Economic Complexity of the City Cluster in Guangdong–Hong Kong–Macao Greater Bay Area, China

Abstract

With the rapid economic growth in China over the last two decades, exploring the changes in the Chinese economy has attracted great attention from the research community. Among different economic clusters in China, the southern region represents the wealthiest region. Hence, it is essential to conduct an in-depth analysis to explore the region’s sustainability in its economy. This paper applies the economic complexity model to 22 major cities within the Guangdong–Hong Kong–Macao Greater Bay Area cluster. The study is based on seven industrial sectors. Revealed comparative advantage of different product sectors, similarities of product sector specialisation, diversity of the economic composition, and the association to the geographical location are investigated in this paper.

1. Introduction

City clusters are formed as a result of urban development, with regional economies combined to scale. Cities surrounding San Francisco, New York, and Tokyo are seminal examples of city clusters in the world. In practice, over 60% of the world’s economic activities take place in harbour cities, with 75% of large cities and 70% of capitals and populations situated within 100 km from the coastline.

China, being the fastest growing economy over the last two decades, possesses multiple city clusters that attract domestic and foreign investments [1]. In Ren’s work on Urban China, in-depth insight on the formation of this mega-economy is covered, from aspects of governance, landscape, migration, inequality, and cultural economy [2]. Similar to most economies, Chinese city clusters are mostly geographically based to form a producer serving network [3]. Cities within close geographical proximity, such as cities in the Pearl River Delta (PRD) region, have attracted great attention from the research community [4]. PRD, the most representative emerging city cluster in Southern China following the strategic economic reform in 1993, is the largest mega-city region in the world. A study between 1993 and 2012 in the PRD has shown that emerging clusters have negative effects on other co-located industries’ or clusters’ total factor productivity, while mature clusters have positive effects [4]. Thus, it is of great interest to further explore the industrial composition and how that affects the economy over time, and this paper discusses this matter from the economic complexity perspective.

City clusters may be formed by multiple factors, such as geographical proximity and the accessibility through transport infrastructure [5,6,7,8,9], thus spatial interaction among cities leads to a positive influence on urban growth [10]. Although, urbanisation through city clusters also brings in other challenges, such as environmental impacts [10,11]. At the same time, urban development, housing policies, personal status, and family relationships lead to distinctive spatial patterns of ageing differentiation [12].

To evaluate the economic development within city clusters, export records are frequently used as a good estimate of economic productivity. Guangdong’s government provides city-level GDP records from different industrial sectors; these values are used to replace the export value by analysing the attributes of major Chinese cities in the PRD region and discussing their economic complexity. This paper calculates the revealed comparative advantage of each industrial sector of the PRD cities based on the city-level GDP figure and subsequently evaluates ubiquity and diversity of the PRD cities to reflect the city-level economic complexity.

The economic complexity model [13,14], originally introduced by Harvard University, has found applications in profiling the competences of economies at the country level. Economic complexity builds on top of the analysis of diversity and ubiquity of an economy and obtains revealed comparative advantage in different industrial sectors to reflect the knowledge barrier of an industrial sector for an economy to possess. Of the two complementary measures of the knowledge capital, ubiquity measures how many products are produced by an economy and ubiquity measures how many economies export a product. The analysis includes normalisation such that small economies or industrial sectors are fairly considered. Such a model has been applied at the subnational level in Australia [15], Brazil [16], and Mexico [17]. These studies explore domestic trading patterns that are significantly different to international trades, such as service sectors and perishable foods. In addition to the application in economic analysis, the complexity model has been applied to analyse academic output among research institutes [18] and to predict the evolution of the research output by constructing the research space [19].

China is the second largest economy in terms of GDP. Although, according to the economic complexity analysis, the composition of China’s product sectors ranked 33 in the year range 2013–2017, according to the Observatory of Economic Complexity (https://atlas.media.mit.edu/en/), reflecting that its product sectors exhibit relative knowledge and skill deficiency. Although, such an impression is rapidly changing, especially with the recent boost in entrepreneurship in China, subject to the effect of localisation and urbanisation economies as discussed [20]. While the localisation and urbanisation economies form city clusters, there’s an imminent interest in exploring the economic complexity at the intra city-cluster level. Such a study will offer an insight into the regional economy to assist policy makers.

