The imprinting effects of parent firms on the evolution of young spinoff alliance networks

ABSTRACT

We tested the drivers of spinoffs’ (that is, new firms started by ex-employees of incumbent firms) alliance network size through the lens of imprinting theory, using a large longitudinal sample of 145 newly founded spinoffs and 3,405 strategic alliances from 2001 to 2014 in the alliance-intensive mining industry. Our results revealed that whereas parent firms’ network positions in terms of size and centrality leave an influential early imprinting effect on spinoffs’ alliance network size, initial partners’ network position leaves an effect through path-dependent forces. Further, our analysis revealed that the parent’s network characteristics can influence the choice of initial partner. We discuss implications for alliance network emergence, spinoffs, and imprinting theory.

KEYWORDS:

• Strategic alliance
• networks
• spinoff
• imprinting
• mining

Acknowledgments

The authors gratefully acknowledge the comments on the early versions of this manuscript from Professor Hana Milanov, Professor Dean Shepherd, and reviewers at the Australian Centre for Entrepreneurship Research Exchange (ACERE) conference.

Disclosure statement

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

Notes

¹ Strategic alliances are voluntary cooperations among firms to achieve the same strategic goal (Gulati et al., 2012).

² The spinoff literature has recognized different types of spinoffs depending on the spinoff mode and the initiator. Typically, if ex-employees start the new firm, it is called a spinoff, and if incumbents start it, it is called a spinout. For a typology, see Fryges and Wright (2014).

³ ABS 2011 reports (mining cities: Perth, Brisbane, Adelaide, Mackay, Melbourne, Kalgoorlie-Boulder, Mount Isa, Newcastle, Sydney, Wollongong, Townsville). This list has been developed according to the number of permanent residents employed in the mining industry in each urban area.

⁴ Given we have a number of insignificant control variables, we ran a more parsimonious model with just the significant control variables. The results for our hypotheses remained the same..

⁵ For random effect panel models such as ours, Xtivreg uses a two-stage least-squares random-effects estimator for including instrumental variables in panel data models. There are two implementations: G2SLS from Balestra and Varadharajan-Krishnakumar (1987) and EC2SLS from Baltagi and Baltagi (2008). We used G2SLS because it is computationally less expensive. We ran the tests again with the EC2SLS and found the same results.

⁶ We thank one of the reviewers for their suggestion on how interpersonal networks of founders may affect the alliance network of spinoffs and how this could be investigated by future researchers.

Additional information

Funding

The authors appreciate the financial support received from the Queensland Research Council and the Australian Research Council to undertake this research (grant number LP130100415).