12Why Do MNCs Divest or Retain Foreign Subsidiaries?Approaches from Dependency and Redundancy in Subsidiary Networks Naoki Yasudafrom Japan is high (p＜.001), subsidiaries are large (p＜.001), subsidiary return on sales is higher (p＜.05), subsidiary type is networked (p＜.05), the parent firm is relatively small (p＜.001), the parent firm’s R&D intensity is relatively low (p ＜.05), and the host country’s growth is relatively low (p＜.01). I incorporated hypothesis-testing variables in Models 2 through 9. Model 2 delivers a statistically significant negative relationship of the degree of a subsidiary’s hub function in a country (p＜.001), so Hypothesis 1a is supported. Model 3 exhibits a significant and negative relationship of the degree of a subsidiary’s authority function in a country (p＜.001), which supports Hypothesis 1b. In Model 4, the coefficient for redundancy in the subsidiary function in a country is not statistically significant; therefore, Hypothesis 3a is not supported. In Model 6, the coefficient for the degree of regional hub function is negative and statistically significant (p＜.001), supporting Hypothesis 2a. In Model 7, the coefficient for the degree of authority function in a region is negative and statistically significant (p＜.001), supporting Hypothesis 2b. In Model 8, redundancy of intra-regional subsidiary functions shows a significant effect (p＜.10); however, this result is contrary to Hypothesis 3b. Model 5 incorporates the degrees of countrywide hub function, authority function, and redundancy of intra-country subsidiary functions. I found significant negative effects for the degree of hub and authority function in a country (p＜.05; p＜.001). However, the redundancy variable shows a negative and statistically significant effect (p＜.01), which is contrary to Hypothesis 3a. Model 9 incorporates the degree of hub and authority function in a region and the redundancy of intra-regional subsidiary functions. I found negative significant effects for regional network authority (p＜.001). I also found that the redundancy of intra-regional subsidiary functions to be negative and significant (p＜.10).Robustness Check This study employs two alternative methods to enhance the validity of these estimations. First, given that the dependent variables are determined at time t + 3, I used alternative dependent variables. I created dependent variables that show divestment within three years. To do so, I specified a variable with a value of “1” if an MNC divests a focal subsidiary at time t + 1, time t + 2, or time t + 3. Although I do not report the results for the sake of space, the results are similar to the original estimates. Second, the study assumes that subsidiaries form regional networks by exporting and importing activities across subsidiary networks. To check the validity of the assumption, I used alternative independent variables that illustrate the network more directly (Figure 2) and created samples focusing on Asian subsidiaries. To create variables that illustrate regional hub and authority functions, I divided a focal subsidiary’s exports within Asia by an MNC’s total regional exports within Asia. I also divided a focal subsidiary’s imports from Asia by an MNC’s regional total imports within Asia. Table 4 shows that the results are consistent with original estimations because the explanatory variables show negative and significant effects (p＜.001; p＜.05).