The rest of this paper is organised as follows: An overview of the technical background of revealed symmetric comparative advantage is presented first, which leads to the modelling of economic complexity. Collection and pre-processing of the data used in this study is explained, and the result and the analysis are covered. Subsequently, we discuss the limitations of this study, followed by the concluding remarks.

1.1. Revealed Symmetric Comparative Advantage

This paper investigates complexity modelling of major cities around Pearl River in Southern China, based on the GDP records of industrial sectors [13]. Let c denote a city, and p denote an industrial sector (or product). Let X_cp denote the GDP value of industrial sector p of city c, and the revealed comparative advantage (RCA) [21] can be obtained to reflect the degree of specialisation in an industry. Mathematically, RCA_cp for city c and product p can be expressed as:

$$RC{A}_{cp}=\frac{X_{cp}/{\sum }_p{X}_{cp}}{{\sum }_c{X}_{cp}/{\sum }_{c,p}{X}_{cp}}.$$

(1)

Normalisation of $RC{A}_{cj}$ makes a smaller scale city with fewer export values adequately assessed; likewise, a less popular industrial sector can be fairly compared.

According to the economic complexity model [14], a city is known to have revealed comparative advantage for a product sector if its RCA value is equal to or larger than one. This approach faces a limitation that the ratio relative to the threshold is inconsistent. For example, when RCA is halved or doubled from reference point 1, the two operations yield different levels of loss/gain (i.e., 0.5 and 2), making it difficult to interpret the relative position against the reference point. Thus, there is a need for symmetry in RCA, such that the relative change from the reference point is comparable. The Revealed Symmetric Comparative Advantage (RSCA), proposed by Laursen [22], offers symmetry in RSCA values centred closed to zero:

$$\begin{array}{cc}\hfill RSC{A}_{cp}=& \frac{RCA-1}{RCA+1}\hfill \\ \hfill =& \frac{X_{cp}{\sum }_{c,p}{X}_{cp}-{\sum }_c{X}_{cp}{\sum }_p{X}_{cp}}{X_{cp}{\sum }_{c,p}{X}_{cp}+{\sum }_c{X}_{cp}{\sum }_p{X}_{cp}}.\hfill \end{array}$$

(2)

The RSCA values across different disciplines are used to construct the $M_{cp}$ matrix, which are either 0’s or 1’s, to indicate whether a city is considered to possess RCA for an industrial sector:

$$M_{cp}=\{\begin{array}{ll}1& \mathrm{if}RSC{A}_{cp}\ge 0,\\ 0& \mathrm{if}RSC{A}_{cp}<0.\end{array}$$

(3)

1.2. Economic Complexity

Based on $M_{cp}$, which indicates whether a city possesses revealed symmetric comparative advantage for an industrial sector, diversity $D_c$ presents the degree of multi-specialisation in a city, which is defined as the sum of all industrial sectors in which the city possesses RCA:

$$D_c=k_{c,0}={\displaystyle \sum }_p{M}_{cp}.$$

(4)

Ubiquity of an industrial sector, $U_p$, is defined as the number of cities that have RCA in that industrial sector:

$$U_p=k_{p,0}={\displaystyle \sum }_c{M}_{cp}.$$

(5)

By applying the Method of Reflections, the value of $k_{c,n}$ can be expressed in terms of $k_{p,n-1}$:

$$k_{c,n}=\frac{1}{k_{c,0}}{\displaystyle \sum }_p{M}_{cp}{k}_{p,n-1}.$$

(6)

Similarly, the value of $k_{p,n}$ can also be expressed in terms of $k_{c,n-1}$:

$$k_{p,n}=\frac{1}{k_{p,0}}{\displaystyle \sum }_c{M}_{cp}{k}_{c,n-1}.$$

(7)

The economic complexity of a city is evaluated as the normalised diversification, according to Hidalgo [13], where each diversification value is normalised by subtracting the mean and then dividing by the standard deviation:

$$k_{c,n}^{\prime }=\frac{k_{c,n}-\overline{k_{c,n}}}{\sigma \left(K_c\right)},$$

(8)

where $\overline{k_{c,n}}$ denotes the average of $k_{c,n}\forall c$ with a given $n$, and $\sigma$ denotes the standard deviation function.

