Cambridge Journal of Regions, Economy and Society 2019, 12, 401–422

doi:10.1093/cjres/rsz012
Advance Access publication 5 October 2019

Industrial Policy in China: The Planned Growth of

Specialised Towns in Guangdong Province
D
o
w
nl
Elisa Barbieria, Marco R. Di Tommasob, Chiara Polliob oa
and Lauretta Rubinib de
d
a fro
Department of Economics, University of Venice Ca’ Foscari, Fondamenta S. Giobbe, m
873, 30100 Cannaregio, Venezia, Italy. elisa.barbieri@unive.it htt
b
Department of Economics and Management, University of Ferrara, Via Voltapaletto, ps
://
11, 44121 Ferrara, Italy. marco.di.tommaso@unife.it, chiara.pollio@unife.it, ac
lauretta.rubini@ unife.it ad
e
Received on September 20, 2018; accepted on July 4, 2019 mi
c.
ou
p.
We analyse one of the most important policy experiences for industrial clustering in co
Southern China—the Specialised Towns programme—that has transformed some Chinese m/
cjr
clusters into the backbone of global production chains. We offer a long-term, detailed over-
es
view of the policy programme and of Guangdong’s specialised towns, classifying them as /ar
endogenous or exogenous according to their features, and investigate their contribution to ticl
local growth and rebalancing. This analysis of the Specialised Towns programme contrib- e-
utes to the international debate on revisiting industrial policy and suggests that the discus- ab
sion should conceive them as articulated processes to reach long-term societal objectives. str
ac
Keywords: China, Guangdong, industrial policy, specialised towns, clusters t/1
2/
JEL Classifications: O18, O25, R12, R58
3/
40
1/
55
81
Introduction areas in a sort of “controlled” 69
China is today at the centre of several global industrialisation. The underlying idea was that 7
by
net- works of production. Starting from the such develop- ment would then trigger the U
launch of the open-door policy in 1978, the rest of the country (Groenewald et al., 2008; ni
country’s growth and structural change have Knight, 2013). ve
rsi
been re- markable and fast. While for some Simultaneously, since the early 1980s, policy da
scholars this has happened despite state actors have interacted with overseas capitals, d
intervention (Nee and Opper, 2007; Parish allowing them to build large and globally rele- E
and Michelson, 1996), recent contributions vant production bases in the country. The A
N
emphasise the role of planning and policy Guangdong Province, Southeast China, offers us
activities of national and provincial a special viewpoint on this interplay. Half as
governments (Veek et al., 2016; Yueh, 2013). large as Germany but with 20 million more
This process has assumed a pre- cise spatial people, Guangdong is the leading province for
shape. The open-door policy inten- tionally its contribution to national GDP—10.8%, cor-
favoured the initial growth of selected responding to 1.3 billion US$ in 2017. A feature

© The Author(s) 2019. Published by Oxford University Press on behalf of the Cambridge Political Economy Society.
All rights reserved. For permissions, please email: journals.permissions@oup.com
Barbieri et al.

