Price Dispersion and the Role of Stores*

• DOI: http://doi.org/10.1111/sjoe.12356
• Published date: 01 July 2020
• Date: 01 July 2020

Scand. J. of Economics 122(3), 1181–1206, 2020

DOI: 10.1111/sjoe.12356

Price Dispersion and the Role of Stores*

Espen R. Moen

BI Norwegian Business School, NO-0442 Oslo, Norway

espen.r.moen@bi.no

FredrikWulfsberg

Oslo Business School, NO-0130 Oslo, Norway

frewul@oslomet.no

Øyvind Aas

Universit´e libre de Bruxelles, B-1050 Bruxelles, Belgium

oyvind.nilsen.aas@gmail.com

Abstract

In this paper, we study price dispersion in the Norwegian retail marketfor 766 products across

4,297 stores over 60 months. Price dispersion for homogeneous products is significant and

persistent, with a coefficient of variation of 37 percent for the median product. Price dispersion

differs between product categories and overtime. Store heterogeneity accounts for 30 percent of

the observed variation in prices for the median product–month, and for around 50 percent for the

sample as a whole. Price dispersion is still prevalent after correcting for store heterogeneity.

Keywords: Price dispersion; retail prices; store heterogeneity

JEL classification:D2; D4; E3

I. Introduction

It is common knowledge that the price for a particular product or service

can vary substantially between stores or outlets. One explanation for price

dispersion is that stores are different. Stores can be heterogeneous in a

multitude of ways, such as location, opening hours, parking facilities (e.g.,

Dixit and Stiglitz, 1977; Weitzman, 1994), loyalty programs (Basso et al.,

2009), and warranties (Grossman, 1981). In addition, idiosyncratic shocks

*We are grateful for comments from Birthe Larsen, Tore Nilssen, Magnus S¨oderberg, the

referees, and seminar participants at the 2014 Annual Meeting of the Norwegian Association of

Economists, the 8th Nordic Summer Symposium in Macroeconomics, the 9th Nordic Workshop

in Industrial Organization at the University of Oslo, the 29th EEA-ESEM meeting at Toulouse

Schoolof Economics, Norwegian School of Economics, BI Norwegian Business School, Statistics

Norway, and Oslo Business School. Ø. Aas gratefully acknowledges the financial support of

the European Research Council under the European Union’s Seventh Framework Programme

(FP7/2007–2013/ERC grant agreement no. 339950).

C

The editors of The Scandinavian Journal of Economics 2019.

1182 Price dispersion and the role of stores

or unexpected fluctuations in demand (Mackowiak and Wiederholt, 2009)

can also yield price dispersion. Furthermore, store characteristics are often

an intrinsic component of a purchase (product differentiation). For example,

buying a lukewarm Coca-Cola in a supermarket in the middle of the day is

different from buying a cold one from a convenience store or a petrol

station in the middle of the night. Also, eating the same meal at two

different restaurants might be perceived as very different depending on

the characteristics of each restaurant. Thus, store characteristics can reflect

different mark-ups and costs, and result in price dispersion.

In this paper, we identify the contribution of store characteristics to price

dispersion by exploring monthly price observations from a wider set of

products categories than in previous studies. First, we establish six stylized

facts of price dispersion, as follows.

1. There is significant and persistent price dispersion in retail prices in

Norway, and the median standard deviation is 33 percent of the mean

price.

2. The dispersion of prices varies between products and over time as

indicated by the range between the first and third quartiles of the

standard deviation between 19 and 50 percent. There is less price

dispersion for non-durable and durable products than for semi-durable

products and services.

3. There is little variation in price dispersion between regions.

4. 84 percent of the overall variation in the standard deviation is between

products while 16 percent is due to time variation.

5. The dispersion in prices increased from around 25 percent at the star t

of the sample to almost 40 percent at the end of the sample.

6. While the pooled distribution of nor malized prices is unimodal,

product-specific distribution of normalized prices is often bimodal.

Second, we identify a fixed store component of prices by observing

prices of multiple products in the same store over time. Using intuitive non-

parametric methods, we find that the store component accounts for about 30

percent of the price dispersion for the median product. To identify the store

component for the sample as a whole, we also employ a novel parametric

method, which shows that the effect of store characteristics (hereafter, the

store effect) accounts for 50–60 percent of the dispersion in prices. As

further evidence of the importance of store heterogeneity, we find that the

ranking of stores within the price distributions is highly persistent over

time.

C

The editors of The Scandinavian Journal of Economics 2019.