Social structure and alliance formation patterns: a longitudinal analysis - includes appendix

Administrative Science Quarterly,  Dec, 1995  by Ranjay Gulati

• E-mail
• Print
• Link

Control Variables

I included as controls a number of variables known or expected to affect the alliance activity of firms but not included in the discussion of the hypotheses. These included measures of time, sector, industry trends, and firm-level attributes.

I added dummy variables for each year to capture effects of temporal trends. For simplicity of presentation, I then reestimated these effects using a single variable, time, which ranges from 0 to 8, with the default year being 1981, and assumes linearity in the effect of time. I observed no differences in the results based on the two controls for time. I controlled for sector differences with two dummy variables, sector 1 and sector 2. The former indicates the new materials sector and the latter, industrial automation, with the automotive industry the default sector.

To assess claims that the growth in alliances may be the result of a bandwagon effect, for each dyad-year record I created the variable total alliances, which is the total number of alliances announced in the industry in the previous year, information I obtained from a larger data set on all alliances in each industry. An alternative interpretation of this measure is that it captures the net effect of the various macroeconomic factors that may influence the formation of alliances in an industry (Amburgey and Miner, 1992). Yet another interpretation may be that it proximates the ecological notions of carrying capacity, in the sense that a large number of prior alliances in an industry may lead to saturation and a diminishing number of new alliances (Baum and Oliver, 1992).

I also controlled for any firm-level effects not included in the current dyadic formulation. I included two variables, one for each firm in the dyad, indicating the prior alliance experience of each partnering firm, alliance history, which is the number of prior alliances entered by each firm prior to the given year. This captures the possibility of repetitive momentum in individual firms' alliance activities. Including these variables also controls for firm-level heterogeneity (Heckman and Borjas, 1980) and issues related to the finite capacity of firms to enter alliances in a given year (discussed in the Appendix). To avoid possible confounding in interpreting these variables, I computed them net of repeated ties.

Lastly, I included several variables to control for the possible influence of the financial attributes of firms in guiding alliance formation. Several scholars have looked at such attributes as predictors of firms' rates of participation in alliances. In the dyadic context, including firm attributes involved looking at the difference between firms in a dyad on each attribute. I looked at firm size, performance, liquidity, and solvency, each a key indicator of a firm's resource base that can be critical in its alliance decisions. Size indicates a firm's financial and managerial resource endowment as well as its level of economies of scale and scope. Performance indicates its degree of success in the marketplace. Poor performers may seek alliances to improve performance, and good performers may want partnerships to leverage some of their successes. Firms also frequently enter alliances to share the costs of new projects, particularly those involving large resource outlays and risks. In this context, liquidity, which reflects the short-term resources available to a firm, is important. Firm solvency addresses similar concerns but with regard to long-term resources. To measure each attribute, for each dyad, I divided the smaller value by the larger value. The larger this ratio (highest value = 1), the closer the two partners were on the given attribute. A new ratio was computed for each dyad for each year based on the attribute values for each firm for the prior year.