RESULTS AND DISCUSSION

The results from model 1 are reported in Table 2. The concentration of industry employment has a positive effect on the mean level of products developed per firm in an MSA in 1993, but no effect in 1994. The diversity of the economic base has no effect in either 1993 or 1994. This result is somewhat surprising because in a related study using the same data set, we found a significant relationship between the diversity of the economic base and the number of products developed (Heeley & Hill, 1996). The major difference is that in our previous paper we used county, instead of MSA, as our measure of geographic region. One possible cause for the non-significant results is that the number of regions declined from 167 (counties) to 78 (MSAs), which significantly reduces the statistical power. Finally, we find no support for hypotheses 3a and 3b. Total MSA employment has a direct effect on the mean level of products developed in 1994, but does not moderate the relationship between the concentration and diversity of industry employment and product development.

TABLE 2
Model 1 - Poisson General Linear Model
Log (Average number of New Products Developed by a firm in a MSA)

Variable

Average Number in 1993

Average Number in 1994
MSA Industry

Employment

5.00 e-5

(8.63 e-4)

5.97 e-5*

(4.34 e-5)

1.13 e-5

(1.67 e-4)

 
MSA Diversity

-4.99

(5.11)

 

-4.10

(10.91)

 
MSA Total

Employment

-1.05 e-6

(1.34 e-6)

 

3.49 e-7

(2.58 e-6)

3.97 e-7*

(2.39 e-7)

MSA Industry

Employment .

MSA Total

Employment

2.22 e-11

(6.80 e-11)

 

2.73 e-11

(1.24 e-10)

 
MSA Diversity .

MSA Total

Employment

5.39 e-6

(7.30 e-6)

 

-1.52 e-6

(1.38 e-5)

 
N

78

78

78

78

Null Deviance

123.72

123.72

92.14

92.14

Residual Deviance

118.73

120.88

86.53

88.10

Dispersion

Parameter

1.70

1.63

1.96

1.81

* significant at the 10% level, ** significant at the 5% level, *** significant at the 1% level.

The results from model 2 are reported in Table 3. This model allows us to test hypotheses 4a and 4b, and also investigate the previous hypotheses at the firm level. Firm absorptive capacity, which is operationalized as the discounted stock of medical patents, has no effect on the number of products developed in 1993, or 1994. We find the lack of a direct effect to be somewhat surprising because we expected firms with a history of successfully patenting to have an increased likelihood of developing new products.

The results from the most parsimonious model for 1993, consistent with hypothesis 2, shows that increased MSA diversity has a positive effect on the log number of products developed. However, as predicted by hypothesis 3b, this effect is moderated by total MSA employment. These results show that increased diversity has a positive effect in smaller MSAs but when MSA total employment exceeds 751773 (-6.36/-8.46e-6) increased diversity has a negative effect on the log number of products developed in 1993. This result is consistent with our theoretical arguments. Once an MSA gets too large, the benefits of diverse knowledge are exceeded by the costs of distilling large volumes of information.

TABLE 3
Model 2 - Poisson General Linear Model
Log (Number of New Products Developed by Firm)

Variable Number in 1993 Number in 1994
MSA Industry Employment

-5.40 e-5

(5.53 e-5)

-6.68 e-5

(5.43 e-5)

-1.41 e-4*

(7.91 e-5)

-9.51 e-5*

(6.30 e-5)

MSA Diversity

-5.93*

(4.79)

-6.36*

(4.61)

-7.98

(7.75)

 
Firm Absorptive Capacity

-2.96 e-3

(3.58 e-2)

-1.33 e-2

(2.72 e-2)

-4.10 e-2

(8.52 e-2)

 
MSA Total Employment

-1.87 e-6**

(8.18 e-7)

-1.90 e-6**

(8.13 e-7)

-9.17 e-7

(1.20 e-6)

2.77 e-8

(3.48 e-7)

MSA Industry Employment .

Firm Absorptive Capacity

-4.16 e-7

(1.32 e-6)

 

3.16 e-7

(3.49 e-6)

 
MSA Diversity .

Firm Absorptive Capacity

1.94 e-1

(2.64 e-1)

2.48 e-1

(1.99 e-1)

2.91 e-1

(4.99 e-1)

 
MSA Industry Employment .

MSA Total Employment

8.52 e-11**

(4.60 e-11)

8.51 e-11**

(4.58 e-11)

1.02 e-10*

(6.40 e-11)

6.39 e-11*

(4.77 e-11)

MSA Diversity .

MSA Total Employment

8.23 e-6**

(4.26 e-6)

8.46 e-6**

(4.18 e-6)

4.70 e-6

(6.03 e-6)

 
N

308

308

308

308

Null Deviance

766.78

766.78

552.45

552.45

Residual Deviance

695.01

695.31

510.50

513.83

Dispersion Parameter

3.19

3.18

2.64

2.64

* significant at the 10% level, ** significant at the 5% level, *** significant at the 1% level.