Subsequently, each city may be characterised by the vector $k_c=\left\{k_{c,n}^{\prime }\mid n=1\dots N\right\}$ and $k_p=\left\{k_{p,n}^{\prime }\mid n=1\dots N\right\}$, where $N$ is a predetermined number of iterations and $N\in \mathbb{Z}$ and $N>1$. The diversity and ubiquity values converge after several rounds of iteration, and when $N=18$ the diversity value is chosen as the measure of economic complexity, according to Hidalgo [13].

2. Data Collection

In this study, we investigate the city cluster around the Pearl River Delta region in the southern part of China and analyse its GDP-based economic complexity in between 2000 and 2015, covering 7 product sectors in

Table 1. Classification of industrial sectors. (Data source: Wind database and Hong Kong government statistics website).

Classification	Industrial Sector
c1	agriculture/forestry/livestock/fisheries
c2	industrial
c3	building and construction
c4	wholesale and retail
c5	transport and logistics
c6	finance
c7	real estate

. In comparison to the study on the producer service linkage of 9 PRD cities [23], or the study on functional and sectoral divisions of labour between Hong Kong, 2 PRD cities, and 2 PRD regions [24], or 9 PRD municipalities for the study of knowledge-intensive business services [25], this study investigates the economic complexity of 21 PRD cities plus Hong Kong, offering more fine-grained analysis of the region and exploring more insights of economic interactions and sustainability.

The GDP record among the PRD cities were collected between 2000 and 2015. To give an overview of the GDP data, records from 2015 are included in

Table 2. City-level GDP data in the PRD region in 2015.

City	Abbrev	c1	c2	c3	c4	c5	c6	c7
Guangzhou	GU	226.84	5185.63	551.17	3099.92	1255.19	1628.71	1529.42
Shenzhen	SZ	6.65	6742.98	479.72	2361.16	540.8	2501.57	1564.41
Shantou	ST	96.71	877.22	86.56	341.47	42.4	50.95	82.92
Foushan	FS	136.45	4675.14	166.36	643.29	270.47	341.73	628.94
Shaoguan	ZK	151.68	360.33	71.64	146.14	84.13	50.69	55.69
Heyuan	HU	94.01	334.67	35.85	108.04	22.59	43.89	57.31
Meizhou	MJ	188.49	287.99	65	108.71	25.44	42.54	52.28
Huizhou	HJ	151.54	1624.48	102.43	411.41	80.71	121.04	213.79
Shanwei	SW	118.04	320.24	28.7	93.15	20.93	20.6	51.77
Dongguan	TG	21.03	2840.35	88.81	905.37	205.9	401.37	529.25
Zhongshan	ZS	66.48	1566.89	66.07	328.61	72.91	159.97	184.78
Jiangmen	GM	174.5	1023.06	62.02	222.23	85.91	126.31	128.99
Yangjian	EG	205.33	509.38	55	129.97	81.26	34.19	73.62
Zhanjiang	JG	454.67	796.77	115.88	244.34	118.97	76.82	122.85
Maoming	MM	387.41	901.03	99.71	293.41	80.14	62.5	122.4
Yuqing	CG	288.29	930.28	61.17	199.15	56.77	54.91	53.25
Qingyuan	FM	192.52	437.46	48.48	129.91	85.9	61.08	82.06
Chaozhou	T5	64.27	454.05	30.23	107.6	22.97	40.02	39.64
Jieyang	RU	167.69	1062.01	69.2	342.02	19.7	25.14	41.03
Yunfu	ZM	149.11	265.83	37.9	63.07	22.78	35.38	34.12
Zhuhai	5C	45.11	894.07	119.95	249.24	46.51	146.8	160.32
Hong Kong	HK	18.85	508.83	885.74	1495.18	1224.95	3316.8	869.9

. It should be noted, among all GDP data missing records are observed, including Hong Kong’s c4 (wholesale and retail) in 2000, and Hong Kong’s c4 (wholesale and retail) and c7 (real estate) in 2015. While these are the first and the final year of the collected data, the GDP record of the matching product sector in adjacent years (2001 and 2014, respectively) are taken as the estimate of the missing entries.

