Do incubatees form more alliances? Nature of the alliance and start-up ownership as contingencies

ABSTRACT

Incubators are reputed to be key institutions for the creation and growth of viable and successful entrepreneurial ventures. One mechanism through which their beneficial action should unfold is that incubatees could be more likely to stipulate alliances with third parties. We explore, both theoretically and empirically, the possibility that this crucial bridging function performed by incubators is indeed contingent on both (a) the type of alliance that start-ups are seeking for, where we distinguish between R&D and commercial alliances; (b) the specific ownership structure of the start-ups. Our analysis is based on a dataset of 1766 incubatees and non-incubatees young innovative companies. Results highlight that incubators accomplish their bridging role depending on the two above-mentioned contingencies. In particular, incubated start-ups show higher probabilities than non-incubates to stipulate R&D alliances only if they are university-backed, while commercial alliances figure as a prerogative of incubatees only when these latter are business-backed.

KEYWORDS:

• Start-ups
• ownership structure
• incubators
• R&D alliances
• commercial alliances

JEL CODES:

• L25
• L26
• O32
• G32

Acknowledgments

The authors would like to express their gratitude to the editor, the anonymous reviewers, and the participants of the R&D Management Conference 2022 in Trento for their valuable feedback during the development of this paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 16 (ex-)innovative start-ups exceeded the law threshold of 5 years (accounting for less than 1% of the sample) and we include them in the analysis. All findings here exposed are unaffected by the choice of including or excluding these start-ups.

2 The Italian Startup Act also ‘introduced the notion of certified incubator, with the objective of encouraging the recognition and valorisation of enterprises having a solid, proven experience in supporting the creation and development of high-potential innovative startups.' (MISE 2017, 21).

3 More precisely the test of the hypotheses is as follows: for what concerns H1, $H_0:({\beta }_1+{\beta }_2+{\beta }_4)-{\beta }_4=0$ (i.e. the fact that a university backed start-up is incubated has a positive effect on its probability of establishing R&D alliances); while for what concerns H2, $H_0:({\beta }_1+{\beta }_3+{\beta }_5)-{\beta }_5=0$ (i.e. the fact that a business backed start-up is incubated has a positive effect on its probability of establishing commercial alliances).

4 Since our specification includes a combination of continuous and binary variables, we also computed point biserial correlation coefficients for each couple of variables belonging to different formats. Results confirm that pairwise correlation is never an issue.

5 We also computed alternative indicators commonly used to detect collinearity, i.e., the condition index and the determinant of the correlation matrix. All these measures undisputedly indicate that collinearity is not problematic.

6 Some industry dummies predict the outcome (R&D Alliance or Commercial Alliance) perfectly. Thus, depending on the model, we have to drop 14 or 15 start-ups from the sample.

7 We re-estimated our models by excluding observations related to start-ups with mixed ownership, meaning those backed by both universities and businesses. The results remain unchanged.

8 We also compute the correct marginal effects of a change in the two interacted variables – i.e., Incubation and University Shareholders at Entry for the R&D alliance equation and Incubation and Business Shareholders at Entry for the commercial alliance equation – for a probit model, as well as the correct standard errors by using the procedure by Norton, Wang, and Ai (2004). Results are consistent with those reported in Table 5.

9 We are conscious that the same line of reasoning can also lead to the speculation of endogeneity for the variables University Shareholders at Entry and Business Shareholders at Entry. However, two features, one relative to the way in which the variables are operationalized and the other relative to our empirical strategy should mitigate the concern. First, we measure backing at foundation, a stage at which the screening ability of investors should be considerably weaker. Second, in order to test our hypotheses, we are interested in comparing the probabilities of establishing alliances with and without incubation between homogenous groups, i.e. within the groups of university-backed and business-backed start-ups. In other words, by using interaction terms, our control groups are no longer non-incubated start-ups of any sort, but non-incubated university-backed start-ups for hypothesis H1, and non-incubated business-backed start-ups for hypothesis H2.

10 Note that the second stage equation includes two interaction terms as independent variables (i.e. Incubation*University Shareholders at Entry and Incubation*Business Shareholders at Entry) which cannot be included in the first stage by construction. This discrepancy may give raise to the so-called ‘covariate ambivalence’ problem (Angrist and Pischke (2008, 189). Grounding on Grilli and Murtinu (2018), the modification depicted in the main text aims at tackling this issue. Note also that in the first stage, we prefer to use a higher level of disaggregation of business backing at foundation, by discriminating between independent venture capital, business angels and ownership by other companies, so to increase the strength of our instruments related to Incubation. However, given these differences between first and second stage (besides the exclusionary restriction) and our remedies which can only be partial, we recommend to interpret results with caution.

11 In order to increase the strength of our instruments we also chose to estimate the first stage on a sample with a greater size than our benchmark (1766), since more information in the database is available on incubation than alliance activity.

Additional information

Funding

Luca Grilli acknowledges for its contribution to the completion of this research the support of the PRIN MUR Research grant (financed by the European Union program Next Generation EU - M4 -C2 -1.1) titled: Institutions, Education and Entrepreneurship Studies: a Novel Perspective on University Impact (IEES), ID MUR: 2022AYY9RK_01; CUP: F53D23003040006.