This article was downloaded by: [Nipissing University]

On: 03 October 2014, At: 23:18

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Review of Applied

Economics
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/cira20

Knowledge spillovers, absorptive

capacity and total factor productivity
in China’s manufacturing firms
a a b
Hailin Liao , Xiaohui Liu & Chengang Wang
a
School of Business and Economics , Loughborough University , UK
b
School of Management , University of Bradford , UK
Published online: 21 Nov 2011.

To cite this article: Hailin Liao , Xiaohui Liu & Chengang Wang (2012) Knowledge spillovers,
absorptive capacity and total factor productivity in China’s manufacturing firms, International
Review of Applied Economics, 26:4, 533-547, DOI: 10.1080/02692171.2011.619970

To link to this article: http://dx.doi.org/10.1080/02692171.2011.619970

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 International Review of Applied Economics
Vol. 26, No. 4, July 2012, 533–547

Knowledge spillovers, absorptive capacity and total factor

productivity in China’s manufacturing firms
Hailin Liaoa,*, Xiaohui Liua and Chengang Wangb
a
School of Business and Economics, Loughborough University, UK; bSchool of
Management, University of Bradford, UK
(Received 1 September 2010; final version received 31 August 2011)
Downloaded by [Nipissing University] at 23:18 03 October 2014

Applying the stochastic frontier framework, this study explores the diffusion
and absorption of technological knowledge in China’s manufacturing firms,
based on a panel of more than 10,000 local and foreign-invested firms over the
period 1998–2001. Our empirical approach allows us to distinguish between
technological progress (TP) and technical efficiency (TE) in analysing whether
R&D, exports and the presence of foreign direct investment simultaneously
enhance TP through knowledge spillovers in a single framework and whether
different types of domestic absorptive capacity moderate external knowledge
spillovers in relation to TE. The results show that there are positive inter-
industry productivity spillovers from R&D and foreign presence, whereas
evidence of intra-industry productivity spillovers from FDI to Chinese firms is
less robust. We find evidence that absorptive capacity is one of the key determi-
nants to quantitatively explain intra-industry differences in productivity of local
Chinese firms. The findings have important policy implications.
Keywords: knowledge spillovers; R&D; stochastic frontier; absorptive capacity;
China
JEL Classifications: F14, F23, L60

Introduction
It is widely recognised that knowledge originating in one country increasingly tran-
scends national boundaries and contributes to technological development in other
countries. Domestic technological development is no longer purely based on domes-
tic resources. Intensive empirical studies have considered international trade, foreign
direct investment (FDI) and external R&D as the major channels for knowledge
spillovers (Blalock and Simon 2009; Griffith, Redding, and Van Reenen 2003;
2004; Liu and Buck 2007; Miller and Upadhyay 2000; Wei and Liu 2006). How-
ever, the empirical confirmation of external knowledge spillovers has proved elu-
sive. One possible explanation regarding the mixed findings in the empirical
literature may be that not only positive knowledge spillovers are likely to arise from
any of the above-mentioned channels; there could also be negative ones. Purely
focusing on one rather than all channels simultaneously, therefore, might yield
biased results as the overall result could show the combined/offset effects with

*Corresponding author. Email: [email protected]

ISSN 0269-2171 print/ISSN 1465-3486 online

Ó 2012 Taylor & Francis
http://dx.doi.org/10.1080/02692171.2011.619970
http://www.tandfonline.com
 534 H. Liao et al.

either negative or positive spillovers arising from each individual channel sepa-
rately. Hence, these channels should be assessed in a unified single framework.
Despite the fact that absorptive capacity has been addressed in previous studies,
research in this area tends to focus almost exclusively on human capital at the
aggregate level of productivity (Griffith, Redding, and Van Reenen 2003; 2004;
Miller and Upadhyay 2000). What is absent in this area is an examination of spe-
cific absorptive capacity with regard to different channels for knowledge spillovers.
R&D, FDI and exports as channels for knowledge spillovers may require different
absorptive capacity. Hence, this omission represents an important research gap.
Adopting the stochastic frontier framework, we aim to examine how different
components of total factor productivity (TFP) are affected by different channels for
external technological knowledge spillovers, and how different types of absorptive
capacity moderate the impact of external knowledge spillovers on TFP in a unified
single framework. The paper differs from previous studies in a number of ways.
Downloaded by [Nipissing University] at 23:18 03 October 2014

First, extending previous work on knowledge spillovers (Blalock and Simon 2009;
Buckley, Clegg, and Wang 2007; Wei and Liu 2006), we consider different chan-
nels for knowledge spillovers in a single framework, which helps to capture various
spillover effects simultaneously, providing a better understanding of how local firms
acquire external technology through an integrated learning process.
Second, the paper considers different dimensions of absorptive capacity with
regard to different channels for knowledge spillovers and investigates how these
dimensions impact on the TFP of domestic firms. This distinction makes our
research on absorptive capacity unique and represents the first study to systemati-
cally examine different dimensions of absorptive capacity in relation to TEC and
therefore TFP. Specifically, our understanding is that productivity growth has been
related to the increased openness of an economy with associated knowledge spill-
overs and production units’ own characteristics in order to fully reap the benefits
from spillovers. With the exception of Girma and Gong (2008a) and Kneller
(2005), there is an implicit assumption that production units are producing along
the production possibility frontier with full technical efficiency. By adopting the sto-
chastic frontier framework developed by Aigner, Lovell, and Schmidt (1977) and
Meeusen and Van den Broeck (1977), TFP growth can be decomposed into techno-
logical progress (TP) and technical efficiency change (TEC) components. Hence,
our empirical analysis addresses foreign technology spillovers as one of the sources
of TP in an endogenous framework in addition to the changes in (in)efficiency – i.
e. distance from the frontier – which reflects firms’ capacity to learn from techno-
logical leaders. We also further differentiate intra- and inter-industry spillovers in
the presence of absorptive capacity and add new evidence on how the impact of
absorptive capacity differs with industries.
Finally, this paper investigates productivity spillovers by applying a one-step
estimation to Chinese manufacturing sectors, which represents an alternative
approach to most previous studies. The majority of the existing empirical studies
have applied a two-step approach to the examination of the impact of knowledge
spillovers on productivity, i.e. to estimate TFP first, and then regress it on various
factors considered to be linked with the changes in TFP (e.g. Madden, Savage, and
Bloxham 2002; Okabe 2002; Savvides and Zchariadis 2004; Qi et al. 2009). There
is, however, a logical inconsistency in this two-step approach. Productivity is often
derived from the well-known growth-accounting exercise that is exogenous by nat-
ure in the first stage, while in the second stage it is assumed to be a function of a
 International Review of Applied Economics 535

