We tested tbe robustness of tbe cluster solution through
cluster membership comparisons in two ways. First, we
determined the proportion of firms tbat cbanged clusters
from membership assigned during the bierarcbical
and k-means clustering procedures. Of the 371 firms,
16% changed cluster membership from that assigned
during the bierarchical clustering phase to that assigned
during tbe k-means clustering procedure. Second, we
used discriminant analysis to classify firms into groups
and group assignment compared witb the cluster solution.
A high proportion of firms were correctly classified in the discriminant analysis with 98.1% of firms being
classified into tbe same groups. The results of the robustness
procedures indicate an acceptable level of cluster
reproducibility.