Vertical Integration and Firm Productivity

• Author: Zhigang Tao,Yi Lu,Hongyi Li
• Date: 01 June 2017
• Published date: 01 June 2017
• DOI: http://doi.org/10.1111/jems.12191

Vertical Integration and Firm Productivity

HONGYI LI

School of Economics, UNSW Business School

UNSW Australia Sydney, NSW 2052, Australia

hongyi@gmail.com

YILU

Department of Economics

National University of Singapore

Singapore 117570

justinly6@gmail.com

ZHIGANG TAO

Faculty of Business and Economics

University of Hong Kong

Pokfulam Road, Hong Kong

ztao@hku.hk

This paper uses three cross-industry datasets from China and other developing countries to

study the effect of vertical integration on firm productivity. Our findings suggest that vertical

integration has a negative impact on productivity, in contrast to recent studies based on U.S.

firms. We argue that in settings with poor corporate governance, vertical integration reduces

firm productivity because it enables inefficient rent-seeking by insiders.

1. Introduction

This paper examines the impact of vertical integration on firm productivity. To do so,

we analyze two different cross-industry datasets of Chinese manufacturing firms and

another dataset of firms in developing countries, drawn from various surveys conducted

between 1998 and 2006.

We find that firm productivity, as measured by labor productivity, decreases with

vertical integration in each of our datasets. Our results are in contrast to recent em-

pirical findings, largely based on U.S. data (e.g., Hortac¸su and Syverson, 2007; Forbes

and Lederman, 2010) that vertical integration improves firm efficiency. We propose a

simple explanation for this negative relationship: in developing-country settings char-

acterized by poor legal protections for firms’ investors, vertical integration may serve as

an inefficient means for parties in control to extract private benefits.

A number of problems arise when estimating the causal impact of vertical inte-

gration. One concern is the endogeneity problem, whereby omitted-variable bias arises

because firms’ vertical integration decisions are nonrandom. In particular, such non-

randomness may arise because firms condition their integration decisions on unob-

served factors. A special case of the endogeneity problem is what Gibbons (2005) aptly

We thank the coeditor and two anonymous referees for their suggestions and criticisms; Oliver Hart for his

comments on an earlier version of this paper; and Adam Solomon and Lianming Zhu for excellent research

assistance.

C2016Wiley Periodicals, Inc.

Journal of Economics & Management Strategy, Volume26, Number 2, Summer 2017, 403–428

404 Journal of Economics & Management Strategy

labels the “Coase-meets-Heckman” problem: if integration is comparatively advanta-

geous at high-difficulty transactions (which are likely to produce inefficient outcomes

relative to the first-best), and if transaction difficulty is unobserved, then simple correla-

tions will naively produce a negative relationship between vertical integration and firm

productivity.

A second concern is the mismeasurement problem: the extent of vertical integration

may be mismeasured in conventional (indirect) measures for vertical integration. For

example, the value-added ratio is used extensively in the literature as a measure of

vertical integration, but may be sensitive to the stage of the production process in which

a given firm specializes.1Mismeasurement of vertical integration may lead to estimation

biases even when the integration decisions are randomly determined (i.e., even in the

absence of endogeneity problems). For example, when both the integration decision

and the mismeasurement of vertical integration are uncorrelated with unobserved firm

characteristics, estimates of the effect of vertical integration may be biased toward zero.

For each of our three datasets, the mismeasurement problem and the endo-

geneity problem may emerge in various forms. We tailor our approach accordingly.

Section 2 analyzes our first dataset, the Survey of Chinese Enterprises (SCE). The SCE is a

cross-section of Chinese manufacturing firms based on a 2003 WorldBank survey. It con-

tains a direct measure of vertical integration: the percentage of parts that are produced

in-house. This measure allows us to avoid the mismeasurement problem, which may

arise with indirect measures of vertical integration. Further,to mitigate the endogeneity

problem, we use the degree of local purchase (a proxy for the extent of site-specificity)

as an instrument for vertical integration.

Section 3 analyzes our second dataset, the Annual Survey of Industrial Firms (ASIF).

The ASIF is a panel of Chinese industrial firms based on comprehensive annual sur-

veys between 1998 and 2005. We rely on the value-added ratio as a measure of vertical

integration. To control for the mismeasurement problem introduced by our use of the

value-added ratio, we perform our analysis with a detailed set of 4-digit industry dum-

mies, and then with firm dummies. Further, the panel analysis allows us to control

for unobserved persistent firm-level heterogeneity, and thus mitigates the endogeneity

problem.

Section 4 analyzes our third dataset, the Private Enterprise Survey of Productivity

and the Investment Climate (PESPIC). The PESPIC is a cross-sectional dataset with some

time-series elements. It consists of firms in six developing countries (Brazil, Ecuador,

Oman, the Philippines, South Africa, and Zambia), and is based on World Bank sur-

veys conducted from 2002 to 2006. In particular the PESPIC contains firm-level survey

information about changes in a direct measure of the integration of major production

activities as well as changes in firm productivity. We control for unobserved firm-level

heterogeneity by performing first-differenceestimation, thus mitigating the endogeneity

problem. The PESPIC’s direct measure of vertical integration alleviates the mismeasure-

ment problem in our analysis.

Our results do not constitute a “smoking gun” for causality.We summarize the lim-

itations of each of the three datasets in Section 4.2. That said, our overall analysis paints

a picture of a significant negative relationship between vertical integration and firm

productivity that is robust to various specifications and measures of vertical integration,

and holds in a variety of cross-industry developing-country contexts. In particular, the

1. For example, the value-added ratio may be lower for firms specializing in later stages of the production

process (Holmes, 1999).