Estimating the Impact of Algorithm-Driven Production Planning: Evidence from the Glass Industry

Abstract

This thesis empirically estimates the causal impact of an algorithm-driven pro- duction planning system on production stability using proprietary product- location level data provided by AGC Glass Europe. Using a staggered Di!erence- in-Di!erences estimator of Callaway & Sant'Anna (2021), this thesis recovers causal estimates by exploiting the staggered adoption of the algorithm across seven production facilities over a twelve month period. The results indicate that algorithmic planning causally reduces the probability of stock shortages by approximately 3 percentage points, on average. Additionally, the algorithm is found to significantly reduce the magnitude of extreme overstocking events, while having no statistically significant e!ect on routine inventory fluctuations or extreme understocking. These findings suggest that the algorithm's stabiliz- ing impact is concentrated in reducing probability of a shortage and preventing extreme overstocking episodes rather than producing uniform improvements across all dimensions of supply chain stability. Keywords algorithmic planning, production automation, glass industry, inventory instability, staggered Di!erence-in-Di!erences Title Estimating the Impact of Algorithm-Driven Pro- duction Planning: Evidence from the Glass In- dustry Abstrakt Tato práce...

This thesis empirically estimates the causal impact of an algorithm-driven pro- duction planning system on production stability using proprietary product- location level data provided by AGC Glass Europe. Using a staggered Di!erence- in-Di!erences estimator of Callaway & Sant'Anna (2021), this thesis recovers causal estimates by exploiting the staggered adoption of the algorithm across seven production facilities over a twelve month period. The results indicate that algorithmic planning causally reduces the probability of stock shortages by approximately 3 percentage points, on average. Additionally, the algorithm is found to significantly reduce the magnitude of extreme overstocking events, while having no statistically significant e!ect on routine inventory fluctuations or extreme understocking. These findings suggest that the algorithm's stabiliz- ing impact is concentrated in reducing probability of a shortage and preventing extreme overstocking episodes rather than producing uniform improvements across all dimensions of supply chain stability. Keywords algorithmic planning, production automation, glass industry, inventory instability, staggered Di!erence-in-Di!erences Title Estimating the Impact of Algorithm-Driven Pro- duction Planning: Evidence from the Glass In- dustry Abstrakt Tato práce...