

Based on the above criteria, four different dependent variables were used accounting for the scope of foreign activities and their timelines. This helps to obtain a diverse overview of the studied phenomenon and avoid single definition bias. These binary (1 meaning a match with the respective definition and 0 otherwise) dependent variables have been coded as follows: a) at least some internationalization in three years as INT3, b) at least some internationalization today as INTT, c) internationalization commitment matching the born global definition in 3 years as BG3, d) internationalization commitment matching born global criteria (export share, number and location of foreign markets) today as BGT.

3.4. Independent variables

Based on the literature review (both the academic spin-off literature covered in Table 1 and Appendix 1 and export prediction literature reviewed in Table 2 and Appendix 2), it was evident that several themes should be covered: firms’ contribution to society [11,[13], [14], [15], [16]], industry characteristics [48,49], technology [13,31,50], firms’ attitudes toward internationalization [37,40,51], managers’ experience and training [29,31,52] and network relationships [52–54]. Although several export prediction studies (for instance, Refs. [55–59]) also used variables related to home market characteristics – for example, business environment and level of economic development – it was not reasonable to include them as this study was based on data from only one country and thus, all firms faced relatively similar conditions. Host market characteristics [60–65] were also excluded as for high-tech and niche products, customers’ tastes are more homogeneous: thus, firms can export successfully without making (considerable) host country-specific adaptations [66].

Based on the above considerations, the initial set of questions encompassed the following thematic blocks (with the number of questions under each block in brackets): a) social demand and social innovations (6), b) industry characteristics (7), c) firm’s technology (4), d) international orientation (4), e) openness to new international opportunities (5), f) experience and training of management (7), g) management team’s diversity (4), co-operation with different national (8) and international (8) entities. The questions about demand and social innovations and some concerning co-operation were somewhat similar to the ones used by Ref. [67], some others regarding contacts originated from Ref. [68], while the ones regarding industrial dynamics were based on [48,69]. The questions about international orientation were developed following [70,71], while the ones regarding openness to new international opportunities were adapted from Refs. [40,42]. Finally, the questions about technological superiority were taken from Ref. [56] and the ones about experience were developed based on [31].

Out of those initial 53 questions, the selection of suitable ones for further analysis was conducted with robust Welch’s ANOVA. Namely, in case the question indicated a significant difference in the mean values for the groups (i.e., (non-)internationalized firm coded with 1 and 0) of at least one of the four dependent variables, it was included in the further analysis. Such pre-selection of variables is a usual strategy in prediction studies, as variables with insignificant differences in means for the predicted groups are unlikely to be useful predictors [72]. The pre-selection of questions with robust ANOVA resulted in 27 remaining questions, reducing their amount about by half. The final list of independent variables is provided in Table 3.

Table 3.

List of independent variables.

Block	Variable	Statement
1 (social demand & innovation)	1a_Soc	We collaborate with society to identify social demands
	1b_Inn	Our social innovations have a high degree of internationalization
2 (industry characteristics)	2a_Ent	[in our main industry] The ratio of entry, exit and bankruptcy of companies is high
3 (firm’s technology)	3a_Van	We are at the technological forefront of our industry
	3b_Inv	We have invented much of the technology that goes into our products
	3c_Pio	With respect to competitors, we are often the first to introduce product or process innovations
	3d_Sup	We are internationally recognized for our technological superiority
4 (international orientation)	4a_Vis	The management team tends to see the world as a single market
	4b_Exp	The organizational culture of the company is conductive to exploring new opportunities abroad
	4c_Tri	The management team often communicates to employees that the company’s mission is to succeed internationally
	4d_Rec	The management team uses human and other resources to achieve objectives in international markets
5 (openness to foreign opportunities)	5a_Per	The company actively pursues new international opportunities as they arise
	5b_Rec	The company commits resources to exploit new international opportunities
	5c_Opo	The company pursues new international opportunities irrespective of resources
	5d_Eva	The company evaluates new international opportunities as they arise
	5e_Exp	The company believes that exploiting new international opportunities as they arise is important for growth
6 (managerial experience)	6a_Exp_i	[management team’s] previous experience of working in international markets
	6b_Exp_s	[management team’s] previous experience in the main sector in which our company operates in international markets
	6c_For	[management team’s] training received abroad (doctorate, master’s degree, stays abroad …)
	6d_Neg	[management team’s] international business training
7 (co-operation)	7a_int	[contacts with] international university institutions (incubators, other academics, university chairs, research centers, etc.)
	7b_nac	[contacts with] national research results transfer offices
	7c_nac	[contacts with] other national universities
	7d_int	[contacts with] other international universities
	7e_int	[contacts with] international government institutions
	7f_int	[contacts with] international industry agents (competitors, customers, suppliers)
	7g_int	[contacts with] other international investors