of the province’s growth is its outward ST of Shiling; 30% of the toys manufactured
orienta- tion. In 2017, Guangdong accounted globally
for nearly 30% of China’s national exports1
and 15% of its total foreign direct investment
D
(FDI).2 It had a leading role during the open- o
door policy. When, in 1978, Deng Xiaoping w
launched it, Guangdong and Fujian were the nl
oa
first provinces to experi- ment the “special de
policies, flexible measures” (Di Tommaso et d
al., 2013; Li, 1997). Such ini- tiatives were at fro
the core of Xiaoping’s plan of Chinese m
htt
socialist economy modernisation, based on ps
gradual opening up to the global cap- italistic ://
market, via trade and FDI, and power ac
ad
delegation to local governments. Guangdong e
was chosen as a pilot area for two reasons: (i) mi
its strategic location, near Hong Kong and c.
Macao, which was considered a booster to ou
p.
FDI and to the diffusion of entrepreneurial co
practices [Barbieri et al., 2009; Enright et al., m/
2005; Yeung et al. (1998)]—indeed, between cjr
es
1988 and 1999, Hong Kong and Macao /ar
accounted for more than 66% of total FDI ticl
directed to the region, 3 detaching themselves e-
ab
by far from the other Asian contributors—and
str
(ii) its initial low eco- nomic performance, ac
reducing the risk to test the opening-up t/1
process here with respect to other more 2/
3/
powerful areas, such as Shanghai (Chung, 40
1998; Di Tommaso et al., 2013). In this 1/
context, national and local governments 55
81
devoted special attention to the spatial 69
organisation of pro- duction through spatially 7
targeted incentives (Rubini et al., 2015; Zeng, by
2010, 2012, and so on). As a result, U
ni
Guangdong’s industrialisation pattern is today ve
largely based on industrial ag- glomerations rsi
(Lai et al., 2005; Zeng, 2010). da
d
In this article, we focus on a crucial indus- E
trial policy initiative for industrial clustering A
in Guangdong—the so-called Specialised N
us
Towns4 (STs) programme. By July 2017, there
were 416 STs, accounting for 40% of the
provincial GDP. In some prefectures, such as
Dongguan or Foshan, STs represent 90% of
GDP. Furthermore, 70% of the European and
US mass-market luggage is produced in the

402
 Industrial policy in China

come from the ST of Chenghai unique and coherent policy programme.6

(Jankowiak, 2017) and 20% of 2. Evolution from industrial park to cluster
smartphones manufactured globally promotion. The policy originally fostered
every year comes from Dongguan City in- dustrial parks, but, after learning from
(where 93% of the total townships the successful experiences of other
participate in the programme). 5 This countries’
means that a signifi- cant part of
consumption in the West is today
possible because these Chinese clusters
have emerged, with the support of this
programme.
Gradually, Guangdong has become
econom- ically polarised [Organization of
Economic Cooperation and Development
(OECD), 2010]. To correct this, since 2008,
policymakers have used the STs
programme also to promote within-
province rebalancing (Di Tommaso et
al., 2013), encouraging specialisations
more related to existing local economic
and social contexts (Su and Sun, 2016).
International scholars have
increasingly fo- cussed on China’s
clustering and spatial indus- trial
agglomerations (for example, Barbieri
et al., 2010; Butollo, 2015; Long and
Zhang, 2012; Wang et al., 2010; Zhu et
al., 2019). However, the literature on the
specific case of Guangdong’s STs
programme is surprisingly small (cf.
Bellandi and Caloffi, 2010; Bellandi and
Di Tommaso, 2005; Barbieri et al., 2012).
To the best of our knowledge, there is no
systematic up-to-date overview of the
phenomenon as a whole and no
international contribution with empirical
evidence on the linkages between the
STs pro- gramme and local economic
performances.
We attempt to fill this gap by offering a
long-term, detailed overview of the STs
pro- gramme, which we believe is worth of
investi- gation for its distinctive features:

1. Number of towns involved. As far as

we know, no other place worldwide
has sup- ported such a high number of
clusters over such a long time with a

