LAUSR

Firms in advanced economies are increasingly outsourcing software and technology development as well as other knowledge work to a worldwide supply base. Standard economic and learning models predict that focal firms should outsource either all or none of a particular activity unless extra resources are required during cyclical demand peaks or access is needed to some tightly appropriable intellectual property. However, recent evidence shows that, even when these exceptions do not apply, many firms pursue a partial outsourcing strategy. We develop a dynamic optimization model to provide a rational explanation for this observation. In our model, learning from prior projects occurs at both the subsystem level and the overall systems level (e.g., systems integration and architecture). Learning at the two levels interacts such that integration capabilities can dynamically build and decay. The model generates conditions where partial outsourcing is rational and dominates the extreme conditions of complete insourcing or complete outsourcing. Our model also specifies the conditions for regime change between insourcing and outsourcing and cycling between insourcing and outsourcing, and overshooting or undershooting the long‐run outsourcing target. Furthermore, we show that these results are highly path‐dependent under short horizons. The model also provides explanation for interesting questions such as why the rate of outsourcing might be U‐shaped in the rate of technological change and why startups so often insource in contrast to more established counterparts in similar industries.