Geography Reference
In-Depth Information
Table 16
.
9
Socially connected citing (control) patents: frequency of geodesic distances
from the originating patent (movers)
Geodesic distance
Citing
Control
1-5
34.42
16.13
6-10
36.28
38.71
11-20
27.44
43.87
>20
1.86
1.29
Mean
8
10.19
Median
7
10
Std deviation
5.18
4.9
Obs = 215 citing (155 controls).
Table 16
.
10
Geographical matching at MSA level, for movers' current and prior
location: % frequency and test of proportions*
No. of obs
Citing: % match
Control: % match
z
- statistic (
P
>
z
)
Movers' current location
All
594
7.74
8.42
−0.43
Connected
272
10.66
8.82
0.72
Disconnected
322
5.28
8.07
−1.42
Movers' prior location
All
594
11.28
7.58
2.18***
Connected
272
13.97
8.46
2.04**
Disconnected
322
9.01
6.83
1.02
Notes:
* Statistically signii cant at 1%***, 5%**, 10%*. One tail
z
- test
P
(Z >
z
).
cent), and not when there is no social connection (9 per cent vs. 7 per cent, but the
z
- test
suggests the dif erence is not signii cant).
These results look sharper than those achieved by Agrawal et al. (2006), both for
purely statistical reasons (the smaller sample) and for substantial ones.
Our smaller sample may explain why we see the citing patents' 'co-location' premium
for movers' current locations to become insignii cant, while Agrawal et al. found it to
diminish, but still resist.
As for prior locations, we do not only i nd coni rmation of Agrawal et al.'s result of higher
co-location of citing patents; we also i nd that this holds only for connected patents. This
suggests that the social ties that bind movers to their prior locations originate from within
their professional networks, that is they can be traced back to their former co-inventors,
and to those co-inventors' collaborators up to the sixth degrees of separation or so.
5. Conclusions
Our analysis has shown that the results obtained by both the original JTH experiment
and its variation by Agrawal et al. (2006) can be largely explained by the importance of
social ties from within the network of inventors.