403
 clusters, the Chinese government restruc- The article is organised as follows. After
tured it to support, drive and accelerate dis- cussing the international literature useful
clus- ters growth (Rubini et al., 2015). to in- terpret the STs’ experience, we describe
3. Central role of the Department of Science data and methodology. Next, we offer a D
and Technology of Guangdong detailed analysis of the programme and its o
Government (DSTGG). Since the launch of evolution and analyse the features of w
the pro- gramme, innovation has been at nl
endogenous versus exogenous STs; we, then oa
the centre of the policy through publicly present and discuss the empirical analysis. de
funded innov- ation platforms for the The article ends with some implications for d
specialised sector (also when low-tech policy design and for the inter- national
fro
productions). m
debate on industrial policy, with ideas for htt
4. Changing long-term aims. From promo- future research lines. ps
tion of specialisation, competitiveness and ://
investment rationalisation to territorial ac
ad
rebalancing. Specialised towns in the context of e
industrial clusters mi
To encouraging local growth, STs have also c.
Since Marshall (1890), the economic advan- ou
been used as means for structural transform-
tages related to industrial clusters7 have been p.
ation, economic upgrading and promotion co
widely discussed (for example, Boschma and
of general economic and societal objectives m/
Fornahl, 2011; Delgado et al., 2014; Martin cjr
(Department of Science and Technology of
and Sunley, 2003; Porter, 2000; Vicente, es
China (DSTC), 2011a, 2011b; DSTGG, 2003, /ar
2018).8 Recently, the debate has interested
2006a, 2006b, 2008). In this sense, these tools ticl
experts on both the industrialised world e-
have a broader industrial policy interpretation. (Eisingerich et al., 2010; Slaper et al., 2018) ab
Hence, our contribution is also valuable for and emerging economies (Kesidou and str
the international debate on revisiting indus- ac
Szirmai, 2008; Nadvi, 1999; Van Dijk and
trial policies (Bailey and Cowling, 2006; t/1
Rabellotti, 2005). Although scholars do not 2/
Chang, 1994; Cimoli et al., 2009; Rodrik, unanimously agree on these ef- fects, there 3/
2007; Vicente, 2018), which, we argue, needs 40
seems to be growing consensus that clusters’
to look to new economic giants’ experience, 1/
contribution to local economic per- formance 55
such as that of China, often neglected by is idiosyncratic to the hosting context (Puig 81
Western scholars (Heilmann and Shih, 2013). and González-Loureiro, 2017). 69
We focus on two empirical research ques- 7
Since Becattini’s seminal contributions by
tions: (i) Is the ST programme associated with
(Becattini, 1987, see also Becattini et al., 2014) U
improved territorial economic performances?
on industrial districts and Schmitz’s (1999) on ni
(ii) Can it contribute to territorial rebalancing ve
clusters, a well-established literature has as-
within the province? With the first question, rsi
serted the relevance of joint actions and trust da
we investigate the capability of local
to foster collective efficiency and produce d
industrial policy to promote local growth; E
positive effects of clustering on performance
with the second, we tackle the relationship A
(Porter, 2000). To achieve collective efficiency, N
between industrial policy and territorial
co-located economic actors need to engage in us
rebalancing (Bellandi and Di Tommaso, 2005;
joint actions, either spontaneously or following
Bianchi et al., 2000). In doing so, we focus
policy incentives.
on the so-called endogenous clusters, as they
A specific literature has focussed on clus-