number of endogenous variables. Koop, Osiewalski, and Steel (1999) argued that
this approach is unsatisfactory as it would lead to substantial (generated) regressor
problems due to the use of estimated means as input data in the second stage, and
ignorance of large uncertainty in the point estimates.
This study considers the above issues in the case of the Chinese manufacturing
sector. Chinese manufacturing firms provide an excellent setting to study the effect
of R&D, FDI and exports on the structure of productivity and growth not only
because they have received the largest FDI inflow among developing countries
since the mid-1980s, but also because the manufacturing sector still has a large
stake in the Chinese economy – it employs over 30% of the non-agricultural work-
force, and its linkages with the other sectors of the economy set the tone for the
overall business cycle and affect the stability of other sectors. Hence, the findings
from the study have important policy implications.
Downloaded by [Nipissing University] at 23:18 03 October 2014

Literature review
In the increasingly globalised world economy, individual countries’ productivity
depends both on their own and on foreign R&D efforts. Some studies have postu-
lated the impact of foreign R&D on domestic TP. Based on a theoretical model of
innovation-driven growth, Coe and Helpman’s (1995) work provided empirical evi-
dence of the positive effect of trade-related foreign R&D spillovers on TFP in a
sample of 22 industrial economies over the period 1971–1990. Using a similar
approach, Coe, Helpman, and Hoffmaister (1997) found strong and robust evidence
for spillovers from the North (industrialized countries) to the South (77 developing
countries). In addition, some empirical studies have shown that imports are a con-
duit for knowledge spillovers (Bitzer and Geishecker 2006; Lumenga-Neso, Olarre-
aga, and Schiff 2005).
The above studies mainly consider imports as the sole channel for knowledge
spillovers across countries and may have neglected the relative magnitude of inter-
national spillovers through other channels, such as exports and FDI. Keller (1998)
challenges Coe and Helpman’s (1995) results by generating simulated and randomly
selected trade partners and estimating the international R&D spillovers. He finds
that these ‘counter-factual estimations’ give rise to larger positive international
R&D spillovers and explain more of the variation in productivity across countries.
Recent studies at firm level have considered exports as a channel for technology
spillovers. Exporting may lead to technology spillovers in two ways. First, linkages
with buyers and suppliers are found to be an important source of learning for
exporters. Exporting firms may obtain technical assistance from foreign buyers or
buyers may specify high quality products. Evidence indeed shows that technology
transfer from buyers does take place (Blalock and Gertler 2004; Salmon and Shaver
2005; Westphal 2002). Second, exporting firms are forced to be productive to
remain viable in competitive international markets, as these firms are exposed to the
intense competition of exporting markets. While non-exporting firms may be insu-
lated from such competition by trade and geographical barriers, exporting firms
may find it difficult to survive without adopting best-practice technology and
increasing their efficiency (Blalock and Gertler 2004).
In addition, numerous studies have focused on the impact of both inward and out-
ward FDI on the productivity growth of a host country. Girma and Gong (2008b) find
that the productivity of domestic firms can be improved to a significant degree by the
 536 H. Liao et al.

competition and demonstration effects from inward FDI. Lichtenberg and van Pottels-
berghe de la Potterie (2001) extend Coe and Helpman’s (1995) analysis by incorporat-
ing both inward and outward FDI flows to the trade flow. They find that the elasticities
of foreign R&D capital stocks with respect to outward FDI flows and imports are both
significant. Other studies using firm-level data yield conflicting results, depending on
the country considered (Aitken and Harrison 1999; Girma and Wakelin 2007).
A major weakness in the previous studies is that the absorptive capacity of host
countries is treated with great simplicity. Absorptive capacity, represented by invest-
ments in R&D and/or human capital, can provide the basis of fundamental knowl-
edge or technology necessary to assimilate and exploit external knowledge. When
the rate of knowledge spillovers is treated as exogenous, however, one implicitly
assumes that knowledge flows to the firm as some kind of ‘manna from heaven’,
and firms do not have to engage in any form of costly learning or a search process
to benefit from positive externalities. However, knowledge spillover effects are
Downloaded by [Nipissing University] at 23:18 03 October 2014

limited by domestic absorptive capacity (Castellani and Zanfei 2003; Cohen and
Levinthal 1989). Although some studies consider the impact of absorptive capacity
on knowledge spillovers (Borensztein, Gregorio, and Lee 1998; Grunfeld 2006; Gir-
ma and Gong 2008a), they did not differentiate absorptive capacity with regard to
technological progress and technical efficiency, and how different aspects of absorp-
tive capacity affect the significance and magnitude of the impact of various chan-
nels on knowledge spillovers.
To fill this gap, we distinguish between TP and TE and argue that TP repre-
sented by a shift in the production frontier is usually stimulated by knowledge accu-
mulation and innovation, and thus can be viewed as an ‘innovative’ effect of
productivity growth. On the other hand, efficiency gain, which helps reduce the gap
between the actual level of production and a steady state production frontier can be
viewed as a ‘catch up’ effect of productivity growth. These two distinct factors are
the important sources of TFP growth. Thus, detailed analysis about how spillovers
and different types of absorptive capacity related to TP and TE affect these two
sources of productivity growth can help enhance our understanding of the real con-
tribution of each factor to the productivity growth of the recipient economy. By
doing so, this study applies the stochastic frontier model to examine the impact of
FDI, R&D and exports on the sources of productivity and growth in the Chinese
manufacturing sector. In particular, the adopted empirical technique allows us to
distinguish TP (innovation) and TE (catch-up), thus further assessing the influence
of technology spillovers on these two distinct components of TFP growth.