Economic Complexity in the Pearl River Delta Region

The RSCA distribution across different cities is illustrated using a heat map as shown in Figure 1. Across different product sectors, it is found that most cities exhibit strong revealed comparative advantage in c1 (agriculture/forestry/livestock/fisheries), whereas Hong Kong, Shenzhen, and Dongguan significantly fall behind in this sector. This observation shows that major cities have limited space, thus resulting in high-cost operations for sustaining an agricultural industry. On the contrary, for the c6 (finance sector), Hong Kong is the only city among the list with strong revealed comparative advantage, whereas other cities have picked up in this sector over time.

Among the PRD cities, Hong Kong retains its leadership in diversity, with revealed comparative advantage in 5 out of 7 product sectors in all 4 sample years, which is ahead of other cities. The distribution pattern is also unique to most PRD cities other than Guangzhou. The finding supports that Hong Kong’s functional and sectoral specialisation reveals complementarities over neighbouring cities, whereas Guangzhou (and to a lesser extent Shenzhen) represents competition to Hong Kong, according to the study by Schiller et al. [24]. Hong Kong’s strength in c6 (finance sector) has consistently yielded high revealed comparative advantage over PRD cities. While this sector has picked up among all PRD cities over time, it demonstrates a possible linkage that Hong Kong’s leadership helps to promote development in the finance sector among the PRD cities.

In general, Figure 1 shows that PRD cities are moving away from c1 (agriculture, forestry, livestock, fisheries), while shifting the focus towards c6 (finance), c7 (real estate), and c3 (building and construction). Such findings suggest that during the rapid-growth phase of PRD cities, the agricultural industry is the least economically sustainable sector, whereas finance, real estate, and building and construction are the most economically sustainable sectors. Such a finding aligns with Ren’s study of the rural-to-urban transition, yet the rural–urban dichotomy may need further adjustment for the evolving economy in China. [2] This is supported in our study that the role of agriculture-focused cities has been evolving with reduced emphasis over time but still holds a significant role amid industrialisation and urbanisation. The RSCA shift could be influenced by both governmental policies as well as private investments. The findings can be utilised by policy makers of future emerging economies to learn from the experience in PRD city clusters, to position their strategic industrial focus over time. In addition, subject to the availability of similar datasets from other emerging city clusters, it would be an interesting future study to explore the differences in industrial shift, offering further advice to policy makers about how the strategic industrial focus could be fine-tuned.

The similarity of RSCA distribution across different product sectors among PRD cities can be found in Figure 2. Euclidean distance of the M_cp values between each city pair is calculated, and links are generated if the distance value falls below a given threshold. Across different years, it has been found that city pairs sharing similar distribution patterns exhibit a slight decreasing trend, with 24 links in 2000, 17 links in 2005, 14 links in 2010, and 16 links in 2015. The linked city groups, or the clusters formed by these links, include 5 clusters in 2000, 7 clusters in 2005, 5 clusters in 2010, and 4 clusters in 2015. There is no clear pattern showing whether the clusters are expanding or diminishing.

Similarities across different time varies, with no common links across all four subplots in Figure 2. The observation reflects that the M_cp distribution of each city varies, and they may be grouped in different clusters at different times. In between 2000 and 2005, 4 common links are found (Heyuan–Meizhou, Qingyuan–Heyuan, Qingyuan–Meizhou, Yangjiang–Shanwei); in between 2005 and 2010, 3 common links are found (Zhongshan–Foshan, Yunfu–Meizhou, Yangjiang–Shanwei); in between 2010 and 2015, 5 common links are found (Zhongshan–Foshan, Yunfu–Meizhou, Jiangmen–Huizhou, Chaozhou–Huizhou, Chaozhou–Jiangmen). Among these common links, 3 links have appeared twice (Yangjiang–Shanwei, Yunfu–Meizhou, Zhongshan–Foshan), and the findings demonstrate that these cities share closer similarities in terms of M_cp distribution over time.