3.5. Data analysis settings

As the study is focused on identifying useful predictors and determining their prediction accuracy, the following empirical strategy has been implemented. Ideally, when the applied variables would be uncorrelated, they could be used in a single logistic regression analysis, which suits for all four dependent variables as they are binary. This was not the case, as many variables indicated very high and significant correlations, being quite usual for Likert-scale questionnaires, and therefore having a serious effect on regression estimates (see Ref. [73]). Thus, a widely applied method of outlining marginal effects of variables from logistic regression was implemented, but separate logistic regressions were run with each of the questions for each of the dependent variables to obtain unbiased estimates. This resulted in 27 × 4 = 108 unique logistic regressions.

The marginal effect indicates the change in a dependent variable (probability to be an internationalizer), in case an equal marginal increase in the value of each independent variable occurs [74]. The latter enabled us to outline the “pure” effect of each variable on the probability to be an internationalized firm, being not affected by the correlations between the independent variables. As marginal effects can be negative or positive (depending on how the underlying question is posited), we used the absolute values of obtained marginal effects to rank them by usefulness from 1 to 27 for each of the four dependent variables.

In the next step, we composed different logistic regression prediction models, first by using only questions from a specific block out of seven, and second by including all 27 variables for each of the dependent variables (see Table 3). Thus, in total 8 × 4 = 32 prediction accuracies are provided. We used the total accuracy from the classification matrix (i.e., the sum of true positives and negatives divided by the studied population) as the measure of prediction goodness, which is a usual measure in the literature [75]. In prediction studies, the classical performance measures of logistic regression like pseudo R² that are more characteristic to explanatory modeling [26] are not applied, while they have obvious positive correlations with the accuracy.

4. Results and discussion

4.1. Characteristics of the studied dataset

The dataset of 128 firms breaks down as follows: INT3 = 81 firms, INTT = 93 firms, BG3 = 32 firms and BGT = 58 firms. The same company can belong to different categories: all BGTs are INTTs and all BG3s are also INT3s. Moreover, while some companies de-internationalized before responding to the survey and thus, did not become BGTs or INTTs, some others internationalized in more than three years and never became BG3s or INT3s despite getting the status of a BGT and/or INTT. Still, firms tend to sustain their statuses, namely 95% of firms with INT3 = 1 are also INTT = 1 and this indicator is 81% in case of BG3 and BGT. From the rest of the BG3 population, 9% de-internationalized and 10% proceeded to the INTT category with a lower international commitment.

Around 80% of firms are micro entities with their total assets remaining below 2 million euros, while among the rest, most are small firms by the total assets criterion with a fraction in the medium-sized category as well. The median foundation age is 2011, thus, the majority of firms can be considered adolescent. The three largest represented sectors are professional, scientific and technical activities (44 firms), information and communication (39 firms), and manufacturing (21 firms), with the rest of respondents are active in the educational, health, and social sectors or in administrative, auxiliary, and artistic activities.

4.2. Individual importance of predictors depicted with marginal effects

Table 4 presents the marginal effects of the individual logistic regression prediction models for all 27 variables. It shows that for predicting internationalization, questions regarding firms’ international orientation are the most useful in terms of providing 4 out of the 6 best predictors in the total ranking. The questions about new international opportunities are in the second place. In both of these blocks, all the questions matter. Thus, it is very important for academic spin-offs to be oriented towards international markets from the beginning of their activities and to constantly pursue new international opportunities. Several authors have also emphasized the importance of firms’ positive attitudes toward internationalization [7,9,24,37,40,51,76]. Consequently, this result is logical: interested and active firms are more likely to enter their first foreign markets – become born globals or slower internationalizers – and, thereafter, keep expanding to other countries and/or increase their involvement in their existing markets. On the other hand, low commitment to internationalization can slow the process down [24,77,78].

Table 4.