The model
Our estimations are confined to the impact of knowledge spillovers on the produc-
tivity of local Chinese firms only, as there are considerable differences in technolog-
ical and managerial capability between domestically and foreign-owned firms
(Wei and Liu 2006). These two types of firms have been found to differ in average
performance in terms of productivity, capital intensity, R&D intensity and export
orientation (Liu and Buck 2007; Liu, Wang, and Wei 2009).
We start our knowledge spillover analysis with the simplest production function
as depicted in equation (1), in which output in the firm i at time t is a function of
the production technology. A measures TFP in firm i and K and L are the usual
input factors, measuring the factor inputs, physical capital and labour, respectively.
 International Review of Applied Economics 537

Yit ¼ Ait f ðKit ; Lit Þ ð1Þ

Previous empirical studies, e.g. Coe and Helpman (1995), Eaton and Kortum
(1999), Keller (2002) and others, suggest that knowledge can be assumed to be a
function of domestically generated knowledge, Dit, and knowledge generated by
firms that lie on the technical frontier F. Then
Ait ¼ gðDit ; FÞ ð2Þ

Substituting equation (2) into equation (1), we have

Yit ¼ gðDit ; FÞ f ðKit ; Lit Þ ð3Þ

The role of investment in research and development has long been recognised in
Downloaded by [Nipissing University] at 23:18 03 October 2014

the literature of innovation and national systems of innovation (Schumpeter 1934;

Freeman 1987). The recent endogenous growth theory (Romer 1990; Aghion and
Howitt 1992) has drawn upon insights from Schumpeter and the literature of
national systems of innovation and has formally shown that advances in technical
knowledge are generated by investing resources in R&D. Hence, the level of
domestic knowledge is assumed to be a function of the sum of previous investment
in R&D – the stock of R&D – where b measures the return to R&D.
Dit ¼ RDbit ; 0\b\1 ð4Þ

It follows that if domestically generated knowledge is a function of the stock of

R&D then frontier technology is, in turn, a function of the R&D stock in the for-
eign country, which can be spread via various channels, including R&D, interna-
tional trade and the presence of FDI. Cross-firm knowledge spillovers occur
because firms that lie behind the technical frontier attempt to imitate the technolo-
gies adopted by firms in countries/industries that lie on the technical frontier, as the
marginal costs of knowledge incurred are almost negligible, and therefore they ben-
efit from these positive externalities.
F ¼ hðRDSP; EXSP; FDISPÞ ð5Þ

So far we have explicitly assumed that knowledge spillovers are absorbed fully by
recipient firms, and there is no inefficiency in knowledge flows. However, if a pro-
duction unit operates beneath the production frontier, then its distance from the
maximal measures its technical inefficiency (Kumbhakar and Lovell 2000). Put dif-
ferently, the frontier approach is capable of capturing both efficiency change and
technological change as components of productivity change, which introduces an
additional dimension to analysis from the policy perspective (Nishimizu and Page
1982). Specifically, technological progress represented by a shift in the production
frontier is usually stimulated by knowledge accumulation and innovation, and thus
can be viewed as an ‘innovative’ effect of productivity growth. On the other hand,
the actual domestic productivity level is assumed to be lower than the frontier one,
ceteris paribus, because domestic firms are inefficient in their use of frontier tech-
nology, which in turn depends on their absorptive capacity. Therefore, efficiency
gain which helps reduce the gap between the actual level of production and a
steady state production frontier can be viewed as a ‘catch up’ effect of productivity
growth. Actual output (Y) can be given by
 538 H. Liao et al.

Yit ¼ gðDit ; FÞ f ðKit ; Lit Þgit ð6Þ

where g (0< g<=1) represents technical efficiency, reflecting the difference in the
outcome from the application of technical knowledge. If g=1, it implies that the
domestic firm uses all of the inputs in the production process efficiently, otherwise
impediments to absorption will cause the firm to produce within the production
frontier.
In our empirical analysis, we opt for a parametric approach by considering the
time-varying stochastic production frontier, originally proposed by Aigner, Lovell,
and Schmidt (1977) and Meeusen and van den Broeck (1977), in semi-translog
form, as
yit ¼ b0 þ b1 kit þ b2 lit þ b3 kit2 þ b4 lit2 þ b5 lit kit þ b6 RDINTit þ b7 RDSPit
þ b8 EXSPit þ b9 FDISPit þ ðvit  uit Þ ð7Þ
Downloaded by [Nipissing University] at 23:18 03 October 2014

where lower-case letters represent logarithms. The distribution of technical ineffi-

ciency effects, uit, is taken to be the non-negative truncation of normal distribution
Nðl; r2u Þ modelled, following Griffith, Redding, and Van Reenen (2003, 2004), as
dependent on the firm’s R&D-based absorptive capacity of foreign knowledge.
Griffith, Redding, and Van Reenen (2003, 2004) argued that own R&D may play
an important role in the absorption of the fruits of foreign knowledge as countries/
firms can increase their ability to assimilate and understand the discoveries of other
firms by engaging in R&D, thereby raising the speed at which technology transfer
occurs. Hence, we can define u as

