Production network data register each firm at a single headquarters address, even when it operates establishments in several locations. Previous research shows that this headquarters effect biases the share of the network reached by a simulated disruption, but not how much value is misestimated or where the error arises.
We show that the headquarters effect biases estimated disruption losses mainly through propagation along supply chains, rather than through the initial hit. Using link probabilities from a logit model estimated at the firm level, we disaggregate transaction values proportionally across establishment pairs. Unlike existing rule-based allocations, our approach needs no data on input requirements, extends beyond manufacturing, and propagates estimation uncertainty to the results. It is validated on a synthetic benchmark and applied to the Dutch interfirm network.
Under localized shocks, the bias in chemical-sector losses has no fixed sign: establishment-level losses are 29% lower than headquarters-level estimates for Rotterdam and 28% higher for Amsterdam. Of these differences, 93% and 57% respectively arise during propagation. At least 7% (€42.5bn) of all transaction value is attributed to a different province than the headquarters view implies. Supply chain risk assessments based on headquarters-registered data may therefore misstate the size and direction of regional exposure.
About Leonie Beck
Leonie Beck received her BA in Business Administration from Hochschule Heilbronn, Germany. In 2026 she graduated from the MSc Transport and Supply Chain Management (cum laude) at Vrije Universiteit Amsterdam. Her master’s thesis, carried out in collaboration with Statistics Netherlands (CBS), reconstructed the Dutch supply network at the level of individual establishments. She currently holds an appointment at the Department of Operations Analytics of Vrije Universiteit Amsterdam, where she is developing this work into a journal article with VU and CBS.