Cross‐selling in the US home video industry

• Author: Gabriel Natividad,Luís Cabral
• Published date: 01 February 2016
• Date: 01 February 2016
• DOI: http://doi.org/10.1111/1756-2171.12117

RAND Journal of Economics

Vol.47, No. 1, Spring 2016

pp. 29–47

Cross-selling in the US home video industry

Lu´

ıs Cabral∗

and

Gabriel Natividad∗∗

We identify significant cross-selling effects in the home video industry: a 10% increase in the

demand for a studio’s old titles leads to a 4.7% increase in new title sales. We argue this is due

to supply-side effects: studios with strong titles are better able to “push” other titles through

retailers; and the latter “push” these additional supplies to consumers by means of lower prices

and/or heavier advertising. Our strategyfor identifying causality is based on “star power” effects:

increases in old movie demand caused by recent success of movies with a similar cast and/or

director.

1. Introduction

Many industries are characterized by bundling or related practices. For example, movie

studios sell movies to rental stores based on full-line forcing contracts (Ho, Ho, and Mortimer,

2012a); cable providers sell TV channels to viewers based on different packages; and publishers

sell academic journals to libraries in large bundles. In some instances, bundling takes place at

the retailer-consumer level (e.g., cable TV); in other cases, at the wholesale level (e.g., movie

rentals).

When bundling takes place at the wholesale level (asi nthe case of movie rentals), the precise

details of the contractual relationship are not alwayseasy to obtain. Moreover, anecdotal evidence

suggests that various aspects of the contractual relationship between wholesalers and retailers are

not spelled out explicitly.In this context, one may ask whether and to what extent bundling takes

place, and what its downstream effects are.

In this article, we estimate the degree of bundling between wholesalers and retailers in

the home video sales industry, where a product is given by a video title of a particular movie.

(Although the industry’svalue chain can be complex, in essence there are three levels to consider:

retailers such as Kmart purchase DVDs from distributors such as Warner Bros. and sell them to

individual consumers.) Unlike Ho, Ho, and Mortimer (2012a), our approach allows us to estimate

∗New York University; lcabral@stern.nyu.edu.

∗∗Universidad de Piura; gabriel.natividad@udep.pe.

Wethank Editor Aviv Nevo,three referees, Matt Grennan, Julie Holland Mortimer, Scott Shriver, Brian Silverman, Jidong

Zhou, and seminar participants at NYU, Boston College, Pontificia Universidad Catolica de Chile, and the Columbia

Strategy Conference (2014) for helpful suggestions.

C2016, The RAND Corporation. 29

30 / THE RAND JOURNAL OF ECONOMICS

the degree of upstream bundling without knowledge of the precise nature of the contracts between

wholesalers and retailers.

Conceptually,our strategy is based on the idea that upstream bundling gets “passed through”

to downstream sales in the form of studio-level cross-selling effects. Suppose that there is no

bundling at the downstream level. We show that, if wholesalers bundle titles when selling to

retailers, then a positive shock to the consumer demand of title xshould lead to an increase in

consumer sales of title y even if there is no relation between x and y in the eyes of the consumer

(i.e., even if the two titles are neither related in terms of consumer utility nor in the way they are

sold to the final consumer).

The idea of bundling “pass through” is fairly simple. A positive shock to the consumer

demand for xleads to an increase in retailer-derived demand for x. To the extent that the retailer’s

purchases of title xare contractually linked to purchases of title y, a consumer demand shock to

xleads in turn to an increase in the retailer’s stock of y. Although there is no positive shock to

the consumer demand for y, excess inventory leads the retailer to market ymore aggressively,

which in turn results in higher sales of y. In sum, an estimate of the degree of upstream bundling

between xand yis the degree of correlation in downstream sales between xand y.

Our conceptual framework implies an additional testable prediction: retailers have different

instruments to increase the demand for y. In the home video sales market, these include the

number and location of copies displayed, the used of “corrugated boards,” and other promotion

devices. In particular, wewould expect retailers to use price as a means to “push” excess inventory.

Therefore, an additional testable prediction of our bundling “pass through” story is that a positive

shock to the demand for xis correlated with a decrease in the price of y.

Our estimates are consistent with final consumer cross-selling effects that are statistically

significant and economically important. If xrepresents demand for library titles and ysales of

new releases, we estimate a one-standard deviation demand shock to xto increase sales of yby

two thirds of a standard deviation and to decrease the price of yby four thirds of a standard

deviation.

A simple way to estimate the retail cross-sales effect would be to regress yon x.However,

such analysis would be subject to the usual criticism that causality may go either way or may

simply be absent, the correlation resulting from an omitted variable bias (e.g., the distributor’s

sales force ability). We therefore proceed by taking an instrumental variable approach. Our

identification strategy is based on an important assumption regarding movie demand,namely, that

it depends on the movie’s credits (the movie’s “star power,” i.e., its director and top cast) but not

on corporate identity (i.e., the movie’s distributor). We believe this is a reasonable assumption:

people want to watch James Cameron or Tom Cruise movies, not movies distributed by Warner

Bros.

Given this assumption, we use the “star power” channel to create an instrument for demand

shocks. Suppose that The Vow (2012), distributed by Sony and starring Rachel McAdams, per-

forms particularly well at the box office (it did). Forreasons similar to those studied by Hendricks

and Sorensen (2009) in the context of the music industry, this leads to a “backward spillover”

effect whereby the demand for Rachel McAdams moviesincreases. Warner Bros., the distributor

of Wedding Crashers (2005)—also starring McAdams—receives a positive shock to the demand

for its movie library.We argue this shock makes a good instrument for our regression of new video

sales (our yvariable) on library sales (our xvariable) because (1) it is exogenous to Warner Bros.;

and (2) it is uncorrelated with current Warner Bros. releases not featuring McAdams (or any of

the top talent in The Vow). We refine our instrumental variable design to bolster the exclusion

restriction that demand shocks are not directly affecting the studio’s new release sales.

As an additional check to our interpretation of the causes of cross-selling effects, we run a

series of placebo matched regressions where we estimate the effect of studio idemand shocks

on studio jsales, with jbeing different from but very similar to i. To continue with the earlier

example, we observe that a shock to the demand for Wedding Crashers, a Warner Bros. movie, is

associated to higher sales by Contagion, a demand-unrelated movie bythe same studio. However,

C

The RAND Corporation 2016.