DSpace 立正大学 日本語

立正大学学術機関リポジトリ >
学部 >
経済学部 >
立正大学経済学会ディスカッション・ペーパー >
No.5 A Compositional Data Analysis of Market Share Dynamics >

Please use this identifier to cite or link to this item: http://hdl.handle.net/11266/6602

Title: A Compositional Data Analysis of Market Share Dynamics
Authors: Arata, Yoshiyuki
Onozaki, Tamotsu
Keywords: Market share dynamics
Subbotin family
Compositional data analysis
Issue Date: Dec-2018
Publisher: 立正大学経済学会
Abstract: The existing literature has shown that there are several statistical regularities in industrial dynamics, which are an important clue to understanding for the underlying mechanism. This paper focuses on market share changes and shows that its distribution has some remarkable properties. Because of the constrained nature of market share, that is, the sum must be unity, this paper applies the recently developed method called compositional data analysis (CDA) to market share data. We find the distribution does not follow a Gaussian but a tent-shaped distribution with a fatter tail, which is closely related with the findings of firm growth rate distribution. With some exceptions, this statistical feature can be observed across different sectors. Furthermore, this property can be observed when we focus on the relation between the top subgroup and lower-ranked firms. This shape of the distribution implies that market share growth cannot be described by an accumulation of small shocks: Rather, lumpy jumps that transform the market structure are crucial in market share dynamics. Put differently, radical change in market structure is a relatively frequent phenomenon. Such implications based on statistical properties of observed data help us further investigate industrial dynamics theoretically.
URI: http://hdl.handle.net/11266/6602
Appears in Collections:No.5 A Compositional Data Analysis of Market Share Dynamics

Files in This Item:

File Description SizeFormat
DP5(2018年度小野崎先生).pdf938.88 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.