Geography Reference
In-Depth Information
Table 16
.
5
Sample of cited, citing and control patents: summary statistics
No. of
patents
No. of
patent triples
No. of
inventors
No. of
inventors
per patent
% of connected
triples
Cited
1933
-
4375
2.96
-
Citing
2868
10988
7436
3.45
35.0
Control
3283
10988
8690
3.24
24.2
Table 16
.
6
Socially connected citing (control) patents: frequency of geodesic distances
from the originating patent
Geodesic distance
Citing
Control
1-5
24.40
15.44
6-10
49.6
45.29
11-20
23.3
36.14
>20
2.7
3.13
Mean
8.72
10.01
Median
8
9
Std deviation
4.94
4.93
Obs = 3894 citing (2656 controls).
Social ties in citing vs
.
control patents
Social connections between inventors are conducive to knowledge dif usion. It has been
shown that the probability of observing a citation link between any pair of patents with
dif erent dates is a positive function of the existence of a personal or social tie between
the two patents (Singh, 2005). In particular, social links matter when they are very close,
that is when they exhibit low (≤5) geodesic distance (Breschi and Lissoni, 2004).
As a consequence, we expect that the probability of observing a personal or social link
within a cited-citing patent pair is higher than the probability of observing a similar link
in a cited-control pair. Table 16.5 shows that 35 per cent of citing patents are linked to
originating ones through a social chain of i nite length; the same value for control patents
is only 24 per cent.
Table 16.6 shows that the mean geodesic distance for connected patent pairs is signii -
cantly lower for citing patents (8.72) than for control ones (10.01). While over 24 per cent
of social ties between citing and originating patents is below six degrees of separation, the
same i gure for the control-cited pairs is under 16 per cent.
Finally, we have calculated the geographical co-location of both the citing and the
control patent samples, with respect to the originating patents. As expected, closely con-
nected patent pairs, both in the citing and the control sample, are highly co-localized
with the cited one. For inventors at less than six degrees of separation, the percentage of
co-location is around 29 per cent both between citing-cited patents and for control-cited
patents (Table 16.7). The only major dif erence emerges for connected patents at more
