Innovation type and external knowledge search strategies in KIBS: evidence from Canada

The sample comprises both innovative and non-innovative KIBS. This approach allows to gather information on a large variety of aspects related to innovation, including its activities, sources, and determinants. If the potential group is restricted to innovating KIBS only, the effects can be underestimated, because the innovation status of a KIBS can change from non-innovative to innovative and vice versa. It is possible that the 3-year backward looking window on innovation outputs limits to capturing all KIBS establishment-level innovations that can be attributed to the adoption of open innovation strategies over the period. At the same time, having in the same the firms that did not introduce innovations in this 3-year period increases variance and helps to better establish the relationships between innovation patterns and external sourcing and external partnering.

In the paper, we do not study whether the various types of innovation are themselves correlated—the principal objective of the paper is to examine whether different types of innovation require distinct kinds of external knowledge search strategies. But as Amara et al. (2009) have shown, KIBS that implement one type of innovation often implement other types of innovations. This should be borne in mind when interpreting the results.

We removed seven outliers using a decision tree algorithm (John 1995). The final number of observations in our sample is 385.

All the new models show good model fit except for the T-KIBS and external sourcing, which suffers from a poorer model fit. At the same time, using other statistical techniques (e.g. operationalizing the dependent variable as low–high sourcing and using probit model) yields similar coefficients, which again, points to the robustness of the results.