Figure 3 shows the geographical distribution of economic complexity of the city cluster. Hong Kong—together with two super-cities, Shenzhen and Guangzhou and their surrounding cities—have formed a cluster that exhibits a high concentration in combined economic complexity. Another cluster is formed by Shantou and Chaozhou on the eastern region of the Pearl River Delta. The remaining cities either possess low economic complexity or situated far apart, thus mostly appear as individual nodes or with weak links to the adjacent cities.

The overall economic complexity, measured in terms of normalised diversification as shown in Equation (8), is summarised in

Table 3. Economic complexity as the normalised diversification.

City	2000	2001	2002	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015
GU	2.09	1.49	1.27	1.39	1.93	1.41	1.97	1.95	2.32	1.98	1.58	1.19	1.67	2.07	2.09	2.08
SZ	−0.08	0.04	0.04	1.30	1.28	2.01	1.86	1.88	2.18	2.19	−0.50	0.17	2.05	1.83	1.91	2.08
ST	−0.26	−0.37	−0.41	−0.53	−0.19	−0.34	−0.39	−0.39	−0.37	−0.40	0.11	0.03	−0.12	−0.07	−0.02	0.02
FS	−0.01	−0.34	−0.47	−0.14	0.07	0.38	0.09	0.14	−0.07	−0.58	−1.67	−1.79	−1.08	−0.95	−0.72	−0.39
ZK	−0.36	−0.31	−0.32	−0.26	−0.67	−0.48	−0.57	−0.57	−0.56	0.21	1.13	1.04	0.53	0.39	0.06	−0.10
HU	−0.54	−0.30	−0.25	−0.32	−1.05	−0.91	−0.57	−0.57	−0.56	−0.72	−0.85	−0.87	−0.99	−0.58	−0.67	−0.64
MJ	−0.54	0.91	−0.25	−0.32	−1.05	−0.91	−0.90	−0.92	−0.81	−0.78	0.57	0.64	−0.18	−0.40	−0.64	−0.76
HJ	−0.47	−0.67	−0.75	−0.71	−0.67	−0.48	−0.51	−0.48	−0.51	−0.85	−0.85	−0.87	−0.99	−0.96	−0.90	−0.76
SW	−0.54	−0.30	−0.07	−0.27	−0.50	−0.59	−0.54	−0.56	−0.47	−0.35	0.70	0.63	−0.12	−0.16	−0.90	−0.76
TG	2.09	1.49	1.77	1.63	1.28	2.01	1.86	1.88	1.60	1.05	0.41	−0.03	1.04	0.92	1.37	1.54
ZS	−0.26	−0.34	−0.47	−0.14	0.07	0.38	0.09	0.14	−0.07	−0.58	−1.67	−1.79	−1.08	−0.95	−0.72	−0.39
GM	−0.26	−0.37	−0.41	−0.53	−0.19	−0.42	−0.51	−0.48	−0.51	−0.85	−0.85	−0.87	−0.99	−0.96	−0.90	−0.76
EG	−0.54	−0.30	1.31	0.91	0.13	−0.59	−0.54	−0.56	−0.47	−0.35	0.70	0.63	−0.18	−0.58	−0.67	−0.17
JG	−0.38	−0.39	−0.38	−0.72	−0.50	−0.42	−0.51	−0.57	−0.81	0.21	1.13	1.04	0.53	0.39	0.06	−0.76
MM	−0.26	−0.39	−0.38	−0.72	0.61	−0.59	−0.54	−0.56	−0.47	−0.29	0.47	0.34	0.04	0.24	0.43	−0.64
CG	−0.47	−1.00	−1.04	−0.72	−0.32	−0.59	−0.54	−0.56	−0.47	−0.35	0.57	0.64	−0.99	−0.96	−0.90	−0.76
FM	−0.54	−0.30	−0.25	−0.32	−1.05	−0.91	−0.57	−0.57	−0.56	−0.72	−0.18	1.04	0.53	0.39	0.06	−0.10
T5	−0.47	−0.67	−0.75	−0.71	−0.58	−0.26	−0.28	−0.27	−0.27	−0.85	−0.85	−0.87	−0.99	−0.96	−0.90	−0.76
RU	−0.26	−0.37	−0.41	−0.53	−0.19	−0.26	−0.39	−0.39	−0.27	−0.39	−0.24	−0.37	−0.34	−0.96	0.04	0.16
ZM	−0.93	−1.00	−1.04	−1.29	−1.05	−0.91	−0.90	−0.92	−0.81	−0.78	0.57	0.64	−0.18	−0.40	−0.64	−0.76
5C	0.00	0.10	0.12	0.11	−0.07	0.22	0.13	0.15	0.06	1.31	−1.67	−1.79	−0.27	0.67	0.66	0.75
HK	2.95	3.38	3.12	2.89	2.69	2.21	2.28	2.22	1.93	1.89	1.40	1.20	2.13	2.01	1.92	1.89