might be interpreted to some extent as a form
tering as a target of industrial policies, given
of place-based policy (Bailey et al., 2015).
their potential economic gains (Humphrey
and Schmitz, 1996; Rodriguez-Clare, 2007).
Some
scholars have also warned that without proper misses some districts’
design and implementation, cluster policies
might even produce negative consequences
(Brakman and van Marrewijk, 2013). Within
D
such literature, some papers use case studies o
to analyse the effect of clustering policies w
(Elola et al., 2017; Kiese, 2017). However, nl
oa
the evidence is still too limited and difficult to de
interpret given the complex design of these d
schemes and the variety of their effects fro
(Uyarra and Ramolgan, 2016; Vicente, 2018). m
htt
Within this debate, China is particularly ps
interesting for the widespread use of spatially ://
tar- geted initiatives (Wang, 2013), ac
ad
heterogeneous for types and aims (Barbieri et al., e
2012). In the case of the STs, Chinese mi
policymakers have drawn on previous c.
ou
international clustering experiences to enhance p.
existing industrial agglomeration and co
specialisation, typical of clustering policies m/
(Ketels, 2013). cjr
es
In doing so, policymakers have been /ar
focussing on promoting upgrading, starting ticl
from local production contexts. In this sense, e-
ab
the STs’ experience recalls the current de- str
bate about place-based policies (Bailey et al., ac
2015; Barca et al., 2012; Hildreth and Bailey, t/1
2014; OECD, 2011), which recognises the 2/
3/
im- portance of places’ context in shaping and 40
pro- moting development (Garretsen et al., 1/
2013; Hildreth and Bailey, 2013). The 55
81
recognition of local productive specialties at 69
the centre of the ST programme is in line with 7
place-based and space-tailored programmes. by
However, STs detach from place-based U
ni
policies in many re- gards. While the ve
interaction between local and provincial rsi
governments is consistent with the multilevel da
d
governance strategy of place-based E
approaches, there is no proof that other rele- A
vant stakeholders, such as firms and workers, N
participate in the design and implementa- us
tion of the STs scheme. Additionally,
although Guangdong policymakers have
considered the experiences of Italian districts
in designing the programme (Di Tommaso et
al., 2013), the evo- lution of these places
key elements, in particular the existence
of a community of firms and people
giving rise to the industrial atmosphere A note on data and methodology
(Becattini, 1987). In Guangdong, in fact, The remainder of the paper adopts a mixed-
clusters often result from a joint action method approach (Cairns, 2018; Goertz,
of planned policies and external forces, 2017;
such as foreign capitals and migrations.
In this context, STs can be classified as
ex- ogenous or endogenous (Barbieri et
al., 2009; Di Tommaso and Bazzucchi,
2013; Lyu and McCarthy, 2015; Zeng,
2010; Zhu et al., 2017). The exogenous
clusters mainly grew attracting FDI, and
policies aimed at facilitating the rela-
tionships between foreign and local
firms. The endogenous clusters stemmed
from either the development of local
firms, the evolution of ancient
productive systems, the privatisation of
town and villages enterprises or some
loca- tional and geographical
advantages.
The few studies on STs in English
illustrate the rationale and the content of
the policy (Barbieri et al., 2012; Bellandi
and Di Tommaso, 2005; Eng, 1997; Long
and Zhang, 2012) but not its possible ef-
fects on the territory and, in particular, the
possible heterogeneity of endogenous
versus exogenous clusters. In particular, we
expect two integrated processes to be at
stake in endogenous STs:

1. Being rooted in the local territory,

history and values, cluster actors are
more likely to engage in mutual trust
relationships and col- laboration—key
success factors of industrial districts
(Becattini, 1987; Mistri and Solari,
2003 and many others);
2. Instituting some forms of collective
action initiatives, such as innovation
platforms, will reasonably have higher
additional impact in terms of
collective action on endogenous ra-
ther than exogenous clusters, thanks
to the embeddedness of relevant actors
to the local context.
Hesse-Biber and Johnson, 2015) by using developing new
both qualitative and quantitative data.
To analyse the STs, we built a novel data-
base, containing information about the year D
of recognition, location and specialisation on o
399 towns established by 2016.9 It results from w
several fieldwork visits, from 2004 to mid-2017, nl
oa
allowing us to collect official documents and de
statistics; to interview provincial and local pol- d
icymakers, economic actors, scholars and think fro
m
tanks as well as to visit some STs.10 htt
In the empirical part, we matched this data- ps
base with county- and district-level data on ://
ac
economic performances from the Guangdong ad
Statistical Yearbooks [National Bureau of e
Statistics (NBS) 2001–2016]. The result was mi
c.
a panel data set of 75 counties and districts in ou
Guangdong from 2000 to 2015. p.
co
m/
The experience of Guangdong’s cjr
es
specialised towns /ar
ticl
The STs programme consists of an ex-post e-
offi- cial certification by the provincial ab
government, provided that the applicant town str
meets three criteria (Di Tommaso et al., ac
t/1
2013): 2/
3/
1. It must be a township from an 40
1/
administrative point of view or, less 55
frequently, a county or urban district; 81
2. At least 30% of its industrial output (or 69
7
employment) must be concentrated in one by
“specialized sector” (defined to the three- U
digit-level equivalent of international ni
ve
classi- fication systems); rsi
3. The annual industrial output value has to da
exceed 2 billion yuan. d
E
A
With the certification, the specialised town N
re- ceives a 30,000 yuan subsidy from the us
DSTGG, provided an additional proportion be
financed by the local government11 (Wang and
Yue, 2010). These funds are to establish the
so-called tech- nology innovation platforms—
the core of the STs policy—to assist firms in
technologies and upgrading production. They effects on the whole
also encourage cooperation among the town’s
firms, private and public engineering centres
and public research institutions in innova- tive
activities to improve the reputation of the
town’s entire productive system (Barbieri et
al., 2010; DSTGG, 2003, 2006a, 2017;
Guangdong
Academy of Social Sciences (GDASS), 2017;
Wang, 2004). The policymakers we
interviewed have stressed that, to increase
efficiency, the DSTGG has pushed the
innovation centres to operate following a
competitive market logic: several centres are
created and encouraged to compete in serving
the companies in the town. Meanwhile, their
general strategies are policy driven: the
political funding actors keep con- trol over
their operations through the board of
directors, which is in charge of defining the
strategic priorities and areas of intervention. In
2015, the innovation platforms have
developed 620 projects for a total value of
above 3 million yuan (Su and Sun, 2016).
According to the most recent data, the
average degree of specialisation of the ex-
isting 416 towns exceeded by about 10% the
minimum 30% required by law (GDASS,
2017). Figure 1 shows their localisation and
some data about their growth in numbers over
time.

