On empirically estimating bullwhip effects: Measurement, aggregation, and impact

Abstract Empirical estimation of the bullwhip effect poses several challenges, although the bullwhip effect has been well studied in modeling papers. Using a dataset from a large supermarket chain, we estimate the product level bullwhip effect using various methods, analyze consequences of its different measurements and aggregations, and examine its impact on supply chain performance in terms of inventory ratio and stockouts. We have three major findings. (a) Bullwhip effect estimates exhibit different magnitudes dependent on how they are measured. The material bullwhip effect is greater in magnitude than the information bullwhip effect in our data, where demand correlations are sufficiently low. (b) The aggregated bullwhip effect ratios by store and by time are lower than the disaggregated bullwhip effect ratios, indicating that the aggregated bullwhip effect ratios underestimate the bullwhip effect. The aggregated bullwhip effect ratios by product are lower than the disaggregated bullwhip effect ratios, indicating the bullwhip effect is not as strong as theory predicts due to order pooling. (c) The bullwhip effect is associated with poor supply chain performance, as measured by elevated inventory ratio and stockouts. However, if the bullwhip effect is measured inaccurately, these benefits can be underestimated as much as 75% for inventory and 25% for stockouts.