Marginal effects for logistic regression models with dependent variable in the column and the single independent variable in the row.^a

		INT3			INTT			BG3			BGT			Total
Block	Question	dy/dx	p-value	Rank	dy/dx	p-value	Rank	dy/dx	p-value	Rank	dy/dx	p-value	Rank	rank
1 (social demand & innovation)	1a_Soc	−0.023	0.468	26	−0.053	0.072	23	−0.033	0.242	23	−0.095	0.006	18	24–25
	1b_Inn	0.112	0.000	14	0.085	0.001	17	0.020	0.475	25	0.039	0.208	23	21
2 (industry characteristics)	2a_Ent	−0.051	0.128	21	−0.058	0.062	21	−0.064	0.024	19	−0.103	0.004	14	19
3 (firm’s technology)	3a_Van	0.101	0.019	15	0.071	0.049	19	0.038	0.365	22	0.088	0.069	20	20
	3b_Inv	0.079	0.038	19	0.071	0.033	20	0.119	0.003	7	0.110	0.012	13	15
	3c_Pio	0.157	0.002	8	0.100	0.020	13	0.170	0.000	2	0.096	0.065	17	9
	3d_Sup	0.091	0.010	17	0.134	0.000	8	0.066	0.038	18	0.141	0.000	9	11–12
4 (international orientation)	4a_Vis	0.200	0.000	3	0.126	0.000	11	0.134	0.000	5	0.133	0.002	10	6
	4b_Exp	0.199	0.000	5	0.215	0.000	1	0.175	0.000	1	0.299	0.000	1	1
	4c_Tri	0.200	0.000	4	0.147	0.000	6	0.145	0.000	4	0.189	0.000	5	4
	4d_Rec	0.246	0.000	1	0.198	0.000	2	0.126	0.001	6	0.222	0.000	3	3
5 (openness to foreign opportunities)	5a_Per	0.219	0.000	2	0.177	0.000	4	0.154	0.000	3	0.280	0.000	2	2
	5b_Rec	0.139	0.000	11	0.131	0.000	10	0.110	0.000	8	0.205	0.000	4	7
	5c_Opo	0.185	0.000	6	0.185	0.000	3	0.101	0.005	9	0.175	0.000	7	5
	5d_Eva	0.145	0.000	9	0.145	0.000	7	0.030	0.397	24	0.120	0.006	12	11–12
	5e_Exp	0.167	0.000	7	0.134	0.000	9	0.048	0.188	20	0.090	0.034	19	13
6 (managerial experience)	6a_Exp_i	0.071	0.023	20	0.090	0.001	15	0.093	0.001	10	0.097	0.004	16	16
	6b_Exp_s	0.093	0.005	16	0.071	0.014	18	0.078	0.007	13	0.063	0.063	21	17
	6c_For	0.010	0.738	27	−0.023	0.425	27	0.067	0.023	17	−0.004	0.897	27	27
	6d_Neg	0.045	0.157	25	0.057	0.052	22	0.046	0.092	21	0.019	0.548	26	26
7 (co-operation)	7a_int	0.085	0.052	18	0.032	0.394	25	0.070	0.025	16	0.102	0.016	15	18
	7b_nac	−0.048	0.204	24	−0.113	0.001	12	0.012	0.713	27	−0.035	0.381	24	23
	7c_nac	0.051	0.182	22	−0.044	0.159	24	0.071	0.016	15	0.050	0.178	22	22
	7d_int	0.140	0.001	10	0.029	0.398	26	0.087	0.002	11	0.132	0.001	11	14
	7e_int	0.137	0.017	12	0.161	0.005	5	0.079	0.034	12	0.188	0.001	6	8
	7f_int	0.119	0.001	13	0.097	0.002	14	0.073	0.010	14	0.156	0.000	8	10
	7g_int	0.050	0.307	23	0.087	0.103	16	0.013	0.745	26	0.024	0.609	25	24–25

^aFor each dependent variable, the independent variable rank has been determined by ordering the absolute values of marginal effects from largest to smallest. Total rank is obtained by ordering the means of four ranks. The marginal effect has been noted with dy/dx and has been obtained from the Stata statistical package with the “mfx” command, i.e., the marginal effect at the mean value of each variable. As Likert-scale variables are concerned, the alternative command “margins” resulting in the average marginal effect provides very similar results, i.e., the rankings are not altered.

Table 4 also shows that some questions regarding technological novelty (especially, being the first to introduce product or process innovations) and co-operation (mainly, with international government institutions, competitors, customers, and/or suppliers) are also relatively useful predictors but their usefulness varies considerably for some types of internationalizers: the former is especially important for born globals (this was expected as according to Ref. [14], technology-based university spin-offs are more international) while the latter for somewhat slower but still highly committed internationalizers. This was also logical as several authors have emphasized the usefulness of network relationships for academic spin-offs’ fast internationalization [36,54,76,79].

