When Sectoral Recovery Fails to Aggregate

Does monotone sectoral recovery survive aggregation? With fixed nonnegative weights, it does. This paper studies the same preservation question when aggregation weights are endogenous Domar weights generated by a production network with gradual pass-through of intermediate-input cost changes. The aggregate derivative decomposes into a recovery push and a network drag, and the resulting reversal index gives an exact condition: aggregate displacement deepens if and only if the drag exceeds the push. This characterisation yields a sharp preservation boundary. Among nonnegative input-output matrices with spectral radius below one, the only matrix that preserves recovery for every admissible sectoral path is the no-network matrix $\Omega=0$. The drag passes through the network one more time than direct sectoral recovery, so it scales at a higher order in network connectivity. Two channels contribute in levels: growth of total Leontief-weighted cost exposure as pass-through rises, and reweighting toward sectors further behind in recovery. In an illustration using the 55-sector US input-output matrix and cross-sector robot density data, the drag peaks at nearly eight times the direct recovery effect under benchmark assumptions.