uit ¼ d0 þ ðd1 RDABit  d2 EXABit  d3 FDIABit Þ ð8Þ

Data
The data used are mainly from the Annual Report of Industrial Enterprise Statistics
compiled by the State Statistical Bureau of China, covering firms in nine two-digit
industries during the period 1998–2001. The sample period was between the Asian
financial crisis in 1998 and China’s WTO entry in 2001 and represented a relatively
stable period in terms of FDI inflows, and during which many Chinese firms were
recovering from the crisis and preparing for the WTO entry. Hence, choosing such
a sample period enables us to make a useful contribution, which reflects the dyna-
mism seen in business activities in China. These industries are food processing,
food manufacturing, beverage production, garments and other fibre products, medi-
cal and pharmaceutical products, ordinary machinery manufacturing, transport
equipment manufacturing, electric machines and apparatus, and electronic and tele-
communications equipment. For each industry, the Bureau collected detailed data
on each firm in operation. The data include information on ownership classification,
value added, output, capital stock, the number of employees, total sales, intangible
assets, new product sales and exports. As for deflators, price indices for total manu-
facturing fixed assets and industrial output are obtained from the China Statistical
Yearbook 2002.
As a result of entry and exit and ownership restructuring, the number of firms
in operation changes over time. In this study, the same firms have been identified,
 International Review of Applied Economics 539

based on their identifiers, to produce a final balanced set of 15,761 firms for each
year, of which 5861 are foreign-owned and 9900 are domestically owned. A firm
has been defined to be domestically owned if its foreign equity participation, if any,
is below 25%.
We use three sets of spillover variables. FDISP, RDSP and EXSP represent spill-
overs due to the presence of foreign-owned firms, R&D and exports, respectively.
We use the share of foreign-owned firms’ capital in total capital in an industry, or
in a region, as the proxy for FDISP, while the ratio of exports by all other firms
(the firms’ own exports are excluded) to total sales in an industry or in a region for
EXSP. In terms of measuring R&D – an output indicator – intangible assets are
used as R&D expenditure is only available for year 2001. RDINT is therefore a
proxy for the level of domestic knowledge, measured as the ratio of intangible
assets to fixed assets in an industry or in a region, whilst the variable for R&D
spillovers (RDSP) is measured as the unweighted sum of the R&D stocks of all
Downloaded by [Nipissing University] at 23:18 03 October 2014

other firms; that is, the ratio of intangible assets held by all other firms (the firms’
own intangible assets are excluded) to fixed assets in an industry or in a region.
Furthermore, three sets of R&D-based absorptive capacity variables are constructed.
According to Griffith, Redding, and Van Reenen (2003, 2004), RDAB, EXAB and
FDIAB stand for the absorptive capacity related to own R&D and R&D spillovers,
export spillovers and FDI spillovers, respectively.
While most previous empirical studies focused on intra-industry spillovers only
(Acharya and Keller, 2009; Kneller, 2005; Zhu and Jeon, 2007), China is geograph-
ically large and consists of provinces/areas with substantial heterogeneity in eco-
nomic development and their endowment of the factors that are most commonly
held to influence both knowledge spillovers and absorptive capacity. Due to the fact
that China’s industrial development is more regionally oriented, three macro areas
can be broadly defined: (1) the coastal area (for example, Beijing, Tianjing, Hebei,
Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shangdong, Guangdong, Guangxi
and Hainan); (2) the central area including Hanxi, Inner Mongolia, Jilin,
Heilongjiang, Anhui, Jiangxi, Henan, Hubei and Hunan); and (3) the western area,
such as Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia and
Xinjiang. Domestic firms located in coastal provinces are, in general, considered to
have a higher degree of openness and a higher level of productivity (Wei and Liu
2006).

Empirical results
The results from the trans-log function using both non-frontier and frontier
approaches are presented in Tables 1 and 2. Table 1 summarises the results of intra-
industry knowledge spillovers, while Table 2 presents the results of inter-industry
knowledge spillovers across regions.
First, we start with the regression of production function using domestic R&D
as the only determinant of knowledge without controlling for differences in knowl-
edge spillovers and absorptive capacity, which is regression (1) in both tables. This
is a restricted version of equation (7) for comparison. Next, we estimate the impact
of knowledge spillovers through various channels, including foreign R&D, trade
and FDI in regression (2). Then we examine the magnitude and significance of
R&D-related absorptive capacity by adding the relevant explanatory variables
directly into the underlying production function in regression (3). In both tables, we
 Downloaded by [Nipissing University] at 23:18 03 October 2014

540

Table 1. Results from production function estimates for local firms in intra industry.
Non-frontier approach Frontier approach
(1) (2) (3) (4) 2-stage (5) 1-stage (6) 1-stage
⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄
LK 0.2853 0.2845 0.2852 0.2961 0.2977 0.2968⁄⁄⁄
H. Liao et al.

(0.006) (0.006) (0.006) (0.006) (0.005) (0.005)

⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄
LL 0.7286 0.7280 0.7272 ⁄⁄⁄ 0.6727 0.6757 ⁄⁄⁄ 0.6778⁄⁄⁄
(0.009) (0.009) (0.009) (0.008) (0.007) (0.007)
⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄
RDINT 0.0448 0.0429 0.0528 ⁄⁄ 0.0537 0.0871 ⁄⁄⁄ 0.021⁄⁄⁄
(0.009) (0.008) (0.023) (0.009) (0.013) (0.007)
⁄⁄⁄ ⁄⁄⁄
RDSP 1.1289 1.1489 ⁄⁄⁄ 1.2409 1.2469 ⁄⁄⁄ 1.1574⁄⁄⁄
(0.157) (0.158) (0.147) (0.1494) (0.151)
⁄⁄
EXSP 0.0483 0.062 ⁄ 0.0689 0.0775 ⁄⁄ 0.0973⁄⁄⁄
(0.038) (0.039) (0.036) (0.035) (0.0351)
⁄⁄
FDISP 0.0026 0.0025 0.0692 0.0829⁄⁄ 0.0927⁄⁄⁄
(0.037) (0.038) (0.036) (0.035) (0.035)

