Vertical structure and innovation: A study of the SoC and smartphone industries

• Date: 01 September 2020
• Published date: 01 September 2020
• Author: Chenyu Yang
• DOI: http://doi.org/10.1111/1756-2171.12339

RAND Journal of Economics

Vol.51, No. 3, Fall 2020

pp. 739–785

Vertical structure and innovation: A study

of the SoC and smartphone industries

Chenyu Yang∗

This article studies how vertical integration and upstream R&D subsidy affect innovation and

welfare in vertically separated industries. I formulate a dynamic structural model of a dominant

upstream firm and oligopolistic downstream firms that invest in complementary innovations. I

estimate the model using data on the System-on-Chip (SoC) and smartphone industries. The

results suggest that a vertical merger can increase innovation and welfare, mainly driven by the

investment coordination of the merged firms. I also find that subsidizing the upstream innovation

increases overall private investment, innovation, and welfare.

1. Introduction

In vertical industries, upstream and downstream innovations are often complementary. Up-

stream firms upgrade the core technology essential to performance enhancement, and down-

stream firms combine the technology with innovative designs in new consumer products. Ex-

amples of complementary innovations include traction batteries (upstream) and electric vehicles

(downstream), CPUs (upstream) and personal computers (downstream), and System-on-Chips

(SoCs, upstream) and smartphones (downstream). The upstream innovations in these industries

bear many similarities to the concept of “General Purpose Technologies” (GPT, Bresnahan and

Trajtenberg, 1995): different downstream firms are consumers of the same upstream technology

(pervasiveness), the upstream firms improvethe technology over time (continuous improvement),

and upstream innovations enabledownstream innovations (innovation complementarity). Bresna-

han and Trajtenberg (1995) points out an important under-innovation problem in a decentralized

economy: firms fail to internalize the innovation complementarity because of the arms-length

transactions between the GPT innovator and its users, and they under-innovate relative to the

level of joint profit maximization. This article presents an empirical model of innovation in verti-

cally separated industries and uses counterfactual simulations to quantify how vertical integration

and upstream R&D subsidy may affect innovation and welfare.

∗Department of Economics, University of Maryland, College Park; cyang111@umd.edu.

I would liketo acknowledge the generous financial support provided by Rackham Graduate School and Michigan Institute

for Teaching and Research in Economics (MITRE) at the University of Michigan. I am grateful for the guidance and

support from my advisers Jeremy Fox,Daniel Ackerberg, Ying Fan and SrinivasaraghavanSriram. I would like to thank

Danial Asmat, Will Strauss and Will Peichun Wang for insight into the mobile phone industr y. I benefited greatly from

the comments of many seminar participants, two anonymous referees and the editor Marc Rysman. All errors are mine.

© 2020, The RAND Corporation. 739

740 / THE RAND JOURNAL OF ECONOMICS

The empirical context of this article is the innovation in the SoC and smartphone industries.

Since the introduction of the original iPhone, the smartphone industry has grown explosively.

The global smartphone sales were 120 billion dollars in the last quarter of 2017 (Koetsier, 2018).

The SoC is a key smartphone component that combines a mobile application processor (es-

sentially a CPU), GPU, modem, and other chips (Yang et al., 2014). The SoC and smartphone

industries provide an interesting setting to study innovationin vertically separated industries for a

number of reasons. First, the SoC and smartphone innovations are strongly complementary. The

enhanced processing power, energy efficiency, and other functionalities from SoC innovations

not only directly improve smartphone qualities, but also enable handset makers to adopt new

designs. Secondly, vertical integration is a historically controversial subject in the mobile phone

industry. Qualcomm, the dominant upstream firm in the SoC industry, was ver tically integrated

until 1999, and foreclosure was a main concern for many of Qualcomm’s handset maker cus-

tomers when Qualcomm was vertically integrated (Dingee and Nenni, 2015). Finally, the recent

US–China trade disputes have renewedattention to government subsidy programs for R&D in the

semiconductor sector (Jamrisko and Torres, 2018). Even before the recent events, the US gov-

ernment had been formulating plans to help upstream innovators, such as Intel and Qualcomm,

with the belief that upstream innovations stimulate innovations across the economy (Holdren and

Otellini, 2016). I use counterfactual simulations to explore the policy implication of an upstream

subsidy to Qualcomm for the innovation in the SoC and smartphone industries.

To model the innovation and pricing of the SoCs and handsets, I consider a dynamic game

of investment that nests a bargaining and pricing stage. The upstream industry consists of a domi-

nant firm (“Qualcomm”) and a non-strategic fringe, and the downstream firms are a finite number

of oligopolistically competitive handset makers. In every period, Qualcomm and its downstream

clients first negotiate SoC prices via Nash bargaining, and the handset makers then set wholesale

prices in the Nash–Bertrand equilibrium. The subgame perfect equilibrium of the overall static

Nash-in-Nash (Collard-Wexler, Gowrisankaran, and Lee, 2014) pricing game determines the pe-

riod profits. Modeling the bilateral negotiation between Qualcomm and handset makers allows

me to quantify how a change in the market structure affects pricing in the counterfactual. I embed

the implied profit functions in the dynamic game of upstream and downstream innovations. In

the dynamic game, the upstream and downstream innovations are complementary: Qualcomm

invests to increase the quality of its SoCs; downstream handset makers invest to increase the

quality of their handsets, but the technological frontiers of some handset makers depend on that

of Qualcomm. When deciding whether to innovate, upstream, and downstream, firms weigh the

gains in the present discounted values of future profits due to the innovation against the sunk

cost, and the dynamic innovation decisions form a Perfect Bayesian Equilibrium (PBE).