The least useful predictors were questions concerning the management team’s foreign academic degrees and international business training. Several authors have emphasized the importance of these factors for academic spin-offs’ internationalization [29,31,52], but studying abroad does not always result in creating useful business contacts, especially if at that time, future entrepreneurs have not yet decided to establish academic spin-offs. In addition, questions about collaborating with society to identify social demands and contacts with foreign investors were not useful predictors. The former can be explained by the fact that as countries differ considerably, knowing what the Spanish society expects does not guarantee successful foreign sales, while the latter by the fact that vast majority of the respondents did not have any contacts with foreign investors, and some of those who had contacts, had not involved them (yet) as shareholders. While among Israeli high technology start-ups, involving foreign shareholders is frequent [80], this does not seem to be the case of Spanish academic spin-offs.

4.3. Block-specific and total prediction accuracies from the logistic regression

Table 5 indicates what is the prediction accuracy by using specific question blocks for the prediction of each of the four dependent variables. Finally, the total accuracy has also been provided, in case of which all question blocks have been used as predictors simultaneously. Two different types of ranks for goodness of predictors have been provided. The rank in the row after each question block shows which of the dependent variables can be predicted most accurately with the specific question block (1 as the most accurate and 4 as the least accurate prediction). The rank in the column after each dependent variable shows which of the independent variable blocks is the best for predicting the specific dependent variable (1 as the most accurate and 7 as the least accurate predictor).

Table 5.

Prediction accuracies (AC, %) with rankings for each dependent variable by each question block and all blocks combined.^a

	INT3	Rank	INTT	Rank	BG3	Rank	BGT	Rank
1 (social demand & innovation)	73.4	3	78.9	4	75.0	3–4	66.4	5
Rank	3		1		2		4
2 (industry characteristics)	63.3	7	72.7	6	75.0	3–4	58.6	7
Rank	3		2		1		4
3 (firm’s technology)	68.0	5	77.3	5	78.9	2	68.0	4
Rank	3–4		2		1		3–4
4 (international orientation)	81.3	1	83.6	1	70.3	6–7	75.8	2
Rank	2		1		4		3
5 (openness to foreign opportunities)	72.7	4	81.3	2–3	81.3	1	75.0	3
Rank	4		1–2		1–2		3
6 (managerial experience)	66.4	6	71.9	7	71.9	5	61.7	6
Rank	3		1–2		1–2		4
7 (co-operation)	74.2	2	81.3	2–3	70.3	6–7	76.6	1
Rank	3		1		4		2
Total prediction accuracy	93.8		96.9		85.2		91.4
Rank	2		1		4		3

^aRanks in rows order the accuracies by dependent variable, while ranks in columns by blocks of questions.

Table 5 demonstrates that similarly to the results presented in Table 4 and to the findings of academic spin-offs’ internationalization literature presented in Table 1 and Appendix 1, the most useful questions concern firms’ international orientation, openness to foreign business opportunities, co-operation and technology [9,14,24,36,37,40,51,54,76,79]. However, the usefulness of these question blocks differs for the studied types of internationalizers.

Table 5 also shows that prediction accuracies vary: they are higher for less active internationalizers –especially for INTTs whose prediction accuracy ranks 1–2 through all question blocks – and lower for born globals. For the latter, rankings of the seven question blocks also differ considerably: for instance, while co-operation was ranked first for firms that are currently still very active internationally in terms of export share, number and location of foreign markets, this block was ranked 6.-7. For reaching a born global status during the first three years since foundation. This is not in accordance with the literature on born globals and other internationalizers as according to several authors, network relationships are very important for firms’ initial and subsequent internationalization: through network relationships, firms can get access to knowledge about foreign markets and internationalization processes, and network contacts help them to identify and follow attractive international business opportunities [[2], [3], [4], [5],77]. However, as the accuracies of different blocks for predicting BG3 vary in maximum by 11% points (this indicator was 18 for BGT), this result is not an indication that academic spin-offs should not co-operate with others.

