I show that the number of positive shares in the input-output (production network density) matters for volatility. Service oriented economies with denser connections display lower volatility, while denser manufacturing oriented economies display higher volatility. I build a multisector model with non-unitary elasticity of substitution between labor and intermediates and show that with an elasticity above one, as in service sectors, a higher network density diversifies shocks, while with an elasticity below one, as in manufacturing sectors, a higher density amplifies shocks. Calibrating the model for the countries in the sample yields a density-volatility correlation similar to the data.