Specialised towns’ development

and growth
We identify four phases in the STs
programme’s evolution. The first dates back to
the 1990s, be- fore the initiative’s official
launch. The pro- gramme stemmed from a
previous national government intervention—
the Spark Plan— which founded several
Technology Investment Zones in rural areas to
upgrade agriculture production and promote a
gradual structural shift towards manufacturing
(Barbieri et al., 2009; DSTGG, 2006b; Zhang
and Ling, 2003). The underlying idea was that
strengthening the innovative capacity of
agriculture would have positive spill-over
 D
o
w
nl
oa
de
d
fro
m
htt
ps
://
ac
ad
e
mi
c.
ou
p.
co
m/
cjr
es
/ar
ticl
e-
ab
str
Figure 1. Specialised towns in Guangdong. ac
Source: Authors’ elaborations on data from Guangdong Provincial ST Development Promotion Association. t/1
2/
3/
economy, including industrial production (Di 40
at the centre of Guangdong industrial growth 1/
Tommaso et al., 2013). In the Spark Plan
since the beginning of China’s open-door 55
zones, firms could benefit from subsidies and 81
policy (Figure 2(a)) (Zheng et al., 2016).
loans (Barbieri et al., 2010). Some DSTGG 69
The concentration of STs in the PRD re- 7
experts realised that most of them tended to
flected policymakers’ effort to rationalise this by
specialise in specific sectors—a process now U
area’s fast—and often chaotic—growth (Chen,
known as “one city, one product” (Su and ni
2007; Enright et al., 2005). Such economic ex- ve
Sun, 2016; Wang and Yue, 2010), which rests
pansion must be related with the career incen- rsi
at the foun- dation of the ST programme.
tives given by national rules to local officials. da
From 2000 to 2003, the second phase—a Indeed, up to the beginning of the 2000s, local d
pilot development of the programme—took officials were mainly promoted according to
E
A
place, introduced by two formal documents: their ability to produce economic improve- N
“Plan for the ST Technological Innovation: ments (mainly GDP growth) at the local level us
Pilot Test” and “Managerial Methods for (Li, 2014). To this aim, local officials often used
the ST Technological Innovation: Pilot Test” transactions in the land market, which indeed
(DSTGG, 2008). Most of the STs recognised boomed between 1988 and 1992. Land con-
in this phase are in the Pearl River Delta veyance fees generated by such transactions
(PRD),
 D
o
w
nl
oa
de
d
fro
m
htt
ps
://
ac
ad
e
mi
c.
ou
p.
co
m/
cjr
es
/ar
ticl
e-
ab
str
ac
t/1
2/
3/
40
Figure 2. Geographical distribution of specialised towns. 1/
Source: Authors’ elaborations on data from Guangdong Provincial ST Development Promotion Association. 55
81
69
were used for infrastructural financing and 7
to encourage development of lagging areas by
upgrading, ending in GDP growth (Ahmad (Barbieri et al., 2012; Sarcina et al., 2014). U
et al., 2018; Cao and Zhao, 2011). Hence, this ni
The new STs were established in the so-called
phase of the STs programme served the pro- ve
Great PRD and on the coast and peripheral rsi
motion of industrial growth and upgrading, areas through 2006 (Figure 2(b)), with rural da
while rationalising industrial development bases and mountainous territory through 2009 d
(Barbieri et al., 2009). The sectoral specialisa- E
(Figure 2(c)). They had different productive
tion of these first towns reflected the features of A
peculi- arities, such as being focussed on N
Guangdong’s—and the PRD’s—manufacturing agriculture- related, resource-based activities us
systems at the time, mainly based on low-tech or services (Figure 3).
production (Figure 3).
Some official documents12 highlighted the
In the third phase (2003–2008), the STs double path behind the policy implementa-
pro- gramme extended to some non-PRD tion of this phase and recognised the strategic
cities and, along with upgrading, started to value of STs in Guangdong’s development
be used
 100%
5
90% 26 56
34
80% 64
27 43
D
70% 69
o
34
60% w
96
nl
50% 39 19 oa
40% 13 de
17 46 d
fro
30%
62
m
6
20% 102 htt
46 ps
10% 9 ://
ac
8 17 29 ad
2
0%
2003 2006 2009 2015 e
mi
Services Agriculture High-Tech Medium-Tech Low-Tech Resource Based c.
ou
Figure 3. Sectoral distribution of specialised towns.
p.
Source: Authors’ elaborations on data from Guangdong Provincial ST Development Promotion Association.
co
m/
(Barbieri et al., 2010; Lin, 2006; Su and Sun, cjr
1. to further decentralise the economic man- es
2016). They stressed how provincial and local agement of STs to local authorities; /ar
institutions started to collaborate in clus-
2. to optimise the industrial and urban struc- ticl
tering processes to avoid sectoral overlap and e-
ture of STs;
