Correlation Between Network Variables And Market Value

1210 WordsMay 21, 20175 Pages
Table 1 reports correlation coefficients among the variables. The correlation between network variables and market value varies and, excluding Closeness, is between 0.23 and 0.32 something well below 0.80, that could be a concern for multicollinearity. However, Degree and Betweenness have correlation 0.85, indicating that the best way to examine network variables is by using one network variable in every regression, to avoid using spurious regressions. We also found that Management Compensation (Comp) has a high correlation with market value, so we replace it in our study by the Compensation over market value that does not correlate significantly with other variables. Table 4 reports estimates of the impact of high network values on…show more content…
When we consider the effect of the Non-Network variables, the coefficients of Network variables become smaller, but they still remain significant and enormous. In particular, the effect of including Non-Network variables results in a decrease of 5%-30% of the previously reported effect of High Degree and High Eigenvalue to the company value, whilst the effect over High Betweenness and High Centrality coefficients is enormous (-65% to -70%). The first conclusion is that regardless of the model used, High Degree and High Eigenvector have an enormous effect on company value, whilst the size of the effect and significance of High Centrality and High Betweenness depends on the model. To assess if the causal relationship between the network variable and the market value is valid under different assumptions and to measure if this association is consistent with a causal mechanism or is rather an effect of non-random choice, we use several methods. First, we found that the effect of the same Network variables is positive, high and significant if we also run Quantile Regressions. We also run IV-Quantile regressions to check the validity of Non-Network variables as instruments of a model that assumes that Network variables are endogenous. IV-Quantile model diagnostic tests do not give evidence of the significance for any of the examined Non-Network variables to be used as instruments of
Open Document