Introduction

Industrial policies are increasingly used to reshape plant location patterns in industries with multi-stage production and economies of scale. This research develops a quantitative method to analyse these policies, specifically applied to the battery electric vehicle (EV) industry. The paper addresses the challenge of characterising global supply chains, an NP-hard facility location problem, by adapting mixed integer linear programming (MILP) methods. Significant recent policies include the 2022 US Inflation Reduction Act (IRA) and Canada’s commitment of roughly $30 billion in production subsidies.

Literature

The study sits at the intersection of global value chain (GVC) modelling, plant location sourcing with interdependencies, and recent EV-directed industrial policies. It moves beyond previous models that assumed constant returns to scale by incorporating fixed costs, which generate increasing returns and complicate location decisions.

Description Of The Global EV Industry

The analysis utilises IHS-Markit data from 2015 to 2023 across 24 major markets. In 2023, China was the largest market in almost all dimensions, with sales of 5,353,000 vehicles and an EV share of 21.0%. In contrast, the United States had an EV share of 8.1% and Norway led with 83.7%. Capital investment for new plants is substantial; news reports indicate an average EV assembly plant involves an investment of $660 million, while an average battery cell plant requires $1.85 billion.

Facts Guiding The Model

Data indicates that single-sourcing is the dominant practice in the industry. For battery cells, the multi-sourcing share is approximately 38% at a broad level but drops to 3% when disaggregated by model, shape, material, and power. Cell plants typically serve a single carmaker, often through joint ventures, rather than supplying generic cells to multiple buyers. Additionally, 70% of cell plants produce only one material category (NMC or LFP).

Multi-product, Multi-stage, Multinational Production

The researchers specify a model of multinational firms’ plant location and sourcing decisions. Firms maximise variable profits net of activation costs across multiple production levels. Using the MILP formulation, the model solves complex combinatorial problems involving approximately 1.93 x 10^25 possible configurations in roughly one second.

Estimation Of Path Costs Via Sourcing Regressions

Sourcing decisions are used to estimate costs related to distance, borders, and trade agreements. Border effects are significant, with results implying domestic sourcing is 2.6–2.8 times as likely as importing. Tariff elasticities were estimated at -8.49 for battery cells and -8.56 for vehicles. Geographic proximity is a critical factor, and distance travelled by components fell substantially between 2015 and 2023.

From Path Costs To Variable Profits

The model uses a constant elasticity of substitution (CES) form for demand, setting the elasticity at 4.0 based on a median of 18 literature estimates. Hat algebra is employed to recover variable profits in levels, allowing market shares to absorb demand shifters and cost parameters that are difficult to calibrate directly.

Simulated Method Of Moments Estimation

The Simulated Method of Moments (SMM) targets 47 moments, including dyadic flows and production line counts. Results show that fixed costs increase with distance to headquarters and vehicle quality. Fixed costs for assembly and cell plants are estimated to be significantly lower in Asia than in North America or Europe.

Counterfactual Policy Simulations

The report simulates three main scenarios with a 20% subsidy rate. Policy 1 (unconditional subsidy) is the most effective at expanding EV adoption, increasing North American EV expenditure by 86% and production lines in the Americas by 16% for assembly and 7% for cells. Policy 2 (domestic assembly requirement) increases assembly lines by nearly a third but reduces the adoption stimulus to 71%. Policy 3 (domestic supply chain requirement) triples cell production expansion to 27% but adoption stimulus falls further to 32%, effectively undoing more than half of the stimulus from pure subsidies.

In Europe, Policy 2 generates a 40% increase in vehicle production lines. Europe is notably more competitive in cell manufacturing, with average costs 39% lower than those in North America and only 6% higher than in Asia.

Conclusion

Increasing returns to scale are empirically important in the EV industry. A 20% unconditional subsidy decreases consumer prices by 25% for North America and 23% for Europe due to production path optimisations. While local content requirements can achieve protectionist goals, they drive up costs and reduce overall EV adoption. The assembly contingency achieves protectionist aims but shrinks cost savings, while full local content requirements fail to deliver significant assembly relocation and eliminate cost savings.