to coordinate the regional innovation system. ab
3. to restore public accounts and realise infra- str
This is reflected also in the establishment of
structural investments; ac
the Guangdong Provincial ST Development t/1
4. to promote quality upgrading of traditional
Promotion Association (POTIC) (Su and Sun, 2/
industries and 3/
2016). Furthermore, many of these
5. to increase GDP, innovation potential and, 40
documents pointed to the diffusion of good
more generally, competitive capacity (Tsai, 1/
practices re- lated to knowledge diffusion and 55
2013).
protection and territorial branding. Finally, 81
69
they emphasised collaboration between local
Similar to this is the “One Town, One 7
actors and univer- sities and research by
Institute” initiative (2012), aimed at founding
institutions. U
a third- level degree institution in each ni
Since 2008, a fourth phase has begun, with township (Su and Sun, 2016). Finally, the ve
a new emphasis on endogenous growth and rsi
initiatives issued by the provincial agencies
innovation, in line with the whole change in da
stress two elem- ents to empower endogenous d
post-crisis Chinese policy orientation. The
sources: (i) pro- moting micro, small and E
most important measure in this framework is medium enterprises (SME) and establishing A
the “One Town, One Policy” (2010), N
dedicated public innovation services, and (ii)
incentiv- ising each town to find its own us
enhancing the quality and innovative
developmental path while supporting the
performances of those towns specialised in
coordinated eco- nomic development of locally rooted productions (Di Tommaso et
Guangdong Province (DSTC, 2011a, 2011b).
al., 2013; DSTGG, 2017; Su and Sun, 2016).
This implies the fol- lowing actions:
 Nowadays, almost all Guangdong coun- To identify the two groups, we referred to
ties host at least one ST, and while a high Becattini’s (2015) contribution about
number is still in the PRD (particularly in industrial districts as originated by the
Dongguan), many others are also localised in historical profile of the production D
the north-eastern peripheral prefectures of o
atmosphere of the hosting areas. We
Meizhou and Chaozhou (Figure 2(d)). While w
classified each township as endogenous if its nl
keeping a good proportion of agriculture- specialisation had at least one of these oa
related STs, the sectoral distribution of towns features:13 de
has changed towards low-, medium- and d
high-tech specialisation (Figure 3). fro
1. rooted in the history of the town; m
The programme has inspired a 2-fold sec-
2. based on local resources (for example, htt
toral specialisation of the townships: on the ps
aqua- culture in townships next to water);
one hand, the specialisation in resource- ://
3. traditional, typical or mainly oriented to- ac
based and agriculture-related activities,
wards national markets (for example, rice ad
consistent with the policy’s aim to upgrade e
cookers or pottery);
trad- itional production in areas less involved mi
4. derived from a production rooted in town’s c.
in manufacturing development; on the other,
history (for example, machinery for stones ou
the growth of mid- to high-tech
processing in areas previously used as p.
manufacturing to promote technological co
quar- ries) and
upgrading and innova- tive activities in core m/
5. linked to the territory’s physical character- cjr
manufacturing clusters in Guangdong.
istics (for example, port logistics on es
coastal areas). /ar
ticl
Endogenous and exogenous e-
specialised towns Some specific productions [such as ab
Information and Communication str
Notwithstanding the common policy frame- Technologies (ICT)] are considered per se ac
work, STs are heterogeneous for size, urban- t/1
exogenous, as they stemmed from the activity 2/
isation degree, development period, etc. An of foreign actors. Conversely, it is likely that 3/
in-depth analysis of the townships shows that 40
some activities, naturally oriented to serve the
some of them, which we call endogenous, fol- local or national market (as the agriculture
1/
lowed an industrial development rooted in 55
related), are automatically endogenous. 81
their own economic and cultural history; in Nonetheless, the distinction be- tween these 69
some others, which we call exogenous, 7
two groups only partially overlaps with
external forces such as foreign capital and by
sectoral distribution (Table 1). U
national pol- icies triggered their growth.
Endogenous STs are 209 out of the total ni
399 and have grown since the mid-2000s in ve
rsi
cor- respondence with the shift of policy da
towards d
E
A
N
Table 1. Sectoral distribution of endogenous versus exogenous specialised towns. us