Finally, although, overall, all prediction accuracies (including all seven blocks) can be considered very high, they are somewhat higher for the current situation and lower for the firms’ statuses during the first three years (INTT compared to INT3 and BGT compared to BG3). However, in the latter pair, there are also some exceptions, especially regarding industry characteristics, technology and managerial experience. These factors could be more important for overcoming initial entry barriers [29,31]. The role of certain other factors – including international orientation and co-operation – may, in turn, become more visible during later internationalization stages [36,51,54].

4.4. Scientific implications regarding the prediction of academic spin-offs’ internationalization

Based on the empirical findings, we can conclude the following about predicting the academic spin-offs’ internationalization types.

• 1.Internationalization as just entering at least one foreign market can be predicted with a higher accuracy when compared with more active forms characterized by more markets and minimum 25% export share from turnover (e.g., the born global type). This would point to the fact that in the world of high technology and inventions, academic spin-offs’ internationalization is complex. Thus, even the implementation of all necessary conditions to become globally successful does guarantee becoming a born global. On the other hand, some firms can achieve this status due to chance and/or luck or due to the specific characteristics of their business model despite lacking considerable efforts to internationalize early [66,81].
• 2.Internationalization in the longer horizon can be predicted with a higher accuracy than in the short run. This would point to the fact that different steps taken by the management need varying time to manifest, while some of them might not succeed [77].
• 3.In case of less committed internationalization, factors directly related to cross-border activities play a pivotal role. Among them could be pursuing exporting already from the beginning, having an organically international product or co-operating with different industrial players outside the home country [14,76].
• 4.In case of more committed internationalization (manifested both in speed and magnitude), factors such as technology, competition and demand have a more pronounced role. Still, these can be considered complementary rather than exclusive of the factors directly related to cross-border activities. This would point to the necessity to possess a bundle of pivotal determinants rather than some standalone key success factors [21].

5. Conclusions and implications

5.1. Conclusions

This paper contributed to academic spin-offs’ internationalization literature, export prediction literature and the literature on born globals by finding out which variables are the most useful for predicting which firms will achieve at least some internationalization in three years or later and which firms will become born globals in three years or achieve some born global criteria (export share, number and location of foreign markets) later. It covers all the main domains important for and previously used for predicting academic spin-offs’ internationalization. Unlike other studies, which focus on a single layer of analysis, this paper outlines the individual importance of variables, domain-based and total accuracies.

The results showed that academic spin-offs which are oriented towards international markets from the beginning of their activities and constantly pursue new international opportunities are more likely to internationalize fast (become born globals or less committed internationalizers) and remain international for a longer time period than other firms. Being the first to introduce product or process innovations and co-operating with international government institutions, competitors, customers, and/or suppliers is also relatively useful for such internationalizers. Differently from some previous studies, the least useful predictors were the management team’s foreign academic degrees and international business training.

The paper also demonstrated that predicting if an academic spin-off will become a less active internationalizer results in a higher accuracy than predicting if it will become a born global. Moreover, predicting if a firm will belong to a certain internationalization type (become a more versus a less committed internationalizer) later results in a higher accuracy than predicting if this will happen during the first three years since the spin-off’s foundation.

5.2. Managerial and policy implications

Based on the results, several managerial and policy implications can be developed. Academic spin-offs’ managers should become more oriented toward international markets if they want to achieve fast internationalization. Moreover, they should be active in pursuing new international opportunities, introducing product or process innovations and co-operating with international government institutions, competitors, customers, and/or suppliers. Still, they should understand that internationalization success is not guaranteed: foreign activities can fail due to various firm-, network- or external environment-related reasons.

As for several dependent variables, the prediction accuracy in case of using all independent variables is exceeding 90%, the application of specific models could be beneficial for different organizations in practice. Namely, at that precision and accounting for equal accuracies in both classes of the binary dependent variable, technology transfer offices of respective universities are able to preselect over 90% of firms with internationalization potential for funding. The first and second type of misclassifications account both for less than 10%, meaning that around one tenth of potential internationalizers remain undetected, while there is similar share of firms funded without the respective perspective. The result is in line with the highest prediction accuracies obtained in earlier studies about internationalization of non-academic start-ups [21], while usually the prediction accuracies for start-up success or failure have remained more modest (e.g., Refs. [82,83]).

Policy-makers should pay more attention to supporting academic spin-offs’ international activities. Many academic spin-offs have academic contacts – for example, created during studies abroad or participating at academic conferences – but they lack useful international business contacts. Thus, policy-makers could organize practitioner-oriented networking events and discussion forums where spin-offs’ managers could present their technologies and products to business people and create useful contacts. Moreover, they could support spin-offs’ managers’ visits to international trade fairs and offer funding for academic entrepreneurs’ international stays to develop relationships with market players. That would allow them to identify international business opportunities. Sharing information about how to enter foreign markets – for instance, which documents should be filled and which quality certificates are required – could be also useful as some academic entrepreneurs lack such knowledge.