Lambda  1.9430 9.1260 3.3213

Sigma  1.7898 8.3622 3.0399
⁄⁄
RDAB 0.1719 0.0968 2.3825 3.2385⁄⁄⁄
(0.132) (0.047) (3.915) (0.218)
⁄⁄ ⁄⁄
EXAB 0.1487 0.0874 4.5406 0.4500⁄⁄⁄
(0.072) (0.042) (5.572) (0.0645)
FDIAB 0.0080 0.0204 4.7754
(0.054) (0.027) (6.526)

R2 0.5895 0.5902 0.5903

Note: 1. The number in brackets represents standard errors.
2. ⁄⁄⁄, ⁄⁄ and ⁄denote significance at the 1%, 5% and 10% levels, respectively.
 Downloaded by [Nipissing University] at 23:18 03 October 2014

Table 2. Results from production function estimates for local firms in inter industry.
Non-frontier approach Frontier approach
(1) (2) (3) (4) 2-stage (5) 1-stage (6) 1-stage
⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄
LK 0.2854 0.2829 0.2826 0.2932 0.2989 0.2896⁄⁄⁄
(0.006) (0.006) (0.006) (0.006) (0.005) (0.005)
⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄
LL 0.7283 0.7303 ⁄⁄⁄ 0.7304 0.6762 0.6711 ⁄⁄⁄ 0.6650⁄⁄⁄
(0.009) (0.009) (0.008) (0.008) (0.007) (0.007)
⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄
RDINT 0.0449 0.0414 ⁄⁄⁄ 0.0192 0.0483 0.0161 ⁄⁄⁄ 0.0115⁄
(0.008) (0.008) (0.003) (0.009) (0.006) (0.006)
⁄⁄⁄ ⁄⁄⁄
RDSP 1.7374 ⁄⁄⁄ 1.8416 2.2318 2.2324 ⁄⁄⁄ 2.2300⁄⁄⁄
(0.220) (0.225) (0.205) (0.206) (0.207)
EXSP 0.0368 0.0373 0.0313 0.0200 0.0303
(0.110) (0.111) (0.105) (0.105) (0.104)
⁄⁄⁄ ⁄⁄⁄
FDISP 0.9482⁄⁄⁄ 0.9220 0.8381 0.8300 ⁄⁄⁄ 0.8095⁄⁄⁄
(0.086) (0.087) (0.082) (0.084) (0.083)

Lambda 1.8983 2.9315 2.5700

Sigma 1.7615 2.7619 2.3627
⁄⁄
RDAB 0.6179 0.4866 ⁄⁄⁄ 7.7360⁄⁄⁄ 6.5647⁄⁄⁄
(0.302) (0.138) (2.427) (0.961)
EXAB 0.0949 0.0369 2.4911
(0.179) (0.079) (2.018)
⁄
FDIAB 0.3077 0.1114⁄ 6.7553⁄⁄⁄ 4.2045⁄⁄⁄
(0.170) (0.069) (2.032) (0.686)

R2 0.5878 0.5987 0.5989

International Review of Applied Economics

Note: 1. The number in brackets represents standard errors.

2. ⁄⁄⁄, ⁄⁄ and ⁄denote significance at the 1%, 5% and 10% levels, respectively.
541
 542 H. Liao et al.

have two empirical settings – non-frontier and frontier approach. For the former, all
relevant variables are included in the underlying production function while the latter
views absorptive capability as the determinant of inefficiency involved, which in
turn has been assessed via either two-stage estimation in regression (4) or 1-stage
estimation in regressions (5) and (6). The results are fairly robust to changes in
specifications, with most estimated parameters being qualitatively similar for these
regressions. Therefore, in order to avoid repetition, similar coefficients will not be
taken into account for other regressions where there is little difference after discus-
sion of the parameter estimates in Table 1. This will allow us to concentrate on the
overall results and differences between the regressions and settings.
We estimated the production functions/frontiers and obtained important and
interesting results. First, most of the parameter estimates are significant as
expected.1 Second, labour contributes significantly to the output. The high labour
input share ranging from 0.67 to 0.73 reflects the labour-intensive nature of the
Downloaded by [Nipissing University] at 23:18 03 October 2014

manufacturing activities of Chinese firms. This result is in line with the results
reported in several previous studies. For example, Young (1992) estimated a labour
share of 0.609 for the Hong Kong economy during the period 1986–91. Wang and
Tsai (2003) also obtained an average 0.65 estimate for labour share at firm level for
Taiwan. Third, the significance of the sigma2 parameter in regressions (4) and (5)
indicates that there was substantial production below the frontier, and the value of
lambda implies that TE is also important in explaining the total variability of out-
put. Hence, the production frontier is significantly different from the traditional
average functions – regressions (1) and (3). We also note that the results are fairly
robust to changes in specifications, with most estimated parameters being qualita-
tively similar for these regressions.3
The estimated elasticities of domestic R&D are positive and statistically
significant in all the regressions, in the range of what one would expect. According
to the results, a 1 percentage point increase in domestic R&D raises productivity
and therefore output by a range from 0.042 to 0.087 of a percentage point. When
only domestic R&D is considered in regression (1) the elasticity is 0.045, which is
in line with the literature, although our estimate, is on the lower boundary. For
example, Griliches (1988) suggested that the estimated elasticity of domestic R&D
lies between 0.06 and 0.1. Although allowing for technology transfer and heteroge-
neities in absorptive capability has little impact on the sign and significance of the
estimated elasticities of domestic R&D through regressions (2) to (6), the more
noticeable impact is on the magnitude of those elasticities. They tend to be larger in
intra-industry analysis but yield mixed results in inter-industry estimations in
Table 2.
Although under the frontier setting, two-stage estimation in regression (4) and
one-stage estimation in regressions (5) and (6) yield qualitatively similar results, we
can still offer an interesting interpretation of them. As mentioned above, two-stage
analysis means the generic frontier is estimated without the covariates, then in the
second step we regress the estimated (in)efficiencies on the covariates. Wang and
Schmidit (2002) have cautioned against this, arguing that the omission of the covar-
iates at the ‘first step’, is tantamount to the omitted variable problem in the ordinary
regression.
Given the main purpose of this study, we are particularly interested in the
behaviour of knowledge spillover variables RDSP, EXSP and FDISP, and R&D-
related absorptive capacity variables RDAB, EXAB and FDIAB. As shown in
 International Review of Applied Economics 543

