while Model 2 represents the best specification obtained. The econometric results support the hypothesis that relatedness among sectors depends on:
industry-specific factors. The proximity in the productive chain, VERT, shows a positive coefficient significantly different from zero at P B 0.01 in both the
models. Likewise, the proxy for the market similarity, MKT, is always significant at P B 0.01;
the firm-specific dimension. The variable COMPET is always significant at P B 0.01 in both the models.
Interestingly, the technology-specific dimension proxied by TECH –LINK does not seem to influence the dependent variable significantly. Nonetheless, that might
well reflect shortcomings in the measurement of technological specificity.
5. Coherence in technological and product diversification
The analysis of relatedness among sectors paves the way for the empirical investigation of coherence in the diversification patterns pursued by the firm.
Since coherence has been defined as the presence of relatedness in firms’ lines of business, it increases as the number of common characteristics found in each
product line increases. Therefore, a firm fails to exhibit coherence when com- mon characteristics and competencies are allocated randomly across its lines of
business.
Following Teece et al. 1994, coherence in the firm diversification patterns of product activities is defined as the firm’s co-presence in related sectors. A possible
measure of coherence can be the weighted average relatedness in products WARP. Specifically, WARP
i
is defined as the relatedness of sector i to all the other sectors in which the firm is active:
WARP
i
=
j
RELAT
ij
s
j j
s
j
where s
j
represents the firm’s sales in sector j. Likewise, we could define the coherence for the firms’ technological diversifica-
tion as the weighted average relatedness in technologies WART. Specifically, WART
i
is defined as the relatedness of the technological field i to all the other fields in which the firm patents. WART
i
is defined as follows: WART
i
=
j
COMPET
ij
p
j j
p
j
where p
j
is the number of patents granted to the firm in technological field j. In order to investigate the dynamic aspects of coherence in firms’ diversification
patterns, both WARP
i
and WART
i
have been calculated in three different years 1977, 1986, 1995.
Concerning the firms’ product diversification, the average value of coherence decreases over the three periods. Indeed, Table 4 shows that the average value of
WARP decreases from 11.12 to 5.09 over the whole period. Additionally, the
number of firms showing an index above the average also decreases, thus confirm- ing a decreasing coherence over time.
On the contrary, technological diversification in the same period remains a rather coherent process. The average value of WART tends to remain rather stable over
the whole period, while the number of firms with a coherence index above the average slightly increases.
This preliminary descriptive statistical analysis highlights that corporate coher- ence is increasingly associated with the firms’ moves into new technological areas
more than into new product activities. Additionally, the analysis can offer suggestions about the dynamics of coherence
in the decades considered. The analysis of the correlation matrix for those variables in each period see Table 5 suggests that coherent strategies pursued by firms are
cumulative or path-dependent. In particular, this result holds more strongly at the beginning of the whole period considered for the diversification of product activities
the correlation coefficient between WARP
t1
and WARP
t2
is 0.742, while it is only 0.206 between WARP
t2
and WARP
t3
. Conversely, coherence in the firms’ diversifi- cation patterns of technological activities seems to be higher in the most recent
period, though it is also considerable in the previous one the correlation coefficient
Table 4 Coherence in product diversification WARP and in technological diversification WART, in the
three periods considered Technological diversification WART
Product diversification WARP t2
t1 t3
t2 t3
t1 243
243 240
No. observations 200
213 217
11.57 10.08
10.39 Mean
11.12 10.79
5.09 2.67
2.23 6.42
6.91 S.D.
7.55 8.40
2.78 1.00
0.00 Min
− 2.86
− 6.74
− 3.82
56.95 31.48
29.00 79.52
Max 21.39
27.57 Coherence\mean 6alue
77 95
133 100
No. of firms 141
99 50.00
55.42 58.02
44.60 41.42
35.48 Table 5
Correlation matrix: product and technological coherence in each period WARPt1
WARPt2 WARPt3
WARTt1 WARTt2
WARTt3 WARPt1
1 WARPt2
1 0.742
0.206 0.258
WARPt3 1
WARTt1 0.020
0.030 0.270
1 0.433
1 WARTt2
0.026 −
0.032 0.215
0.413 0.758
1 WARTt3
− 0.161
− 0.137
0.192
Table 6 Results of Kolmogorov–Smirnov tests
a
WARPt3 WARTt2
WARTt3 WARPt2
0.083 0.424 WARPt1
0.421 0.000 WARPt2
0.235 0.000 WARTt1
0.042 0.987 WARTt2
a
Notes: the adjusted P-value are reported in parentheses two-sided test. Significant at PB0.05.
is 0.758 between WART
t2
and WART
t3
, and 0.433 between WART
t1
and WART
t2
. In order to strengthen this result, we run the Kolmogorov – Smirnov test
11
for the equality of coherence patterns over the period considered see Table 6. It emerges
that the coherence in diversification of product activities is more persistent at the beginning of the period considered. The equality of the two distributions WARP
t1
and WARP
t2
can not be rejected, while it is rejected between WARP
t2
and WARP
t3
at P B 0.05. On the contrary, coherence in the firms’ diversification patterns of technological activities is persistent in the most recent period. The observed data
provide some evidence in support of the null hypothesis that is, the equality of the distributions WART
t2
and WART
t3
, while they do not seem to sustain it in the first period in which the null hypothesis is rejected at P B 0.05.
The findings support the view of the firm as a repository of accumulated competencies which are developed by firms themselves through a gradual learning
process. This accumulation process takes place within the firm and provides it with the characteristics of technological persistence or path dependency Fai, 1998;
Cantwell and Fai, 1999a,b.
6. Conclusions