Firm strategy and age dependence: a contingent view of the liabilities of newness, adolescence, and obsolescence

Administrative Science Quarterly,  June, 1999  by Andrew D. Henderson

• E-mail
• Print
• Link

Because members of larger communities may enjoy system-level economies of scale (David and Greenstein, 1990), I controlled for community size, measured by the natural log of total annual sales of personal computer products in each community (Wade, 1995). The density of firms within a community may affect processes of competition and legitimation (Hannah and Freeman, 1989; Baum and Oliver, 1992), thus I controlled for the number of community members. A community's age may also affect its legitimacy, as well as the datedness of its technology (cf. Stinchcombe, 1965), so community age was measured in years. Because conditions at founding can have lasting effects on firm performance (Carroll and Hannan, 1989b), I controlled for community age at founding, measured as the age of the community that a firm was born into, and number of members at founding, which measured the density of firms in that community in the year a firm was founded. squares of each of the community controls listed above were also examined and included wherever they were significant.(5)

Wade (1995, 1996) also described the importance of community sponsors, which invite second-source entry by licensing their product designs to others. In this industry, proprietary strategists did not license their designs, but IBM and Sun Microsystems did. To control for this, I coded the community sponsor variable 1 for IBM starting in 1981 (the year it introduced and began licensing the PC) and for Sun Microsystems starting in 1989 (the year it introduced and began licensing its Sparc technology), and 0 otherwise. The Z80 standard was based on a design consensus among a number of firms during the industry's early years (Anderson, 1995). Thus, there was no dominant or leading sponsor of that standard. I did not include the sponsor control in the failure rate models. Since neither IBM nor Sun failed, those models would not converge if it were included.

Control variable x strategy interactions. I also expected that the effects of some controls would differ by strategy. For example, the IBM era might be more disruptive to proprietary strategists than standards-based strategists. In analyses not shown here, each of the controls was interacted with strategy. None of those interactions was significant in the sales growth models. In the failure rate models, the interactions of strategy with founding density, IBM installed base, and IBM era were all significant, so those interactions were included there. Non-significant interactions were omitted to avoid collinearity.

Model Specification and Estimation

Specification. Hypotheses 1b and 1c predicted that sales growth would have curvilinear relationships with age. Failure rates would also be curvilinear if firms exhibited a liability of adolescence. Consequently, growth and failure rate models were specified as follows:

r(t) = [[Beta].sub.0] strategy + [[Beta].sub.1] age(t) + [[Beta].sub.2] [age.sup.2](t) + [[Beta].sub.3] age(t) [multiplied by] strategy + [[Beta].sub.4][age.sup.2](t) [multiplied by] strategy + [Gamma]X(t) (1)