5.3. Future research directions

This paper focused on Spanish academic spin-offs. Focusing on one country can be considered a limitation of the study although this is relatively typical for the literature on such firms (see Appendix 1). Future researchers could collect survey data from several countries, as this would also allow taking the differences between their internal environments into account. Our sample size is comparable to the ones used in previous studies (see Appendix 1) but it could be increased in future research. In case of small samples, the prediction accuracy is strongly dependent of the classification of individual cases: in this study, the false classification of each firm becoming a born global during the first three years would lead to a reduction in the classification accuracy for the respective firm group by around three percentage points. Larger samples would also enable researchers to study the interconnection between different predictors more systematically.

In addition, while our sample’s median age was eight years, studying older firms would also allow researchers to find out if it is possible to predict which firms will de-internationalize and later re-internationalize as the latter phenomena can occur in a longer time frame. It must be also emphasized that several questions portray the status quo of a firm’s strategy, not its dynamics, while we cannot fully exclude the possibility that during the relatively short average lifetime, certain changes in it have occurred. Thus, future surveys might also focus on dynamics. Moreover, they could encompass other variables and sub-topics not covered by this study. In addition, while this paper used logistic regression analysis, some other prediction methods – for instance, machine learning methods like neural networks, decision trees, and rough sets that have been used in some export prediction studies (for an overview, see Ref. [21]) – could be applied for forecasting, as they might lead to higher accuracy due to being able to account for the linkages between different independent variables.

Appendix 1

Methodological and theme variety of the literature on academic spin-offs’ internationalization¹