Table 1, the significant coefficient on RDSP in all intra-industry regressions

indicates that there is evidence of intra-industry R&D spillovers, which also holds
for inter-industry regressions as shown in Table 2.
Also from Table 1, EXSP is negative and significant in the ‘industry’ regres-
sions with the exception of regression (2) in which the estimated coefficient of
EXSP is still negative but insignificant. These results are very interesting, as they
may be closely associated with some specific features of China as a transitional
economy. Exporting has been adopted as an outward orientated development strat-
egy, and exporting activities are highly encouraged by the Chinese government. For
example, one of the governmental incentives to encourage exporting is the refund
of value added tax for exported products. Partly because of this policy, both
indigenous and foreign-invested firms compete for exports of products in the same
industries, often at reduced prices. As a result, exports may generate negative intra-
industry competition effects, which tend to lower the productivity of indigenous
Downloaded by [Nipissing University] at 23:18 03 October 2014

Chinese firms and therefore tend to lower the production possibility frontier of
indigenous Chinese firms. On the other hand, firms’ absorptive capabilities are
enhanced through exporting, which can be seen from the significant negative
coefficients of the EXAB variable in Table 1, which is closer to the frontier. The
productivity is the sum of lowered frontier plus the enhanced absorptive capabili-
ties, and the final result depends on which factor is dominant.
All coefficients on EXSP in Table 2 are negative but insignificant in the ‘region’
regression or the inter-industry estimation. The results may reflect the fact that
regional trade barriers are evident (Meyer 2008). The introduction of fiscal decen-
tralization in China created strong incentives for local governments to develop
regional economies and to protect local firms from regional competition (Lu and
Tao 2009). Local protectionism impedes firms from freely trading across provinces,
and separates China into one country but many regional economies (Meyer 2008).
As a result, inter-industry linkages across regions are relatively weak (Fu 2004;
Girma and Gong 2008b). This may be one of the reasons why the coefficients on
EXSP in Table 2 are insignificant. Our results are consistent with those of previous
studies (e.g. Wei and Liu 2006; Li and Wu 2010).
Turning to spillover effects from FDI, the coefficient on FDISP is negative but
statistically insignificant in the non-frontier setting of the ‘industry’ regressions 1(2)
and 1(3), and positive and highly significant in the frontier setting of the ‘industry’
regressions through 1(4) to 1(6) and all ‘region’ regressions through 2(2) to 2(6).
This suggests that evidence of intra-industry productivity spillovers from FDI across
regions is either mixed or weak, but there are strong inter-industry positive spill-
overs. Put another way, like R&D and exports, positive productivity spillovers from
FDI are fundamentally localised rather than national, as barriers to free flow of
goods and services across regions contribute significantly to the existence of posi-
tive FDI spillovers on a local rather than a national scale. The spillover effects from
FDI would have been stronger without regional barriers.
The regression results may also reflect the fact that there is an enhanced rela-
tionship between technology transfer and technology absorption. Due to the fact
that knowledge spillovers through FDI are either mixed or weak in the ‘industry’
regression, the variable on FDI-related absorptive capacity is dropped in regression
1(6). The same logic applies to the ‘region’ regression in regression 2(6). As can be
seen, all coefficients of these (in)efficiency terms now have the expected negative
sign and are statistically significant. That is, RDAB and EXAB in Table 1, and
 544 H. Liao et al.

RDAB and FDIAB in Table 2 help reduce the gap between the best-practice
frontier and the actual productivities of domestic firms. In other words, absorptive
capability is also the key factor determining indigenous firms’ productivities. This
result supports the arguments made by Griffith, Redding, and Van Reenen (2003)
that although all countries/firms behind the technological frontier have the potential
to achieve productivity growth through technology transfer, the realisation of this
potential depends on institutions and government policy, which affect the pace of
autonomous technology transfer as well as R&D-based absorptive capacity. By
comparing the estimation results from various specifications, we have also found
that when absorptive capability is taken into account, both underlying production
frontiers’ parameters and knowledge spillover effects remain qualitatively
unchanged. This result partially provides evidence that the two facets of R&D dis-
cussed by Griffith, Redding, and Van Reenen (2004) do receive empirical support
when using stochastic frontier analysis. This result is also similar to that found in
Downloaded by [Nipissing University] at 23:18 03 October 2014

Kneller and Stevens (2006) where human capital is one of the determinants for (in)
efficiency along with the usual R&D variable, and in line with the previous findings
that in-house R&D facilitate technology transfer in China (Hu, Jefferson, and Qian
2005; Liu and Buck 2007).

Conclusion
This paper empirically investigates how technological knowledge can be generated,
transmitted and absorbed through different channels in a single framework. Such an
approach helps capture the integrative learning process of local firms. Moreover, by
considering the two facets of R&D with regard to TP and TE using the stochastic
frontier estimations, we have differentiated the absorptive capacity of local firms in
relation to different spillover channels, and this specification enables us to provide
new insights into how different dimensions of absorptive capacity moderate the
impact of various spillover channels.
We find that while there are positive inter- and intra-industry productivity spill-
overs from R&D and strongly positive inter-industry spillovers from foreign presence
to local Chinese firms, evidence of intra-industry productivity spillovers from FDI is
less robust. However, our findings show a negative impact of intra-industry export
spillovers on productivity. With regard to R&D-related absorptive capacity, the find-
ing shows that the ability and/or efforts to absorb new technology are also one of key
determinants to quantitatively explain both intra- and inter-industry differences in
productivity. Our findings provide further evidence for Nishimizu and Page’s (1982)
argument for the decomposition of TFP. Failure to take account of inefficiency
(absorptive capacity in our paper) may produce misleading and biased TFP estimates.
The findings from our study offer some important policy implications. First,
in-house R&D is found to be an important determinant of the productivity of local
Chinese firms. Hence, the government should provide sufficient incentives to
encourage local Chinese firms to invest in R&D. Our findings also suggest that the
influence of technology spillovers is not automatic, but it is linked with deliberate
learning and active interaction between domestic R&D and foreign firms. Therefore,
improving absorptive capacity and promoting organisational learning may be an
effective way to extract maximum spillover effect from FDI. Second, our results
show limited evidence of export spillovers. It may imply that further liberalisation
of trade should be the top priority for Chinese government if China desires to
 International Review of Applied Economics 545

