INTERNATIONAL KNOWLEDGE FLOWS: RESULTS 3

Geographic localization is also evident in the (3, parameters, presented in the middle panel of Table 3 in the form of the estimated modal lag. Here the diagonal elements are generally the smallest entry in each row and column, meaning modal citation lags are noticeably shorter for domestic citations, relative to citations to and from others. The only exception to this general pattern is the U.S. U.S. inventors are slightly faster to cite Japanese inventors than they are to cite U.S. inventors (p,=T.05), and Japanese inventors are faster to cite U.S. inventors than are U.S. inventors (P^l .19).

There are also systematic variations across the countries that are superimposed on top of the general pattern of localization. While the modal lags for citations made by the U.S. range from 5 to 5.4 years (depending on the cited country), those for Japan range from 3.4 to 4.4. Indeed, it appears that the overall tendency of the U.S. to both generate and receive long-lagged citations is part of the reason why U.S.-to-U.S. citations do not come more quickly than those from and to others.

The fact that domestic citations generally involve both higher a and higher p, creates offsetting effects for the overall number of citations, since the higher p, means that citations fade faster and hence reduces the total holding a constant. The bottom panel of Table 3 combines these effects by presenting the overall cumulative probabilities. The estimates show that, in terms of total citations, the variations in a dominate the variations in P,; the matrix is still strongly diagonal, indicating localization. These differences are quite significant statistically, although it should be noted that this calculation relies heavily on the assumed functional form as it integrates the citation function into the infinite future.