Study	Data	Method(s)	Main results
[8]	Italian databases (508 spin-offs and 400 other firms)	PR, HSM	academic spin-offs are more likely to internationalize – start exporting within 2 years since foundation – than other firms, especially if their parent university is very internationalized (in terms of attracting foreign students and faculty members but also regarding the share of publications co-authored by foreign universities’ staff members)
[9]	Spanish survey (173 responses)	CFA, SEM	academic entrepreneurs’ international human capital and international market relational capital affect their international orientation; their psychological capital is related to international orientation through international market relational capital and international human capital
[10]	Spanish survey (173 responses)	CFA	born globals are relatively common among academic spin-offs but some spin-offs also internationalize more slowly; international spin-offs get financing mainly from governmental institutions, venture capitalists and other financial investors while local spin-offs get it mainly from governmental institutions and academic agents
[12]	Spanish survey (173 responses)	CFA, SEM	human capital (academic entrepreneurs’ international experience and training), network relationships with foreign academic agents, and psychological capital are antecedents of academic enterprises’ internationalization
[13]	Portuguese survey (111 responses)	OLS, LR	younger academic spin-offs and those that receive support from technology transfer offices, are not too much R&D-focused and/or operate in robotics or microelectronics are more likely to internationalize earlier and faster than others
[14]	Spanish database (219 spin-offs, 121 others)	DA, GLS	technology-based university spin-offs achieve higher employment and sales growth than other technology-based firms and they are also more international; however, the share of exporters is also low among university spin-offs (only 14.5% of spin-offs and 0.4% of other technology-based firms export)
[15]	Portuguese survey (111 responses)	DA, OLS	for 55% of respondents, technology transfer infrastructures were helpful to get access to local or foreign business networks; larger and more experienced academic spin-offs achieved better local and international performance
[19]	databases (465 Spanish university spin-offs)	DA, CPHM	exporting increases university spin-offs’ survival probability as it offers additional sales opportunities, but firms need enough resources (financial assets, employees) to export successfully
[20]	9 Greek cases	qual.	internationalization is important for Greek academic spin-offs as the local market is too small; networks are useful for international and local growth: some have so-operated with local firms, some have found foreign partners or shareholders
[23]	–	conc.	several topics should be still studied: the role of the university context and universities’ international and local links in knowledge diffusion, product development and internationalization; the role of founding teams and non-academic actors in academic spin-offs’ internationalization; failures in internationalization
[24]	8 Swedish cases (3 academic firms, 5 others)	qual.	academic contacts are important for academic firms’ internationalization; to succeed in some foreign countries, it is important to commit time and other resources; universities are useful sources for technological knowledge, but less useful sources for internationalization knowledge: thus, some academic firms internationalize sporadically
[29]	Spanish survey (126 responses)	DA, PLS	involving non-academics (e.g., business people) in the founder team had a very important role in supporting academic spin-offs’ internationalization, innovation and overall performance; domestic spin-offs often lack previous management experience
[31]	Norwegian survey (109 responses)	LR	academic spin-offs with fully developed products, good network connections, industry-experienced team members (especially from very similar or very different industries) and board members with different functional backgrounds are more likely to export and create foreign strategic alliances
[33]	7 Polish cases	qual.	firms benefiting from cost and tax advantages and venture capitalists’ support are likely to succeed in global markets; managers’ international scientific co-operation, foreign experience and knowledge are also very important for academic spin-offs’ internationalization
[34]	Italian database and survey (120 responses)	DA	academic spin-offs are more likely to internationalize via exports and foreign partnerships; many firms internationalize in 3–5 years since foundation and expand to several continents, however, for some, this is a sporadic event; internationalized firms are larger and better financed than others
[35]	Spanish database (237 firms)	DA, PR	among several other factors, internationalization positively affects the probability that an academic spin-off will become a high-growth firm (that it will achieve a 20% or higher average annual turnover growth during a 3-year period)
[36]	10 Swedish cases	qual.	academic spin-offs can internationalize fast as they benefit from the university’s scientific knowledge and academic networks; however, some stop growing due to lacking business competence, experience and business network relationships with potential large foreign customers
[37]	survey (5 countries, 895 responses)	DA	78 scholars had their own business (60% had internationalized), 47 had detailed plans, 146 did not have detailed plans yet but also planned to start their own business; most respondents understood the importance of international contacts and most of the potential entrepreneurs also intended to internationalize
[38]	–	conc.	relational capital (e.g., relationships with customers, suppliers, competitors, other firms, public institutions) positively affects academic spin-offs’ internationalization; firms’ international performance depends on their networking capabilities
[39]	survey (62 Dutch and 43 Norwegian responses)	DA, LR	about 74% of academic spin-offs internationalized during the first 5 years since foundation, while some stayed domestic, some firms’ international growth slowed down while some spin-offs’ international involvement decreased; network relationships were very important for internationalization
[40]	2 Danish and 2 Irish cases	qual.	internationalization and innovation are strongly interrelated as scientific knowledge creates opportunities for fast internationalization; founders’ resources, academic networks and confidence in their ideas’ entrepreneurial potential are important for academic spin-offs’ internationalization
[41]	3 Norwegian cases	qual.	young academic spin-offs’ internationalization benefits from network relationships created before or during internationalization (in terms of knowledge, technology, financial, reputational and other resources); pre-founding period should not be ignored as technology and product development can take years
[42]	4 Australian cases	qual.	academics’ network relationships – especially, those created during academic conferences and contacts with important industry people – are crucial for identifying and exploiting initial foreign market opportunities; proactive and innovative firms were able to succeed internationally
[50]	survey (58 Dutch and 41 Norwegian responses)	DA, LR	61% of academic spin-offs were involved in international knowledge networks; firms with more innovative products needed more time for creating such relationships and for internationalizing than less innovative firms; having team members with PhD experience associated with more active network involvement
[51]	3 Irish cases (1 academic spin-off, 2 others)	qual.	during an academic spin-off’s international development, several changes can occur in its products, staff, company culture, financing, network relationships etc.; overall, flexibility, international vision, market knowledge and network relationships are very important for internationalization
[52]	Belgian survey (114 firms: spin-offs and others)	DA, CFA, OLS	young firms that lack experiential foreign market knowledge can use other knowledge sources (for instance, they can learn from their customers, suppliers, investors or other important partners or hire experienced managers): this knowledge helps them to internationalize successfully
[53]	56 British firms (27 from a survey, 29 more from databases)	DA	11% had offices abroad (all of those in the USA, some also in other countries: Israel, Norway, Germany or Japan) while 7% had headquarters abroad (in USA or Belgium), 21% exported but for 90%, internationalization was the main priority; network relationships were important for internationalization; some spin-offs aimed to merge with other firms or to become acquired
[54]	1 Danish case	qual.	academic and business (especially, with venture capital firms) networks are very important for an academic spin-off’s fast and successful internationalization: due to knowledge acquired through such networks, it is possible to reduce experiential learning time
[76]	22 Italian cases	qual.	in the last phases before official foundation, born globals are more proactive (for instance, in terms of learning, using network relationships and pursuing opportunities) than slower internationalizers: as a result, born globals are more market-ready
[79]	4 U S. cases (2 academic spin-offs, 2 others)	qual.	some firms develop innovative products, launch ventures and internationalize very rapidly; it is important not to stay too technology and product development-focused for too long as without launching ventures, the firm’s overall and international development will be delayed; having academic and business network relationships accelerates the pace of internationalization
[84]	international databases (220 initial public offerings)	DA, PR	spin-offs affiliated with more internationalized (in terms of foreign students, international co-authored publications and patents) and prestigious (according to the Academic Ranking of World Universities) universities are more likely to get targeted by cross-border mergers and acquisitions as they can get access to more valuable knowledge and other resources
[85]	6 Australian cases (2 university spinouts, 4 others)	qual.	as technological innovation can take years, university spinouts can have long pre-foundation (the period before legal registration) histories (during which, often, there several milestones: first publication, first patent, assembly of scientific and entrepreneurial team, finding first employees, …): thus, born globals’ internationalization can be much slower than it may initially seem

