The determinants of product lines
| Date | 01 September 2018 |
| DOI | http://doi.org/10.1111/1756-2171.12244 |
| Published date | 01 September 2018 |
RAND Journal of Economics
Vol.49, No. 3, Fall 2018
pp. 541–573
The determinants of product lines
Justin P. Johnson∗
and
David P. Myatt∗∗
We study product line choice allowing for cost asymmetries but ruling out strategic motivations
such as a desire to soften competition. We identify two forces that interact with asymmetric
competition to shape equilibrium product lines. Possible outcomes range from head-to-head
competition to complete separation of product lines and include the intermediate case of partial
separation. In an international trade context, we predict which qualities a disadvantaged foreign
producer will specialize in. If cost factors drive quality-based discrimination then foreign firms
sell only lower qualities. However, if demand factors are the main drivers then foreign firms sell
only high qualities.
1. Introduction
We study the determinants of the (vertically differentiated) product lines of multiproduct
competitors who engage in quantity competition. We are motivated by two observations.
Our first observation is that multiproduct firms may differ in their production costs. For
instance, a firm that exports to a given country may face higher costs (owing to tariffs or
transportation) than a domestic firm. Our second observation is that product-line configurations
among rivalsmay differ substantially in their degree of overlap. Forinstance, some rivals choose to
compete head-to-head by offeringthe same product qualities, whereas other rivals offer completely
separate product lines.
Our goal is to determine when and why such different configurations occur in equilib-
rium. Moreover, we seek to identify precisely which firms will offer which product line. For
example, if there is product-line separation in equilibrium, then we seek to identify whether a
cost-disadvantaged firm (e.g., a firm with proportionally higher costs for the production of every
quality) will focus on high quality or on low quality.
A primary finding is that—despite there being no strategic reason (such as a desire to soften
price competition) for firms to limit their product lines—firms may, nevertheless, avoid head-
to-head competition. That is, although firms do not observe the product line choices of their
∗Cornell University; jpj25@cornell.edu.
∗∗London Business School; dmyatt@london.edu.
We thank participants at the Workshop on Multiproduct Firms in Industrial Organization and International Trade (Uni-
versity of Mannheim), the HKUST 2017 IO Workshop,and the Berkeley 2017 Summer Institute in Competitive Strategy.
In particular, we thank discussants Rapha¨
el Levy and Anthony Dukes, as well as the Editor, Ben Hermalin, and two
anonymous referees for many constructive and useful comments.
C2018, The RAND Corporation. 541
542 / THE RAND JOURNAL OF ECONOMICS
competitors prior to making their own product line and output choices, a firm nonetheless may
choose to omit from its own product line some products that its rivals sell. Asymmetric costs
are necessary for this outcome to emerge. As cost asymmetry varies, equilibrium product lines
may either be fully separated, fully overlapping, or partially overlapping. Thus, our motivating
observations stated above are closelyrelated: the degree of cost asymmetr y determines the degree
of product-line overlap.
Another finding is that the structure of a firm’s product line depends both on its costs
relative to rivals and also on which of two distinct forces for price discrimination is dominant
in the industry under consideration. Even when all firms sell all products, the identity of the
dominant force determines the market shares of different firms overdifferent ranges of quality. We
distinguish between cost-driven and elasticity-driven price discrimination (defined below). When
price discrimination is cost-driven, firms with less advantageous cost structures (by which we
mean firms with costs that are proportionally lowerfor every quality) tend to sell only lower-quality
products, whereas firms with more advantageous cost structures tend to sell only higher-quality
products. On the other hand, if discrimination is elasticity-driven, then cost-disadvantaged firms
tend to sell only higher-quality products, whereas cost-advantaged firms tend to sell all qualities.
An application is to international trade: these forces determine whether exporters (facing tariffs
that make them higher cost than domestic firms) specialize in high or instead low-quality products,
and whether domestic firms sell only high-quality goods or instead both low and high-quality
goods.
To describe our results and the underlying intuitions in more detail, we begin by more
carefully describing the differences between cost-driven and elasticity-driven discrimination.
Price discrimination being cost-driven means that a monopolist offers a full product line
because providing higher quality is significantly more costly than providing lower quality. For
example, if a business-class seat on a transcontinental flight is substantially more spacious than
an economy seat, then it is also much more costly to provide.
In contrast, price discrimination being elasticity-driven means that a monopolist offers a full
product line because higher-quality products have less-elastic demand than lower-quality prod-
ucts, although higher quality might not be significantly more expensive to supply. For example,
if the main quality gain from flying business class is a free glass of wine or tastier snack, then the
provision of higher quality is not particularly expensive. Nonetheless, because some consumers
are willing to pay a significant premium for this extra quality, a monopolist may optimally choose
to deny this perk to economy ticket-holders.
The computer industry provides other examples of these two motivations for price discrim-
ination. For instance, when considering faster-access memory devices, faster microprocessors,
larger hard drives, or larger and higher resolution flat-panel displays, there may be substantive
costs over lower-quality versions, so that the decision to offer multiple versions is cost-driven.