benefit more from export spillovers. Finally, it may be crucial for the government
to adopt a combined strategy that encourages local Chinese firms to obtain new
technology via various spillover channels, while at the same time move towards
developing their own technological capabilities through internal effort.
There are some limitations in the study. Our research is based on secondary
firm-level data. The data availability constrained the scope of the study, which has
mainly investigated R&D, exports and inward FDI as the main channels for tech-
nology spillovers. Further studies should include imports in the estimation and pro-
vide a complete picture of channels for external knowledge spillovers in a unified
framework. In addition, our study only covers the period 1998–2001. Future studies
should use a more recent dataset to further enhance the current analysis.

Notes
Downloaded by [Nipissing University] at 23:18 03 October 2014

1. Since we have already mean-corrected the data prior to the estimation, the estimated
first-order parameters in the (semi-) translog function/frontier can be directly interpreted
as estimates of the production elasticities, evaluated at the sample mean. Estimated coef-
ficients of the square variables included in the equation (7) are available upon request.
2. The estimation of parameters in the stochastic frontier model given by equations (7) and
(8) are carried out via the maximum-likelihood method, using LIMDEP. Instead of
directly estimating r2v and r2u . LIMDEP seeks estimates of r ¼ ðr2v þ r2u Þ1=2 and
k ¼ ru =rv . These are associated with the variances of the stochastic term in the produc-
tion function and inefficiency term. These estimates are the benchmark used to perform
the tests and are referred to as the ‘full’ model. For instance, if the hypothesis k=0 is
accepted, this would indicate that ru is zero and thus the efficiency error term, uit should
be removed from the model, leaving a specification with parameters that can be consis-
tently estimated by OLS.
3. Two sets of tests are performed to identify the adequate functional form and the pres-
ence of inefficiency by imposing restrictions on the parameters of the ‘full’ function.
The likelihood ratio (LR) tests are used as criteria. Each time a set of restrictions is
imposed, a new system of equations is created that is nested in the unrestricted model.
Results are available upon request.

References
Acharya, R., and W. Keller. 2009. Technology transfer through imports. Canadian Journal
of Economics 42, no. 4: 1411–48.
Aghion, P., and P. Howitt. 1998. Endogenous growth theory. Cambridge, MA: MIT Press.
Aigner, D., C.A.K. Lovell, and P. Schmidt. 1977. Formulation and estimation of stochastic
frontier production function models. Journal of Econometrics 6, no. 1: 21–37.
Aitken, B., and A. Harrison. 1999. Do domestic firms benefit from foreign direct invest-
ment? Evidence from Venezuela. American Economic Review 89: 605–18.
Bitzer, J., and I. Geishecker. 2006. What drives trade-related R&D spillovers? Decomposing
knowledge-diffusing trade flows. Economics Letter 93: 52–7.
Blalock, G., and D. Simon. 2009. Do all firms benefit equally from downstream FDI? The
moderating effect of local suppliers’ capabilities on productivity gains. Journal of Inter-
national Business Studies 40, no. 7: 1095–112.
Blalock, G., and P. Gertler. 2004. Learning from exporting revisited in a less developed
setting. Journal of Development Economics 75: 397–416.
Borensztein, E., J. Gregorio, and J.-W. Lee. 1998. How does foreign direct investment affect
economic growth. Journal of International Economics 45: 115–35.
Buckley, P., J. Clegg, and C. Wang. 2007. Is the relationship between inward FDI and spill-
over effects linear? An empirical examination of the case of China. Journal of Interna-
tional Business Studies 38: 447–59.
 546 H. Liao et al.

Castellani, D., and A. Zanfei. 2003. Technology gaps, absorptive capacity and the impact of
inward investments on the productivity of European firms. Economics of Innovation and
New Technology 12, no. 6: 555–76.
Coe, D.T., and E. Helpman. 1995. International R&D spillovers. European Economic Review
39, no. 5: 859–87.
Coe, D.T., E. Helpman, and A.W. Hoffmaister. 1997. North-south R&D spillovers. Economic
Journal 107: 134–49.
Cohen, W.M., and D.A. Levinthal. 1989. Innovation and learning: The two faces of R&D.
Economic Journal 99: 569–96.
Eaton, J., and S. Kortum. 1999. International patenting and technology diffusion. Interna-
tional Economic Review 40: 537–70.
Freeman, C. 1987. Technology policy and economic performance: Lessons from Japan.
London: Pinter Publishers.
Fu, X. 2004. Limited linkages from growth engines and regional disparities in China.
Journal of Comparative Economics 32: 148–64.
Girma, S., and Y. Gong. 2008a. Putting people first? Chinese state-owned enterprises’ adjust-
Downloaded by [Nipissing University] at 23:18 03 October 2014

ment to globalization. International Journal of Industrial Organization 26, no. 2:

