Embedded patterns of international alliance formation

Organization Studies, Fall, 1998 by Richard N. Osborn, John Hagedoorn, Johannes G. Denekamp, Geert Duysters, C. Christopher Baughn

The CATI data provided 762 alliances involving Japan, Western Europe and/or the United States. Here, there were 129 supply contracts, 344 technical agreements and 289 joint ventures; 305 of 762 alliances announced unidirectional product/knowledge flows and 457 announced reciprocal flows. Similarly, for the CATI data, there were 229 in semiconductors, 169 in computers, 227 in communications/information systems and 137 in autos/auto parts.

The interactive hypotheses which predicted embeddedness (H1) was tested using hierarchical log linear analysis. This procedure is ideally suited to examining interactive associations among categorical variables (see Dillon and Goldstein 1984 for a discussion). These results are contained in Tables 1 and 2 for the JETRO and CATI samples, respectively. The findings for the interactive results were completely consistent with H1. There were significant triple-order interactions for both samples. Specifically, industry, product/knowledge flow direction and administrative form were interactively related over and above the additive and second-order effects, as suggested by the institutional embeddedness hypothesis (HI).

An inspection of the data (complete data available from the senior author) indicated mixed evidence for the hybridized and dominance patterns predicted in Hypotheses 2 and 3 when viewed across both samples. Consider first Hypothesis 3, which predicted a hybridized pattern involving reciprocal flows administered via technical agreements within technology-intensive industries with potential commercial inseparabilities. In semiconductors and computers, reciprocal technical agreements were, in fact, disproportionately more popular than any other combination for both samples. Specifically, with an even distribution of forms and flows in semiconductors one would expect 22 reciprocal technical agreements in the JETRO sample while the actual number was 40 (p [less than] .01); in the CATI sample the expected value for semiconductors was 38 and the actual was 85 (p [less than] .01). For computers, the relevant data were 40 expected reciprocal technical agreements with 75 actual (p [less than] .01) in JETRO and 28 expected with 46 actual (p [less than] .05) in CATI. There were too few alliances in communications for the JETRO sample but in the CATI data 25 reciprocal technical agreements were expected while 98 were found (p [less than] .01).

Table 1

Results of Log Linear Analysis for Alliance Formation Using the
JETRO Data US-Japanese Firms, 1988-1989

Variable/Effects       Degrees of      Pearson        Partial
                        Freedom      Chi-square      Chi-square

All one-way effects        6        450.102(***)
Industry (I)               3                        304.097(***)
Governance form (G)        2                        21.630(***)
Flow (F)                   1                        92.159(***)

All two-way effects       11        238.734(***)
I x G                      6                        62.976(***)
I x F                      3                        78.698(***)
G x F                      2                        80.398(***)

Three-way effect
I x F x G                  6         14.612(**)

One-way and higher        23        703.448(***)
Two-way and higher        17        253.346(***)

n = 625; *** significant p [less than] 1%; ** significant p [less
than] 5%; * significant p [less than] 10%
Table 2

Results of Log Linear Analysis for Alliance Formation Using The CATI
Data US-Japanese-European, 1970-1989

Variable/Effects       Degrees of      Pearson        Partial
                        Freedom      Chi-square      Chi-square

All one-way effects        6        460.929(***)
Industry (I)               3                          30.525(***)
Governance Form (G)        2                         108.491(***)
Flow (F)                   1                          33.082(***)

All two-way effects       11         287.402(***)
I x G                      6                         131.130(***)
I x F                      3                          34.118(***)
G x F                      2                         148.800(***)

Three-way effect
I x F x G                  6          23.811(***)

One-way and higher        23         149.716(***)
Two-way and higher        17         297.040(***)

n = 762; *** significant p [less than] 1%; ** significant p [less
than] 5%; * significant p [less than] 10%