Persistent Price Dispersion in Online Markets

TLDR: In this article, the authors provide a non-technical and compelling analysis of the modern macro-economy, and provide their views on the New Economy from a variety of perspectives.

Abstract: What is the New Economy, what makes it new, and what are the implications for antitrust, regulation and macroeconomic policy? Providing a non-technical and compelling analysis of the modern macro-economy, the contributors to this volume, eminent scholars all, provide their views on the New Economy from a variety of perspectives.

Figures

Table 4: Product Life Cycle, Range of Prices, and Number of Firms

Table 5: Percentage of Price Dispersion Explained by Observable and Unobservable Firm Heterogeneities

Figure 4: Measures of Price Dispersion Controlling for Inventory Differences

Table 3: Product Life Cycle and Range of Prices

Figure 3: Measures of Price Dispersion Controlling for Differences in Shipping

Table 1: Sample Prices Posted By Four Firms on November 5, 1999

Full text

Persistent Price Dispersion in Online Markets
by
Michael R. Baye, John Morgan, and Patrick Scholten
April 4, 2002

Authors’ Blurb
Michael R. Baye is a professor in the Department of Business Economics and Public
Policy, Kelley School of Business, Indiana University, Bloomington, IN 47405
John Morgan is the corresponding author. He is a professor in the Woodrow Wilson
School and Department of Economics, Princeton University, Princeton NJ, 08544. e-
mail: rjmorgan@princeton.edu
Patrick Scholten is a graduate student in the Department of Economics, Indiana
University, Bloomington, IN 47405

Abstract
Using data from one of the Internet’s leading price comparison sites for consumer
electronics products, we present evidence for the persistence of price dispersion for
36 homogeneous products. The markets for these products are “thick” with an
average of over 20 firms selling each product. We show that prices do not converge
to the “law of one price” even after an 18 month period. This finding is robust to
controls for differences in shipping charges and inventories. Further, we show that
product life cycle effects lead to changes in the number of competing firms and the
range of price dispersion consistent with the theoretical predictions of the Varian
(1980) model. The average number of competing firms declines from about 28 to 10
during the final five months of our dataset. Over this same period, the average range
in prices decreases from about 75 percent to 30 percent. After accounting for firm
heterogeneities in costs, branding, reputation, trust, product availability and shipping
costs, 28 percent of the variation in prices charged for homogeneous products remains
unexplained. This is also consistent with the Varian model.
Keywords: Price dispersion, Internet, Law of One Price
JEL Numbers: D4, D8, M3, L13

1
1. Introduction
The failure of the “law of one price” has been widely observed in non-Internet
markets. In his seminal paper, Hal Varian quipped that when it came to prices for
identical products in conventional retail markets, “… the ‘law of one price’ is no law
at all.” (Varian 1980, 651) Beginning with Stigler (1961), a large and successful
theoretical literature has arisen explaining how dispersed prices can comprise
equilibrium in the presence of price-sensitive consumers and homogeneous products.
One approach generates price dispersion in pure strategies, where different firms
charge different prices because of heterogeneities in costs or service levels. The other
approach generates price dispersion through randomized pricing strategies by firms.
1
As Varian points out, the former type of price dispersion is unlikely to be long-lived,
as consumers will eventually learn which firms are charging the best prices and shop
from them exclusively. Varian argues that the latter type of price dispersion will
persist over time.
Price dispersion has also been observed in Internet markets (see Smith, Bailey,
and Brynjolfsson, 1999 for a useful survey of results
2
); however, little is known about
the empirical persistence of this price dispersion over long periods of time. In contrast
to Varian’s explanation, Brynjolfsson and Smith (2000) argue that price dispersion is
largely due to retailer heterogeneity with respect to branding, awareness, and trust. If
this is the case, price dispersion should be modest after controlling for such
heterogeneities.

2
The present paper examines the persistence of price dispersion in a well-
established online retail market. Using data collected over an 18-month period from
one of the leading price comparison sites on the Internet, we show that prices for
identical consumer electronics products listed by multiple retailers display
considerable and persistent price dispersion. On average, the highest price for a
consumer electronics product is 57% above the lowest available price. The savings to
a consumer simply from consulting the comparison site are also significant—on
average, the difference in prices paid by consumers shopping at a randomly selected
firm rather than from the firm offering the lowest price is about $31.
Price dispersion persists across products and across time. We find no convergence
to the “law of one price” over an 18 month period even after controlling for
differences in shipping charges and inventories. We also use time variation in the
data to make an additional inference. Specifically, we show that Varian’s model
implies that as the reservation value of consumers seeking to purchase an item
decreases, (1) the number of firms listing the product declines and (2) the range of
prices offered decreases. The methodology we employ to test these predictions is to
use the short life cycles of consumer electronics products as a means of obtaining
variation in consumer reservation values. Reservation values are presumably lower
later in the product lifespan. Empirically, we find evidence consistent with these
hypotheses. The number of competing firms declines by over 60 percent in the final
seven months of our dataset. Over this same period, the average range in prices
decreases by $31 or 54% of the range at the start of the sample period.

Frequently asked questions

Q1. What are the contributions in "Persistent price dispersion in online markets" ?

Using data from one of the Internet ’ s leading price comparison sites for consumer electronics products, the authors present evidence for the persistence of price dispersion for 36 homogeneous products. The authors show that prices do not converge to the “ law of one price ” even after an 18 month period. Further, the authors show that product life cycle effects lead to changes in the number of competing firms and the range of price dispersion consistent with the theoretical predictions of the Varian ( 1980 ) model. 

Q2. What is the methodology the authors use to test these predictions?

The methodology the authors employ to test these predictions is to use the short life cycles of consumer electronics products as a means of obtaining variation in consumer reservation values. 

Q3. How much can a consumer save by buying from the lowest rather than the highest priced of these?

A consumer shopping for a digital camera, the Nikon Coolpix 950 can save about $100 by simply buying from the lowest rather than the highest priced of these four merchants. 

Q4. What is the main finding of this paper?

Their central finding is that price dispersion is remarkably persistent over the 18 month period of their study—even after controlling for shipping costs and firm heterogeneities. 

Q5. What are the key factors in generating price dispersion on the Internet?

Brynjolfsson and Smith (2000) identified branding, reputation, and productawareness as key factors in generating price dispersion on the Internet. 

Q6. How many percent of the variation in prices for homogeneous products remains unobservable?

21To summarize, after accounting for firm heterogeneities in costs, branding,reputation, trust, product availability and shipping costs, 28 percent of the variation in prices charged for homogeneous products remains. 

Q7. How does the model explain the dispersion of prices?

regressions controlling for the presence of banner ads, consumer feedback ratings, firm disclosures about shipping costs, and inventory availability as proxies for these factors explain only about 17% of the observed dispersion in prices. 

Q8. How much is the coefficient associated with the range of prices?

The coefficient associated with D1 is -4.576 indicating that the range in prices is compressed by about $4.58 within ten months of the end of the dataset. 

Q9. How many percent of the variation in prices for homogeneous products is observable?

If price dispersion were an artifact purely of these observable differences, the R-squared values would all have been close to 1. 

Q10. What is the reason why price dispersion is so persistent?

In contrast to Varian’s explanation, Brynjolfsson and Smith (2000) argue that price dispersion is largely due to retailer heterogeneity with respect to branding, awareness, and trust. 

Q11. How many firms in the sample are dummies?

even with individual dummies for every firm in the sample and controls for each product in their sample, 28 percent of the dispersion in prices remains unexplained.