Forecast-Informed Wasserstein Distributionally Robust Optimization for Capacity Investment Under Trade Policy Uncertainty

Long-term capacity investment under trade-policy uncertainty requires decisions before future tariffs and market conditions can be reliably characterized. This study develops a forecast-informed two-stage Wasserstein distributionally robust optimization model linking predictive information to capacity-mode and capacity-level decisions. Rolling market-state forecasts and synchronized market-state residuals are combined with an event-based tariff distribution to construct period-specific conditional reference distributions, while a shared multi-period Wasserstein budget represents distributional ambiguity around these reference distributions. The model compares rigid export, destination localization, and source-region flexible capacity, with operating decisions adjusted after uncertainty is realized. Under standard compactness and continuity conditions, the model admits a Wasserstein dual reformulation and an optimal capacity solution. The ambiguity specification is calibrated using historical decision performance and further examined through robustness diagnostics. In the battery-electric-vehicle application, rigid export, destination localization, and flexible capacity are selected for 38, 10, and 5 of 53 origin–destination pairs, respectively. The five highest-ranked tariff-risk pairs all select localization, whereas flexible-capacity value depends jointly on destination-market uncertainty, origin-market absorption, cross-market dependence, and flexibility costs. A chronological 2024 out-of-sample evaluation shows that replacing the rolling-mean origin-market forecast with histogram-based gradient boosting increases mean realized net value by USD 4.587 million per decision, with exploratory block-bootstrap intervals remaining above zero across alternative block lengths. The framework links conditional uncertainty construction, predictive information, and robust capacity investment in a unified prescriptive model.

Authors

Institutions

Publication Details

Journal
Axioms
Published
2026-11-18
DOI
https://doi.org/10.3390/axioms15110863
Primary Topic
Risk and Portfolio Optimization
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Forecast-Informed Wasserstein Distributionally Robust Optimization for Capacity Investment Under Trade Policy Uncertainty

Xiaowei Li, Juliang Zhang
Axioms
Risk and Portfolio Optimization
article

Forecast-Informed Wasserstein Distributionally Robust Optimization for Capacity Investment Under Trade Policy Uncertainty

Xiaowei Li, Juliang Zhang
article en

Abstract

Long-term capacity investment under trade-policy uncertainty requires decisions before future tariffs and market conditions can be reliably characterized. This study develops a forecast-informed two-stage Wasserstein distributionally robust optimization model linking predictive information to capacity-mode and capacity-level decisions. Rolling market-state forecasts and synchronized market-state residuals are combined with an event-based tariff distribution to construct period-specific conditional reference distributions, while a shared multi-period Wasserstein budget represents distributional ambiguity around these reference distributions. The model compares rigid export, destination localization, and source-region flexible capacity, with operating decisions adjusted after uncertainty is realized. Under standard compactness and continuity conditions, the model admits a Wasserstein dual reformulation and an optimal capacity solution. The ambiguity specification is calibrated using historical decision performance and further examined through robustness diagnostics. In the battery-electric-vehicle application, rigid export, destination localization, and flexible capacity are selected for 38, 10, and 5 of 53 origin–destination pairs, respectively. The five highest-ranked tariff-risk pairs all select localization, whereas flexible-capacity value depends jointly on destination-market uncertainty, origin-market absorption, cross-market dependence, and flexibility costs. A chronological 2024 out-of-sample evaluation shows that replacing the rolling-mean origin-market forecast with histogram-based gradient boosting increases mean realized net value by USD 4.587 million per decision, with exploratory block-bootstrap intervals remaining above zero across alternative block lengths. The framework links conditional uncertainty construction, predictive information, and robust capacity investment in a unified prescriptive model.

AxiomsVol. 15(11)
Beijing Jiaotong University (CN)
National Natural Science Foundation of China
Industry, innovation and infrastructure
Risk and Portfolio Optimization
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon