Technology from external sources can be accessed through a multiplicity of relationships, ranging from industrial theft to mergers and acquisitions (Leonard Barton, 1995). As we are concerned with research spillovers, the focus of this research is relationships that are based on informal or non–contractual relationships between the provider and receptor of the information.

To test the effect of external knowledge acquisition, and intra–firm knowledge dissemination, on research productivity, new measures were developed for this study. To insure the reliability and validity of the measures, multi–item scales were developed (Nunnally, 1978). Individual item fluctuations cancel each other out in multi–item scales (Marsden, 1990). A broad and thorough literature review was used to define the scope of activities and generate the initial items to measure these constructs. The results of this identification were combined with available questionnaires dealing with external knowledge acquisition and intra–firm knowledge dissemination (e.g. Narver and Slater, 1990). The set of questions was then refined in interviews with industry participants and experts. Questions were designed to be easy to understand, concise, and neutrally worded. Each question relied on a single balanced seven–point LIKERT scale that was anchored at each end to simplify scoring.

The final element of absorptive capacity, technical capabilities, represents the existing research knowledge that resides within the organization. While organization knowledge is attracting the attention of numerous researchers, a review of the literature discerned no accepted metric for assessing technical knowledge. Following the research of Pakes and Griliches (1984), R&D expenditures are assumed to reflect the current state of technical abilities within the firm.

The research hypotheses are investigated by using a systematic varying parameter model to examine the relationship between firm sales and research expenditures. Sales as the output of research expenditures reflects the productivity of the firm’s research program, as productivity is the function of outputs to inputs (Kendrick, 1977). The relationship between sales and research expenditures is expressed as a simple Cobb–Douglas production function:

Taking the log of both sides gives:

...........................................(1)

The subscript *i* refers to the business unit, and *t*
reflects the time period. The error term eit
reflects the influence of omitted factors on the response variable. Some of these factors
may be specific to a particular organization or business unit and, as such, the error term
can be expressed as eit
= mi + hit,
where mi is an organization specific
factor that is time invariant and hit
reflects a non–autocorrelated contemporaneous shock (Jacobson, 1994). If mi and hit
are uncorrelated with the explanatory variables, generalized least squares (GLS)
regression produces consistent estimates (Chamberlin, 1984).

The assumption of zero correlation between the explanatory variables and the unobserved business–specific effect has been criticized from both theoretical and empirical grounds (Jacobson, 1994). The resource–based view of the firm highlights the potential of invisible (unobservable) assets as scarce, valuable, and costly–to–copy inputs to production (Wernerfelt, 1984), and Jacobson (1990) has shown empirically that failure to control for unobservable factors can generate biased estimates of the parameters and invalidate the findings. To account for time invariant unobserved business unit factors, equation (1) is modified as follows:

.....................................(2)

In equation (2), the unobserved business–unit factor mi is allowed to be correlated with the explanatory factors. If mi is correlated with the observed variables included in the model, both ordinary least squares (OLS) and GLS regression will generate biased and inconsistent coefficient estimates (Jacobson, 1994). To obtain consistent estimates of the effects of the observed explanatory variables in the presence of unobserved fixed effects, a first difference model is estimated:

...(3)

All factors constant for a business unit at adjacent time periods, which included fixed effects, will be removed by taking first differences. Equation (3) is equivalent to the following growth model which, to aid interpretability of the results, was used in this research:

One consequence of this formulation is that the effect of technical capabilities is not separable from the general effect of R&D expenditure growth and Hypotheses 3 is not tested, however, the gains in being able to control for firm specific time invariant effects clearly outweigh this limitation.

To investigate whether the explanatory variables impact sales growth directly, the following model is estimated:

*EKAi,t2* refers
to the external knowledge acquisition activities of firm *i *in time period *t**2*, *IFKD**i,t2* refers to intra–firm knowledge dissemination
activities in business unit i in time period *t**2*. Both sales and R&D
growth are measured from time *t**1 *(1993) to *t**2* (1995).

To assess the effect of business–unit absorptive capacity on the
productivity of research expenditures the impact of the explanatory variables on the
coefficient of R&D growth, *b1,*
is investigated by estimating the following model:

Top of page | Chapter Listing | Return to 1997 Topical Index |

© 1997 Babson College All Rights Reserved

Last Updated 03/19/98