573–85.
Girma, S., and Y. Gong. 2008b. FDI, linkages and the efficiency of state-owned enterprises
in China. Journal of Development Studies. 44, no. 5: 728–49.
Girma, S., and K. Wakelin. 2007. Local productivity spillovers from foreign direct invest-
ment in the UK electronics industry. Regional Science and Urban Economics 37, no. 3:
399–412.
Griffith, R., S. Redding, and J. Van Reenen. 2003. R&D and absorptive capacity: Theory
and empirical evidence. Scandinavian Journal of Economics 105, no. 1: 99–118.
Griffith, R., S. Redding, and J. Van Reenen. 2004. Mapping the two faces of R&D: produc-
tivity growth in a panel of OECD industries. Review of Economics and Statistics 86:
883–95.
Griliches, Z. 1988. Productivity puzzles and R&D: another non-explanation. Journal of
Economic Perspectives 2: 9–21.
Grunfeld, L. 2006. Multinational production, absorptive capacity, and endogenous R&D
spillovers. Review of International Economics 14, no. 5: 922–40.
Hu, A., G. Jefferson, and J. Qian. 2005. R&D and technology transfer: Firm level evidence
from Chinese industry. Review of Economics and Statistics 87, no. 4: 780–6.
Keller, W. 1998. Are international R&D spillovers trade-related? Analyzing spillovers among
randomly matched trade partners. European Economic Review 42: 1469–81.
Keller, W. 2002. Geographic localisation of international technology diffusion. American
Economic Review 92: 120–42.
Kneller, R. 2005. Frontier technology, absorptive capacity and distance. Oxford Bulletin of
Economics and Statistics 67, no. 1: 1–23.
Kneller, R., and P.A. Stevens. 2006. Frontier technology and absorptive capacity: Evidence
from OECD manufacturing industries. Oxford Bulletin of Economics and Statistics 68,
no. 1: 1–21.
Koop, G., J. Osiewalski, and M. Steel. 1999. The components of output growth: A stochastic
frontier analysis. Oxford Bulletin of Economics and Statistics 61, no. 40: 455–87.
Kumbhakar, S.C., and C.A.K. Lovell. 2000. Stochastic frontier analysis. New York:
Cambridge University Press.
Lichtenberg, F., and B. van Pottelsberghe de la Potterie. 2001. Does foreign direct invest-
ment transfer technology across borders? Review of Economics and Statistics 83: 490–7.
Li, X., and G. Wu. 2010. In-house R&D, technology purchase and innovation: Empirical
evidences from Chinese hi-tech industries, 1995-2004. International Journal Technology
Management 51: 217–38.
Liu, X., and T. Buck. 2007. Innovation performance and channels for international technol-
ogy spillovers: Evidence from Chinese high-tech industries. Research Policy 36, no. 3:
355–66.
Liu, X., C. Wang, and Y. Wei. 2009. Do local manufacturing firms benefit from transactional
linkages with multinational enterprises in China? Journal of International Business Stud-
ies 40, no. 7: 1113–30.
 International Review of Applied Economics 547

Lu, J., and Z. Tao. 2009. Trends and determinants of China’s industrial agglomeration. Jour-
nal of Urban Economics 65: 167–80.
Lumenga-Neso, O., M. Olarreaga, and M. Schiff. 2005. On indirect trade-related R&D
spillovers. European Economic Review 49: 1785–98.
Madden, G., S. Savage, and P. Bloxham. 2002. Asian and OECD international R&D spill-
overs. Applied Economics Letters 8: 431–5.
Meeusen, W., and J. van den Broeck. 1977. Efficiency estimation from Cobb-Douglas pro-
duction functions with composed error. International Economic Review 18, no. 2:
435–44.
Meyer, M.W. 2008. China’s second economic transition: Building national markets. Manage-
ment and Organization Review 4: 3–15.
Miller, S.M., and M.P. Upadhyay. 2000. The effects of openness, trade orientation, and
human capital on total factor productivity. Journal of Development Economics 63:
399–423.
Nishimizu, M., and J.M. Page. 1982. Total factor productivity growth, technological progress
and technical efficiency change: Dimensions of productivity change in Yugoslavia,
Downloaded by [Nipissing University] at 23:18 03 October 2014

1965-78. Economic Journal 92: 920–36.

Okabe, M. 2002. International R&D spillovers and trade expansion: Evidence from East
Asian economies. ASEAN Economic Bulletin 19, no. 2: 141–54.
Qi, J., Y. Zheng, J. Laurenceson, and H. Li. 2009. Productivity spillovers from FDI in
China: Regional differences and threshold effects. China & World Economy 17, no. 4:
18–35.
Romer, P.M. 1990. Endogenous technological change. Journal of Political Economy 98, no.
5: 71–102.
Salmon, R., and J. Shaver. 2005. Learning by exporting: New insights from examining firm
innovation. Journal of Economics & Management Strategy 14: 431–60.
Savvides, A., and M. Zchariadis. 2004. International technology diffusion and the growth of
TFP in the manufacturing sector of developing economies. Mimeo.
Schumpeter, J.A. 1934. The theory of economic development. Cambridge, MA: Harvard
University Press.
State Statistical Bureau of China. 1999, 2000, 2001, 2002. Annual report of Industrial Enter-
prise Statistics. State Statistical Bureau of China.
Wang, H., and P. Schmidt. 2002. One step and two step estimation of the effects of exoge-
nous variables on technical efficiency levels. Journal of Productivity Analysis 18:
129–44.
Wang, J., and K. Tsai. 2003. Productivity growth and R&D expenditure in Taiwan’s manu-
facturing firms. NBER Working Paper 9724.
Wei, Y., and X. Liu. 2006. Productivity spillovers from R&D, exports and FDI in China’s
manufacturing sector. Journal of International Business Studies 37: 544–57.
Westphal, L. 2002. Technology strategy for economic development in a fast changing global
economy. Economics of Innovation and New Technology 11: 275–320.
Young, A. 1992. A tale of two cities: Factor accumulation and technical change in Hong
Kong and Singapore. NBER Macroeconomics Annual: 13–54.
Zhu, L., and B.N. Jeon. 2007. International R&D spillovers: Trade, FDI and information
technology as spillover channels. Review of International Economics 15, no. 5: 955–76.