PRICE DISCOVERY BETWEEN CARBONATED SOFT DRINK MANUFACTURERS AND RETAILERS A DISAGGREGATE ANALYSIS WITH PC AND LiNGAM ALGORITHMS》.pdf

Journal of Applied Economics. Vol XVIII, No. 1 (May 2015), 173-198 PRICE DISCOVERY BETWEEN CARBONATED SOFT DRINK MANUFACTURERS AND RETAILERS: A DISAGGREGATE ANALYSIS WITH PC AND LiNGAM ALGORITHMS Pei-Chun Lai∗ Capital University of Economics and Business DaviD a. BessLer Texas AM University Submitted October 2013; accepted August 2014 This paper considers the use of two machine learning algorithms to identify the causal relationships among retail prices, manufacturer prices, and number of packages sold. The two algorithms are PC and Linear Non-Gaussian Acyclic Models (LiNGAM). The dataset studied comprises scanner data collected from the retail sales of carbonated soft drinks in the Chicago area. The PC algorithm is not able to assign direction among retail price, manufacturer price and quantity sold, whereas the LiNGAM algorithm is able to decide in every case, i.e., retail price leads manufacturer price and quantity sold. JEL classification codes : C1, L1 Key words: causal relationships, PC algorithm, Linear Non-Gaussian Acyclic Models (LiNGAM) I. Introduction This paper addresses price discovery at the interface between two manufacturers and one retailer of carbonated soft drink (CSD) products. We use two algorithms from the field of machine learning: the PC algorithm and a recently developed Linear Non-Gaussian Acyclic Model (LiNGAM). We apply both of them to infer * Pei-Chun Lai (corresponding author): International School of Economics and Management, Capital University of Economics and Business, Beijing 100070, China; e-mail: pclaitamu@. David A. Bessler: Department of Agricultural Economics, Texas AM University, College Station, TX 77843, USA; e-mail: d-bessler@. We are indebted to Shohei Shimizu a