¹CFA: confirmatory factor analysis, conc: conceptual, CPHM: Cox proportional hazards model, DA: descriptive analysis, GLS: generalized least squares, HSM: Heckman selection model, LR: logistic regression, OLS: ordinary least squares, PLS: partial least squares, PR: probit regression, qual: qualitative, SEM: structural equation modeling.

Appendix 2

Variables used for predicting firms’ internationalization types or statuses

Variables	Studies
firm size (e.g., annual sales, number of full-time employees)	[57,59,64,65,[86], [87], [88], [89], [90], [91], [92], [93]]
availability and efficiency of resources: (excess) production capacity, (total factor) productivity; productivity or value added per worker, availability of financial, human and other resources, availability of management time	[55,59,[61], [62], [63], [64], [65],88,92,[94], [95], [96], [97], [98]]
innovativeness (e.g., developing significantly improved or new products, designs, materials, services or processes, using innovative marketing methods)	[48,55,56,59,91,95,99,100]
products’ competitiveness (e.g., differentiation, novelty, uniqueness, niche orientation, quality, price, brand, image)	[48,55,56,64,65,87,97,98,101,102]
other production characteristics: e.g., industry type, manufacturing technology (e.g., level of automation, being a technological leader), R&D intensity, origin of raw materials, following quality standards	[48,57,64,65,86,87,89,91,97,102]
decision-makers’ or the firm’s prior export or import knowledge (e.g., about business practices, documentation etc.), international or local experience	[21,22,48,49,57,59,61,88,90,94,97,[101], [102], [103]]
other skills and/or abilities: e.g., staff’s or the entrepreneur’s education level and foreign language skills	[49,64,65,87,89,90,95,99,103]
(perceived) export (market) risk and/or export barriers: e.g., bureaucracy, regulations, tariffs, transport costs, firms’ perceptions of internal barriers (e.g. uncompetitive products, lack of resources)	[59,61,62,64,65,88,90,99,102,104]
attitudes toward international vs. domestic growth, learning, marketing, planning, control, risk taking; making product, pricing or marketing adjustments, responding to competitors’ actions, getting and using information about foreign markets, etc.	[22,55,56,[60], [61], [62],64,65,88,94,[97], [98], [99],[101], [102], [103]]
foreign market information availability and diversity: e.g., using different information channels, learning from other firms that have export experience in the target country, participating in international networks, contacting export agencies	[48,62,64,65,100,103,105]
technological or other co-operation: e.g., cluster membership of the firm, being an original equipment manufacturer or an assembly firm, co-operation with local banks, business consultants, scientists or other local firms to develop new products etc.	[48,87,89,91,95,102]
network’s international characteristics: e.g., international network’ size and strength, level of network members’ internationalization	[48,58,97,101,103]
home market’s characteristics: e.g., business environment, market size, industry dynamism, demand, competition, level of development, government assistance, political regime, inflation rate etc.	[48,49,[55], [56], [57], [58], [59],61,62,89,94,102]
foreign markets’ characteristics: business environment, market size, cultural and language similarities and differences, technological developments, infrastructure, competition, government’s attitude toward foreign firms (support vs. restrictions), exchange rate etc.	[55,[60], [61], [62], [63], [64], [65],88,95,102]