The more you buy, the more you pay? Commodity procurement under price uncertainty and market impact
研究了有市场影响力的买家在商品采购中如何制定最优策略，发现忽略自身市场影响会导致显著损失，并提出了基于机器学习的最优库存策略。
We examine a commodity procurement problem for a price-setting actor with significant market impact, where decisions influence the market price, resulting in an upward-sloping price curve. Most analytics do not yet address scenarios involving a price-setting buyer, eg, in markets lacking liquidity or when a decision-maker purchases a substantial share of a commodity’s volume. We study such procurement settings using an empirical study on three selected commodities, ie, natural gas, copper, and soybean. For each, we consider a real-world setting and apply market liquidity data to evaluate the market reaction based on the findings of finance literature. The prescriptive modified base-stock policy employed uses external feature data to derive inventory target levels in a data-driven manner. We find that simple smoothing approaches outperform a naïve hand-to-mouth policy, while the developed policy incorporating covariate feature data performs best. Moreover, we show that ignoring one’s market impact leads to significant losses. Our research extends the existing literature by analysing the optimal procurement policy under market impact, modelling market behaviour based on finance and econometrics literature, and employing a new machine-learning-based approach to optimize policy parameters.