On the other hand, sometimes a firm offers a higher-quality version that entails no significant
costs, as when Intel or IBM sold lower-quality microprocessors and printers by first building a
high-quality version and then disabling certain features.1
Settling on the example of airlines, consider competing airlines, each of which can sell both
economy and business class tickets.2These products are substitutes: an increase in the output of
one reduces the price of the other. However, a (higher-quality) business class ticket is equivalent
to a bundle of a (lower-quality) basic economyticket together with an additional quality upgrade
from economy to business class. Usefully, the bundle’s components are neither complements
1In the early 1990s, Intel disabled the mathematics coprocessor of its premium 486DX microprocessor to create
a lower-quality 486SX version. Similarly, IBM used software that limited the printing speed of its 4019 LaserPrinter to
create an economy version. In both cases, owners of the economyversions could, at a later date, purchase an upgrade to
restore the full features of the premium version.
2An airline is able to adjust its capacity on any given route by, for example, using larger or smaller airplanes.
C
The RAND Corporation 2018.
JOHNSON AND MYATT / 543
nor substitutes.3Hence, multiproduct Cournot competition over two (or more) qualities mimics
single-product Cournot markets for economy class and upgrades, respectively. A complication is
that an airline cannot sell more upgrades than basic economy tickets. If competition in the market
for upgrades leads to a push against this constraint, then (in essence) all this airline’s economy
tickets are upgraded; this is equivalent to the elimination of economy class from a product line. It
follows that a gap in a product line, such as the omission of economyclass, arises when a supplier
has a strong incentive to expand the supply of the upgrades relative to the supply of the basic
product.
Suppose for the sake of developing intuition that there are only two airlines and that price
discrimination in the airline industry is entirely cost-driven.4Moreover, suppose that any cost
advantage of one firm over the other is more pronounced in the market for upgrades. Because
price discrimination is cost-driven, the shape (although perhaps not level) of demand for the
economy ticket is the same as for the upgrade market. Just like in single-product Cournot
markets, the firm with lower costs has more market share in the upgrade market than in the
economy market. Following the discussion just above, if this difference is sufficiently large, the
constraint that the lower-costfir m cannot sell more upgrades than basic economytickets will bind,
meaning that the lower-cost firm sells only business-class tickets. In turn, this high production
of business-class tickets may drive the higher-cost firm out of the upgrade market, so that it
specializes in economy-class tickets. This explains whycost-driven price discrimination can lead
to completely nonoverlapping product lines, even in the absence of any strategic motivation to
soften competition.
Now, suppose instead that price discrimination in the airline market is entirely elasticity-
driven: assume that an upgrade to business class has no significant cost, but also assume that the
demand curve for this upgrade is much less elastic than the demand curve for economy travel. In
these circumstances, it may well be that both firms sell both products when their cost structures
are the same. However, consider what occurs when this cost asymmetry increases. Intuitively,
for small increases in this asymmetry, there is no effect on the market shares or outputs for
business-class tickets (because the costs of upgrading are assumed insignificant in this example,
meaning such asymmetry impacts only the number of basic economy tickets sold). However, as
cost asymmetry increases, the firm with higher costs will reduce the number of basic economy
tickets it sells until it hits the upgrade constraint, meaning that it will upgrade all tickets—and
hence, only sell business-class tickets. At the same time, the low-cost firm will continue to sell
both types of tickets.
Our discussion above emphasizes that an attractive feature of our approach (involving
quantity-setting and potentially asymmetric costs) is that it can explain a complete range of
outcomes, including full product-line overlap, partial product-line overlap, and full product-line
separation.5In contrast, standard models that predict nonoverlapping product lines, such as price-
setting models in which ex ante symmetric firms commit to product lines before setting their
prices, do not easily also predict partially overlapping or fullyoverlapping products lines. Rather,
in such two-stage models, no twofirms will sell the same products—equilibrium outcomes exhibit
complete separation of product lines.
3A local change in the upgrade supply influences the price of the upgrade to business class, but this change does
not influence the marginal buyer (and hence, price) of an economyticket. Similarly, a local change to the economy supply
does not influence the marginal buyer of the upgrade.
4Airlines are an example for expositional purposes. Nevertheless, there are examplesin which airlines do specialize
in particular classes of service. Some low-cost airlines (Southwest in the United States is a prominent one) offer
only economy-class seats. There are other instances of airlines that offer only business-class service from a particular
airport (British Airways offers a business-only service from London’s City Airport to New York’s JFK) or where the
configuration of the aircraft involves relatively few economy-class seats. This final point is important, because a more
general contribution of our work is to identify the economic forces which influence “relative market presence,” as
discussed later in this Introduction and elsewhere in the article, as opposed to merely whether firms literally sell zero
units of some particular quality or not.
5If firms are symmetric, then (under familiar regularity conditions) they each offer the same product line.
C
The RAND Corporation 2018.
To continue reading
Request your trial