. Hong Kong, holding the highest normalised diversification in 2000, has dropped its leading position in economic complexity, with a loss in its normalised diversity of 1.06. This is possibly due to the economic boost in other PRD cities. Hong Kong’s decline in its relative economic complexity is also reflected in Figure 1.

Shenzhen, one of the four super cities in China (Beijing, Shanghai, Shenzhen, Guangzhou), has been under rapid growth in economic complexity, with a gain of 2.16 between 2000 and 2015. On the contrary, the economic complexity of Guangzhou, another super city in China, remains almost unchanged with a slight 0.01 drop.

As shown in

Table 4. Statistical analysis of economic complexity.

City	Mean	Standard	Median	First	Third	Variance	Standard	Kurtosis	Skewness	Range	Mini	Max	Sum
		Error		Quartile	Quartile		Deviation
GU	1.78	0.09	1.94	1.47	2.07	0.12	0.35	−1.25	−0.34	1.13	1.19	2.32	28.48
SZ	1.27	0.24	1.85	0.14	2.02	0.94	0.97	−1.23	−0.75	2.69	−0.50	2.19	20.24
ST	−0.23	0.05	−0.30	−0.39	−0.06	0.04	0.20	−1.28	0.36	0.64	−0.53	0.11	−3.70
FS	−0.47	0.16	−0.37	−0.78	0.01	0.40	0.64	0.09	−0.89	2.17	−1.79	0.38	−7.53
ZK	−0.05	0.14	−0.29	−0.50	0.26	0.32	0.57	0.06	1.02	1.80	−0.67	1.13	−0.84
HU	−0.65	0.06	−0.61	−0.86	−0.56	0.06	0.24	−0.71	0.03	0.80	−1.05	−0.25	−10.39
MJ	−0.40	0.15	−0.59	−0.83	−0.23	0.37	0.61	0.24	1.15	1.96	−1.05	0.91	−6.34
HJ	−0.71	0.05	−0.73	−0.86	−0.51	0.03	0.18	−1.40	0.08	0.52	−0.99	−0.47	−11.43
SW	−0.30	0.11	−0.41	−0.55	−0.15	0.19	0.44	1.55	1.29	1.60	−0.90	0.70	−4.80
TG	1.37	0.15	1.52	1.05	1.79	0.34	0.58	0.98	−1.09	2.12	−0.03	2.09	21.91
ZS	−0.49	0.16	−0.37	−0.78	−0.04	0.39	0.63	0.16	−0.87	2.17	−1.79	0.38	−7.78
GM	−0.62	0.07	−0.52	−0.86	−0.42	0.07	0.26	−1.45	0.00	0.80	−0.99	−0.19	−9.86
EG	−0.08	0.16	−0.33	−0.55	0.26	0.39	0.63	0.02	1.13	1.98	−0.67	1.31	−1.27
JG	−0.13	0.16	−0.39	−0.53	0.26	0.38	0.62	−0.17	0.97	1.94	−0.81	1.13	−2.08
MM	−0.17	0.11	−0.34	−0.55	0.27	0.20	0.45	−1.31	0.55	1.33	−0.72	0.61	−2.71
CG	−0.53	0.13	−0.58	−0.92	−0.44	0.25	0.50	1.88	1.48	1.68	−1.04	0.64	−8.46
FM	−0.25	0.14	−0.31	−0.57	−0.06	0.30	0.55	0.76	0.90	2.09	−1.05	1.04	−4.05
T5	−0.65	0.07	−0.73	−0.86	−0.42	0.07	0.26	−1.28	0.50	0.73	−0.99	−0.26	−10.44
RU	−0.32	0.06	−0.36	−0.39	−0.26	0.06	0.24	3.29	−0.57	1.12	−0.96	0.16	−5.17
ZM	−0.65	0.14	−0.86	−0.95	−0.58	0.31	0.55	1.86	1.59	1.93	−1.29	0.64	−10.40
5C	0.03	0.20	0.12	−0.02	0.33	0.62	0.79	2.09	−1.24	3.10	−1.79	1.31	0.48
HK	2.26	0.15	2.17	1.91	2.74	0.37	0.61	−0.43	0.25	2.18	1.20	3.38	36.11