I estimate the model using data from the US smartphone market from 2009 to 2013. The

estimation procedure has three steps. First, price and quantity data of handsets allow me to esti-

mate a random coefficient logit model of consumer demand for smartphones. I refer to a linear

combination of a product’s characteristics, where the weights are given by the estimated demand

coefficients, as the quality index of the product, and I use these indices to construct the quality

frontiers of Qualcomm and handset makers. Next, I recover SoC prices and other marginal costs

of smartphones using equilibrium pricing conditions and data on the markup of SoCs. The first

two steps do not involve estimating the dynamic model. The estimates and the pricing equilib-

rium assumptions imply the period profit functions of the upstream and downstream firms. In

the last step, I use the estimated period profit functions and the evolution of quality frontiers of

Qualcomm and handset makers to estimate the innovation cost functions. To keep the compu-

tation tractable, I estimate a dynamic game among the upstream Qualcomm and three handset

makers: Apple, HTC, and Samsung. Consistent with data, I assume that Apple uses its own

SoCs, HTC exclusively uses the SoCs from Qualcomm, and Samsung sources both Qualcomm

and non-Qualcomm SoCs. The quality of Qualcomm forms the upper bound on the technological

frontiers of Samsung and HTC. I use a Simulated Minimum Distance estimator (Shi and Shum,

2015) to estimate the model.

C

The RAND Corporation 2020.

YANG / 741

Using the estimated model, I simulate two approaches to address the potential under-

innovation problem. The first counterfactual considers a vertical merger betweenQualcomm and

HTC. This merger did not occur, but the simulation allows me to explore the magnitudes of var-

ious economic forces in vertical integration. The merged Qualcomm and HTC fully internalize

their complementary innovations and jointly make innovation decisions.1The vertical integration

also changes the pricing incentives. By endogenizing both the pricing and innovation decisions,

I allow for and contrast the benefit of investment coordination between the merged fir ms and

the potential harm from “raising rival’s costs”: the integrated upstream Qualcomm has an incen-

tive to raise the SoC prices to a competing handset maker that buys SoCs from Qualcomm.2In

the main specification, I find that the upstream Qualcomm’s innovation rate, defined as the av-

erage increase of quality per period, increases by 13% to 35% (95% confidence set), and the

innovation rate of the integrated HTC increases by 14% to 20%. Moreover, Samsung’s innova-

tion rate increases by 9% to 22%. Apple’s innovation rate increases by less than 3%. Consumer

surplus increases by 4% to 8%. I also find that investment coordination accounts for most of the

gains from vertical integration. In terms of policy implications, the findings suggest that antitrust

regulators should fully take into account the potential positive effect of coordinated investment

in vertical integration, especially for innovative industries. My focus in this exercise is to quan-

titatively assess how vertical integration changes innovation by aligning incentives. Potentially

vertical integration could allow the merged firms to specialize and realize additional gains by

reducing marginal costs or lowering innovation costs.

The second counterfactual examines the effect of an upstream R&D subsidy. The existing

empirical evidence is rather mixed on whether public subsidies crowd out private investment

(David, Hall, and Toole, 2000; González, Jaumandreu, and Pazó, 2005). Furthermore, whether

a subsidy is welfare-enhancing (welfare increase greater than the amount of the subsidy) partly

depends on the changes in consumer welfare. I examine whether the effect of a subsidy is more

definitive when the upstream subsidized firm is similar to a GPT provider, whose faster innova-

tion can stimulate the innovations of multiple downstream firms. The results suggest upstream

subsidies increase private investment and welf are. However, the effects on downstream firms are

heterogeneous. A 10% subsidy of Qualcomm’s R&D expenditures increases Qualcomm’s inno-

vation by 15% and Samsung’s innovation by 11%, but has a much smaller positive effect (1.3%)

on HTC. Apple’s innovation slightly declines (-2%). Overall, the welfare effect of an upstream

subsidy is large and positive, and the increase in the total surplus exceeds the amount of the

subsidy.

Related Literature and Contribution. The key modeling novelty in this article is the specifica-

tion of dynamic upstream and downstream firms in vertical industries.3The model of innovation

builds on the alternating-move, finite-horizon dynamic game in Igami (2017) by including an

upstream innovator and considering complementary upstream and downstream innovations. The

model of pricing nested in the dynamic game is related to the empirical bilateral bargaining

framework (Crawford and Yurukoglu, 2012).4This type of models has been widely used to ana-

lyze the pricing of services and physical goods in vertical industries (within a static game).5,6

1This treatment follows the view that vertical integration facilitates the transfer of knowledge input between the

merged firms (Atalay,Hortaçsu, and Syverson, 2014; Natividad, 2014).

2An additional benefit of vertical integration is the elimination of double marginalization.

3This article is related to the vast literature on vertical structures. Lafontaine and Slade (2007) surveys the theo-

retical and (reduced-form) empirical literature.

4This empirical framework is based on Horn and Wolinsky (1988).

5Examples include Draganska, Klapper,and Villas-Boas (2010); Grennan (2013); Gowrisankaran, Nevo, and Town

(2014); Crawford et al. (2015); and Ho and Lee (2016).

6Like many papers in this literature (Brenkers and Verboven, 2006; Murr y, 2015; Asker, 2015; Crawford et al.,

2015, in addition to those cited above), I assume that firms in my model use linear price contracts. I later discuss the

pros and cons of this modeling choice in more details. Another strand of the empirical structural literature on vertical

C

The RAND Corporation 2020.