Input Diffusion and the Evolution of Production Networks

What determines which inputs are initially considered and eventually adopted in the production of new or improved goods? Why are some inputs much more prominent than others? We model the evolution of input linkages as a process where new producers first search for potentially useful inputs and then decide which ones to adopt. A new product initially draws a set of ‘essential suppliers’. The search stage is then confined to the network neighborhood of the latter, i.e., to the inputs used by the essential suppliers. The adoption decision is driven by a tradeoff between the benefits accruing from input variety and the costs of input adoption. This has important implications for the number of forward linkages that a product (input variety) develops over time. Input diffusion is fostered by network centrality – an input that is initially represented in many network neighborhoods is subsequently more likely to be adopted. This mechanism also delivers a power law distribution of forward linkages. Our predictions continue to hold when varieties are aggregated into sectors. We can thus test them, using detailed sectoral US input-output tables. We show that initial network proximity of a sector in 1967 significantly increases the likelihood of adoption throughout the subsequent four decades. The same is true for rapid productivity growth in an input-producing sector. Our empirical results highlight two conditions for new products to become central nodes: initial network proximity to prospective adopters, and technological progress that reduces their relative price. Semiconductors met both conditions.