The effects of industry employment are not as clear. In both 1993 and 1994, contrary to hypothesis 1, increased industry employment has a negative effect on the log number of products developed. However, consistent with hypothesis 3a, this effect is moderated by MSA total employment. In examining this result in more detail we can see that even though increased industry employment has a negative effect in small MSAs, it is extremely close to zero and becomes positive when MSA total employment exceeds 984958 in 1993, and 1488263 in 1994.

The results of our sub-sample analysis can be found in Table 4. In the interests of brevity we will present only the significant models. There are no significant relationships in size class I which may be in part due to the small number of observations.

 

TABLE 4
Poisson General Linear Model For Size Class 3
Log (Number of New Products Developed by Firm)

 

Number in 1993

Number in 1993

Variable

Size Class II

Size Class III

MSA Industry
Employment

-1.23 e-4
(5.53 e-5)

 

2.20 e-5
(8.65 e-5)

3.80 e-5
(7.26 e-5)

MSA Diversity

-26.2**
(4.79)

-24.0***
(4.61)

1.94
(7.53)

2.64
(3.63)

Firm Absorptive
Capacity

-0.138
(0.286)

 

5.00 e-2
(5.99 e-2)

2.78 e-2**
(1.00 e-2)

MSA Total
Employment

-4.57 e-6**
(1.37 e-6)

-4.45 e-6**
(1.26 e-6)

-1.06 e-6
(1.26 e-6)

-5.87 e-7
(4.78 e-7)

MSA Industry Employment .
Firm Absorptive
Capacity

3.70 e-6

(1.61 e-5)

 

-1.91 e-6

(1.73 e-6)

-1.48 e-6

(1.19 e-6)

MSA Diversity .
Firm AbsorptiveCapacity

0.488
(1.53)

 

1.37 e-1
(3.67 e-1)

 
MSA Industry Employment .
MSA Total Employment

6.11 e-11

(6.28 e-11)

 

6.33 e-11***

(7.44 e-11)

4.80 e-11

(6.48 e-11)

MSA Diversity.
MSA Total Employment

2.70 e-5***
(7.39 e-6)

2.71 e-5***
(7.42 e-6)

2.63 e-6**
(6.54 e-6)

 
N

89

89

148

148

Null Deviance

115.17

115.17

467.93

467.93

Residual Deviance

91.05

94.6

409.74

411.86

Dispersion Parameter

1.0

1.1

4.04

3.95

* significant at the 10% level, ** significant at the 5% level, *** significant at the 1% level.

In size class II, we find a significant relationship between industry diversity and the number of products developed in 1993. As hypothesized, this relationship is negative, but consistent with hypothesis 3b, the effect is moderated by total MSA employment. In smaller MSAs, increased diversity has a positive effect on the log number of products developed. This positive effect continues until the MSA total employment exceeds 885609, when the effect becomes negative.

In size class III, we find no significant relationships between industry concentration or diversity, and the log number of products developed in 1993. However, there is a strong positive relationship between absorptive capacity and the log number of products developed. This suggests that for firms in size class III, the likelihood of developing more new products in 1993 increases, as a firm’s level of absorptive capacity increases.

As the measure for absorptive capacity, the discounted stock of medical patents, reflects the amount of knowledge a firm has internally. It appears that larger firms, with increased stocks of internal knowledge, tap internal stocks of knowledge rather than external knowledge. In contrast, smaller firms reap some benefit from tapping external stocks of diverse knowledge. For both size classes II and III, there are no significant results in 1994.

 

CONCLUSIONS AND IMPLICATIONS

The results from this study are not clear-cut. At the regional level we find no support for the existence of a relationship between knowledge spillovers and the mean level of product development in the region. This result implies that external knowledge spillover effects are dependent upon firm specific differences. However, as this result contradicts earlier studies, we suggest interpreting this result cautiously. It is quite possible, given the small number of regions in our sample (N=78), that there is not sufficient statistical power to isolate regional effects.

In our analysis of the relationship between knowledge spillovers, product development, and absorptive capacity, we find some evidence that this phenomenon exists at the firm level. The diversity of economic activity in a region impacts the number of new products developed, and this relationship is moderated by total MSA employment. In addition, we find that the results differ for our size classes, suggesting that firm size affects the abilities of a firm to benefit from knowledge spillovers. Smaller firms benefit from diverse external knowledge stocks, while larger firms benefit from internally generated knowledge.

It is clear from our analysis and results that the spillover phenomenon is complex and the relationships may not be linear. It is likely that there are a number of other factors that have the potential to impact the ability of firms to exploit external knowledge. For example, firms will differ in their ability to acquire and internally distribute external information. These factors should be included in spillover models if we are to fully understand this research question. In addition future research may benefit from using techniques such as classification trees and general additive models that are not restricted to linear functions.

Previous Page | Main Menu | Next Page

 

 

1997 Babson College All Rights Reserved
Last Updated 1/15/97 by Geoff Goldman & Dennis Valencia

To sign-up for the Center for Entrepreneurial Studies' publication lists,
please register with the
Entrepreneurship WebTeam.