Trade Liberalisation, Product Complexity and Productivity Improvement: Evidence from Chinese Firms

• Date: 01 July 2013
• DOI: http://doi.org/10.1111/twec.12046
• Author: Miaojie Yu,Baozhi Qu,Guangliang Ye
• Published date: 01 July 2013

Trade Liberalisation, Product

Complexity and Productivity

Improvement: Evidence from Chinese

Firms

Miaojie Yu

1

, Guangliang Ye

2

and Baozhi Qu

3

1

China Center for Economic Research, National School of Development, Peking University, Beijing,

2

Hanqing Institute of Economics and Finance, School of Economics, Renmin University of China,

Beijing, and

3

China Merchants Group, Hong Kong

1. INTRODUCTION

THIS paper investigates the effect of trade liberalisation on Chinese firms’ productivity. In

the past three decades, China has experienced dramatic trade liberalisation as well as pro-

ductivity gains. The average unweighted tariffs decreased from around 55 per cent in the early

1980s to about 13 per cent in 2002. At the same time, China’s average annual increase in

total factor productivity (TFP) in the first two decades since economic reform in 1978 was

around 4 per cent, although this pace seems to have slowed down after that (Zheng et al.,

2009). It is interesting to see whether or not China’s trade liberalisation has boosted it s pro-

ductivity. Although economists have paid some attention to this issue, the research is far from

conclusive and deserves further exploration.

First, in much of the existing work on TFP, TFP is usually measured as the Solow residual,

defined as the difference between the observed output and its fitted value calculated via ordin-

ary least squares (OLS) regressions. However, this method suffers from a number of econo-

metric problems, including simultaneity bias and selection bias. The first bias comes from the

fact that a profit-maximising firm would respond to productivity shocks by adjusting its

output, which, in turn, requires reallocating its inputs. Since such a productivity shock is

observed by firms and not by econometricians, this creates an endogeneity issue. Moreover,

firms covered in the samples are usually those that have relatively high productivity and

We thank Robert Feenstra, Joaquim Silvestre, Yang Yao, Linda Yueh, Robert Elliott, Tomohiki Inui,

Zhigang Li, Fredik Willhelmsson, Jiegen Wei and Douglas Campbell for their helpful comments and con-

structive suggestions. We thank participants at the seminar at Gothenburg University, Sweden, Conference

on Microeconomics Drivers of Growth in China, held by the University of Oxford, the 7th Korea and World

Economy (2008), the 3rd China’s New Political Economy Symposium, the 2008 International workshop on

Chinese Productivity for their useful comments and suggestions. Grants provided by the Beijing Municipal

Government, China (Project No. XK100010501), and National Natural Science Foundation of China

(Project No. 71001030 and No. 71273270) are gracefully acknowledged. We thank the excellent research

assistance from Wei Tian, Jianfeng Shan, Zhe Yuan and Lingyun Zhang. However, all errors are our own.

The World Economy (2013)

doi: 10.1111/twec.12046

Ó2013 John Wiley & Sons Ltd

912

The World Economy

survived during the period of investigation. Those firms that have exited the market due to

low productivity were not observed and thus excluded from the samples. Ignoring the firms’

entry and exit from the market means that the samples are not randomly selected, and hence,

the estimation results may suffer from selection bias.

Second, previous studies ignored the heterogeneity of goods in their estimations. Complex

products are differentiated and have many characteristics, including size, design, material and

other specifications (Berkowitz et al., 2006). In contrast, simple goods are more homogeneous,

and they are either traded on organised exchanges or are reference-priced. When facing trade

liberalisation, firms that produce complex goods may react differently with those that produce

simple goods. However, there has been no empirical evidence on whether trade liberalisation

affects the productivity of producers of complex goods and simple goods differently.

Third, much of the literature has used output tariffs as an indicator of trade liberalisation.

Recently, Amiti and Konings (2007) took a step forward to take input tariffs into account.

However, a tariff is just one of the many instruments in trade policies, which has already been

reduced to a very low level after the Uruguay Round of the WTO in 1994. Other trade policy

instruments, including various non-tariff barriers (NTBs), also play important roles in prot ect-

ing domestic import-competing industries. Restricting the scope to tariffs only is insufficient

in understanding the impact of trade liberalisation on productivity.

Last but not least, the existing literature has faced an empirical challenge in using China’s

data. Holz (2004) emphasised the bias of using China’s aggregated data since there is a mis-

match between disaggregated and aggregated statistical data. This is consistent with Krug-

man’s (1994) complaint that it is a challenging job to explain China’s economic growth due

to its low-quality data. Young (2003) argued that China’s TFP growth rate was quite modest

and perhaps negative in the post-Mao era. However, his work is based on aggregated indus-

trial data, which would be subject to some bias as well.

In this paper, to mitigate the above-mentioned estimation issues, the effect of China’s trade

liberalisation on its productivity was estimated by precisely measuring TFP, by taking into

account the difference in complex goods and simple goods, by choosing an appropriate indi-

cator of trade liberalisation and by using the most disaggregated firm-level data.

First, to address the two empirical challenges (i.e. simultaneity bias and selection bias)

caused by OLS, we adopt the Olley–Pakes (1996) approach. This approach was also revised

by imbedding a survival probability model to control for the problem of selection bias. Sec-

ond, we estimate the effect of trade liberalisation on firm productivity for complex and simple

goods separately using a classification system in line with Rauch (1999). Third, as stated

above, trade liberalisation includes the removal of various NTBs in addition to tariff cuts.

However, data on NTBs are very difficult to obtain, especially for developing countries like

China. The import penetration ratio, which is defined as industrial imports over its outputs, is

the economic consequence of both tariffs and NTBs. Compared to tariffs, the import penetra-

tion ratio is a better instrument for measuring trade liberalisation (Levinsohn, 1993). In this

paper, the import penetration ratio is used to measure trade liberalisation. Finally, the sample

in this paper is a rich firm-level panel, covering more than 150,000 manufacturing firms per

year from 1998 to 2002. For each firm, the coverage is more than 100 financial variables

listed in the main accounting sheets of all state-owned enterprises (SOEs) and those non-SOEs

firms, whose sales are more than five million yuan RMB per year.

The estimation results suggest that trade liberalisation significantly increases productivity

for firms that produce complex goods. In contrast, we find that trade liberalisation has the

opposite effect on the productivity of producers of simple goods. These findings are robust

Ó2013 John Wiley & Sons Ltd

TRADE LIBERALISATION, COMPLEXITY AND PRODUCTIVITY 913