Sectors

Agriculture-related High-tech Low-tech Mid-tech Resource-based Services Total

Endogenous 102 3 37 8 41 18 209

Exogenous 0 43 32 88 15 12 190
Totals 102 46 69 96 56 30 399

Source: Authors’ elaborations.

 45 PRD Non-PRD
40
35
30 D
o
25 w
20 nl
oa
15 de
10 d
fro
5 m
0 htt
ps

Chaozhou Heyuan
Dongguan Foshan Guangzhou Huizhou (PRD)

://
ac
ad
e
mi
c.
ou
p.
Exogenous STs Endogenous STs co
m/
Figure 4. Geographical distribution of endogenous and exogenous specialised towns, year 2015.
cjr
Source: Authors’ elaboration.
es
/ar
endogenous sources (Supplementary Figure ticl
2. RQ2: Can the policy contribute to e-
A1). Their geographical distribution confirms territorial rebalancing within the province? ab
different specialisation paths between PRD str
and non-PRD (Figure 4). Endogenous town- ac
We use data at the most detailed territorial t/1
ships are mainly concentrated in non-PRD
level—that is, counties and districts— 2/
inner and mountainous areas, testifying to a
which is yet above the township one. We 3/
larger policymaking investment to promote 40
mainly use panel fixed-effects models to 1/
these prefectures by focussing on the local
test for robust correlations between the 55
market and existing productions. The predom-
policy and local economic performance. 81
inance of exogenous specialisation in the 69
Given the structure of the data and
PRD, conversely, clearly indicates its 7
considering that policy implementation by
exogenous eco- nomic growth path, based on
continuously changes through time and U
foreign capital and national migrations, ni
involves the great majority of Guangdong
although with some notable exceptions that ve
counties, there are obstacles in building rsi
founded their eco- nomic growth on
proper counterfactuals to explore da
traditional productions.
causality. 14 However, to address some causal d
interpretation, we add a ro- bustness check E
A
Empirical analysis based on latest techniques to estimate dose- N
response effects, particu- larly fit for us
We now turn to investigate the two empirical analysing non-binary treatments as in our
research questions: case. We use the Cerulli (2015)15 fixed-
effect dose-response method, which,
1. RQ1: Is the ST programme associ- compared with usual General Propensity
ated with improved territorial economic Scores methods (Bia and Mattei, 2008;
performances? Hirano and Imbens, 2004), allows to
include
non-treated units in the analysis and to con- 3 years.20 Summary statistics for the relevant
sider increasing treatment intensity across variables are in Supplementary Table A1.
time. Our outcome variables are industrial To analyse the association between STs
and agricultural output in county i at time t and territorial rebalancing (RQ2), we divide D
(INDOUTi,t and AGROUTi,t) used as a proxy the counties into four groups (low, medium o
of economic performance.16 low, medium high and high) according to the w
We test the association of clusters with eco- nl
level of industrial output in 2000 (t0) using oa
nomic performances at the local level (RQ1) de
the
first with panel data fixed effects with robust distribution’s quartiles as thresholds. We ob- d
standard errors.17 Our independent variables of fro
serve through a fixed-effects panel subgroup m
interest are: analysis whether the effects of SP_TOWN and htt
of RATIO_ENDOG is heterogeneous ps
1. the number of STs in county i (SP_TOWNi) ://
according to the initial level of local output. ac
and Specifically, if we find that the effects are ad
2. the ratio of endogenous towns on the total stronger on lower than on higher groups, we e
number of STs (RATIO_ENDOGi). may conclude that STs help territorial mi
c.
rebalancing. We also sup- port these results ou
Finally, we introduce the number of full-time with the evidence found in relation to RQ1. p.
workers in county i (WORKERSi), to control We summarise information on the subgroups co
for the economic activity, and year dummies. in Supplementary Table A2.
m/
cjr
The variables enter the model with one or es
three lags to account for lagged effects. We /ar
also perform a subgroup analysis (Boyd- The effects of specialised towns and ticl
Swan et al. 2016) to study possible endogenous specialised towns on e-
ab
heterogeneous effect between PRD and non- economic performances (RQ1) str
PRD areas. Table 2 shows the results of the fixed-effects ac
We then support the results with the dose- t/1
es- timation. For industrial performances for 2/
response analysis,18 including the variable of the whole sample, there is a positive 3/
interest: association between the ratio of endogenous 40
1/
towns, at t-1, and industrial output. 55
1. as a dummy variable (value = 0 if the unit Additionally, in the PRD, endogenous towns 81
is not treated and 1 otherwise) and show positive effects at t-3. For the whole 69
2. as intensity of the treatment—the dose— sample, there is also quite robust evidence 7
normalised in order to range from 0 by
that agricultural output benefits from higher U
(minimum dose in the sample) to 100 levels of endogenous STs 1-year lagged. The ni
(max- imum dose in the sample). gains related to endogenous towns, in this ve
rsi
case, appear to be stronger in non-PRD coun- da
The latter is assumed to have a parametric ties. In all such cases, there is no observable d
polynomial form (Cerulli, 2015; Filippetti ef- fect of specialisation per se. E
and Cerulli, 2018) that can be investigated Figures 5 and 6 show graphically the results A
N
through graphical representation. We run two of the dose-response models (the coefficients us
series of models, one for specialized towns are in Supplementary Tables A3 and A4). The
(SP_TOWNi) and the other for endogenous first focuses on STs total number. Regarding
towns (ENDOGi).19 The outcomes of interest in- dustrial output, both at t-1 and t-3, the
are industrial output (INDOUT) and agricul- response is (rather) flat or non-significant up
tural output (AGROUT), forwarded by 1 and to a certain dose of the treatment, and then it
increases. The curves shapes for agricultural
output are similar, although flatter in
correspondence with