TABLE OF CONTENTS One limitation of any static, "levels", study is that the coefficient on any variable reflects the average effect of the data in question. As the Internet develops and the technologies change, the relationships among these variables are likely to change with the scope of the firm (e.g., through network effects, increased efficiency or changes in browser demographics or habits) and over time, respectively. In addition, as described in Section 5, it is possible that some of the variables tested have competing effects which may confuse the results and cannot be easily modeled out, even within a path analysis framework. To examine the marginal effect of these variables, the complete set of regressions (1a) (1g) are estimated using a changes specification, where the changes are defined as the difference between the reported quarterly accounting data and its one-quarter lag value. [19]
From Table 6, Panel A, it can be seen that, under this specification, neither SG&A nor R&D is significantly associated with either time spent online or visits per person. In addition, R&D is positive but no longer significantly related to unique audience, although the coefficient for SG&A and unique audience remains positive and significant. The lack of a coefficient for changes in R&D spending suggests that additional firm spending on R&D is most likely not associated with efforts to improve website activity. Overall the results for the changes specification are not as strong as those in the prior section. The results are, nonetheless, consistent with the interpretation that...
