The following section of the thesis aims at interlacing the results from the network analysis and the multiple correspondence analysis. The purpose is to conclude whether there is a correlation between pairs of actors’
distances in the network and distances in their political positions, hence the extent to which social contagion has occurred between the manufacturers.
To determine whether contagion has occurred, a simple regression analysis has been carried out to test for a linear relationship between the two variables. A positive slope between the two variables would imply that the network distances are positively correlated to the political distances of pairs of actors. This would, other things being equal, indicate that social contagion has occurred, since a short distance in the network correspondently would be equal to a short distance politically (and reverse for long distances). This is in line with the theory of social contagion arguing that actors engaging in frequent communication, hence have a short network distance, will tend to be similar in their behavior, attitude and beliefs, hence have a short political distance in this specific case. A negative slope or a non-existent relationship between the two variables would indicate no signs of contagion, since only a positive correlation between the two variables makes sense in relation to the theory of social contagion.
Figure 1 visualizes all 45 observations (pairs of actors) in a scatterplot with the network distances on the horizontal axis and the political distances on the vertical axis.
Figure 18: Plot visualizing all 45 pairs of actors and the relationship between their distance in the network and their distance in political positions. Units are geodesic distances on the x-axis and Euclidean distance on the y-axis. The
dark grey area represents the 95% confidence interval.
The linear regression model in Figure 18 indicates a negative relationship between pairs of actors’ distances in the network and the distance in their political position (a slope of -0.29), which generally is inconsistent with the argument of social contagion. The R2 of the linear regression is very poor (=0.005). This implies that the linear regression only explains less than 1% of the variation in data, thus the linear model is a poor fit for the data. The visualized data points in Figure 18 supports this, since there exists no clear pattern in data indicating a linear trend. Nor any other trend for that matter. The only other type of regression that would have made sense under the contagion-argument would have been an exponential regression. However, the data in Figure 18 shows no obvious pattern supporting an exponential trend. Therefore, this thesis finds no clear evidence suggesting a correlation between the pairs of car manufacturers’ network distance and the distances in their political positions, hence no clear indications of actors being similar due to social contagion.
Based on the argument of social contagion, this thesis would have expected to find pairs of actors having short network distances to correspondently having short political distances – hence to be located in the in
bottom-left of Figure 18. Oppositely, pairs of actors with long distances in the network, and therefore subsequently long distances politically, would have been expected to be located in the upper-right corner of Figure 18. These expectations are not reflected in the data of Figure 1, where only a small set of actors follow these patterns.
For example, the network analysis found that the following four pairs of manufacturers having short network distances (0.125-0.167), would be likely have a similar political position, in total there are 10 pairs with such short distance. But Figure 18 reveals that this prediction is only valid for three of the pairs (Skoda-VW does not count because they are the same company), Honda-Toyota, Skoda-Toyota and Toyota-VW9. Among the other seven pairs, BMW-Toyota and BMW-Renault are actually pairs of actors having some of the longest political distances, despite having the closest social network distance. The same pattern is reflected with pairs of manufacturers having long network distances, for example Skoda-Suzuki (0.667), Mitsubishi-Suzuki (0.667), Suzuki-VW (0.667). Based on the contagion-argument, these actors would expectedly be rather dissimilar in their political position. However, Figure 1 reveals that none of the three pairs are positioned far away from each other politically. The political distance between Mitsubishi and Suzuki is actually one of the smallest indicating a high degree of political similarity. An exception is the pair with the furthest distance, Mitsubishi-Renault (0.7679), which also have a similarly far from each other political position.
BMW and Suzuki are in general great examples of manufactures who do not support the argument of social contagion. BMW was defined as one of the manufacturers with the strongest relations in the network analysis, hence the shortest average network distance, and therefore also an actor that expectedly would be similar to the other manufacturers politically. However, as the correspondence concluded, BMW has the largest average political distance to the other manufacturers, hence is politically differentiated from the other manufacturers.
Reversely, Suzuki has relatively weak ties to other manufacturers, hence a long average network distance, but a position relatively similar to other manufacturers politically, hence a short average political distance. These two manufacturers are some of the greatest examples of outliers found in Figure 18.
This thesis concludes that there are no overall indications of homogeneity in political positions between pairs of manufacturers as a result of social contagion. However, social contagion solely focuses on the dyad-relations between actors. Results from the network analysis found a high connectedness of network actors as a group, and results from the correspondence analysis found that the group of manufacturers likewise were rather homogenous politically, at least compared to actors outside of the automobile industry. The last section of the paper will translate into a discussion about the most obvious methodological limitations of the analyses
9Skoda’s distances to the other actors in the network is partly flawed, since the lack of network data on Skoda (due to Volkswagen’s ownership of the brand) has implied that Skoda has been provided the exact same network position as
in the thesis, but also a discussion about the theoretical limitations of applying social contagion to explain political similarities among the car manufacturers investigated in the specific case.
8 Discussion
The following section will reflect on the findings of this thesis by discussing the theoretical and methodological limitations. First, the section will discuss whether the theory of contagion has been a rightful theoretical ‘fit’
for the purpose of this thesis by discussing how the other theoretical network mechanisms, adaptation, capitalization and coordination, could have explained the results of the thesis. Second, the section will identify and discuss the main methodological shortcomings of the conducted analyses by focusing on the limitations related to the overall research design of the thesis. Last, the section will reflect on the limitations related specifically to the network analysis and correspondence analysis.
8.1 Theoretical limitations