The search-transfer problem: the role of weak ties in sharing knowledge across organization subunits

Administrative Science Quarterly, March, 1999 by Morten T. Hansen

In addition to these contacts, there were some other regularly occurring ties between the divisions. I identified two such relations from my initial interviews: licensing agreements between divisions and enduring cross-divisional groups that were focused around a specific technology area. I included these two types of fairly informal relations so as not to bias the network information. Sometimes a regular informal contact between groups of people from two divisions had evolved into a licensing agreement, becoming a taken-for-granted and routinized knowledge-sharing activity. Information on licensing agreements between divisions was furnished by the legal department, and data on ongoing technology groups were provided by the three senior corporate R&D managers. These relations were coded 1 if any two divisions had a licensing agreement or participated together in a technology group, and 0 otherwise. The interunit network by and large comprised informal contacts as captured in the network survey. The network in 1993 consisted of a total of 225 relations, of which 201 were regularly occurring informal contacts.

Merging Project and Network Data

I merged the project data with the divisional network data by assigning a division's network relations to its projects that were included in this study. Thus, interdivisional ties became the equivalent of interdivisional project ties. It is important to record the values on the network variables prior to the start of a project because my theoretical arguments assume that a project team uses established preexisting interunit ties to search for and transfer knowledge. I handled this issue by measuring the interdivisional network relations and other division-level variables over several years. These variables were lagged by one year before they were merged with the project data.

Measuring network data over multiple years was a two-step procedure. First, as described above, I asked the divisional R&D managers about regularly occurring ties. It is common practice in social network research to assume that actors are reasonably likely to recall regularly occurring relations, as opposed to specific ad hoc interactions (Marsden, 1990). Second, following the approach of Burt (1992: 173) and Podolny and Baron (1997), I then asked how many years each of these reported ties had been in existence. Thus, for example, I know whether a particular tie existed for the 1993 network, which I used to construct the network positions for projects that started in 1994. This procedure thus generated time-varying network data for the last four years from information that the respondents could recall.

Dependent and Independent Variables

Project completion time. I measured project completion time, the dependent variable, at the project level. Time to project completion is the number of months from the start of concept development to the time of market introduction for a given project (or time to the end of the study period or cancellation for on-going and canceled projects, respectively). I defined starting time as the month when a dedicated person started working part or full time on the project, which typically coincided with the time an account was opened for the project. I defined the end date as the date on which the product was released to shipment, which is a formal milestone date in this company because it signifies that the product is ready to be manufactured and shipped on a regular basis. These definitions turned out to be very clear and provided few problems in specifying the start and finishing times.