, by averaging the normalised diversity throughout 16 years, Hong Kong has the highest average economic complexity, followed by Guangzhou, Dongguan, Shenzhen, and Zhuhai. The cities with the lowest average economic complexity are Huizhou, Chaozhou, Yunfu, Heyuan, and Jiangmen. In terms of volatility in economic complexity across 16 years, Shenzhen, Zhuhai, Foshan, Zhongshan, and Yangjiang have the highest standard deviation, whereas Huizhou, Shantou, Heyuan, Jieyang, and Jiangmen have the lowest standard deviation.

3. Discussion

It should be noted that Macau, a major city in Southern China within close geographical proximity, is excluded in this study. While Macau possesses a high GDP, its industrial composition significantly differs to the PRD city cluster, hence it is not typically considered as the same industrial ecosystem. However, it is unclear how cross-sectors influence the nearby economies, and further exploration may be of great interest.

The analysis conducted in this paper should be taken with caution, as the small number of industrial sectors provide a very high-level abstraction of economic complexity modelling. The classification with seven product sectors used in this paper can be considered comparable to Tier 1 of the Standard International Trade Classification (SITC), which consists of 10 product sectors. In the study of the Atlas of Economic Complexity [14], a six-digit SITC code with over 1000 industry classifications have been used, which provides a more fine-grained product sector composition compared to this study. Collecting finer details of GDP records will be considered for future work following this study.

Selection of the representative cities in the PRD region is also worth discussion. In comparison with the popular choice of nine PRD cities possessing the highest GDP [23], subsequently 12 PRD cities, according to the GDP together with Hong Kong, have been included in this study. As illustrated in Figure 2, these smaller scale cities may possess unique patterns of industrial composition, representing three, one, three, and three isolated nodes in years 2000, 2005, 2010, 2015, respectively. These are comparable to three, two, three, and three isolated nodes from the list of the nine largest PRD cities, excluding Hong Kong. Such an observation indicates that the economic composition of smaller cities is not a replication of the larger ones. These smaller cities could, however, still be considered subsidiary as the result of industrial transformation in larger cities (for instance, offloading labour-intensive industries to smaller counterparts). Such practices may lead to satellite cities with complementing industrial sectors that demand different skillsets. Further investigations of such behaviour may require data set at finer details (i.e., further break down of industrial sectors), which could be a potential future study, subject to the availability of relevant datasets.

A further limitation is the scale of the cities in the study. China, the country with the highest population in the world, hosts cities in much larger scales. For instance, the population of Shenzhen and Guangzhou combined has exceeded the entire population in Australia. Cities at such a scale could be subdivided, and intracity studies, such as the work investigated by Liu et al. [26] may help to give better insights or provide fair comparisons.

4. Conclusions

This paper examines the economic complexity of the Pearl River Delta cities, using the GDP data between 2000 and 2016. Among 22 major cities in this region, it was found that Hong Kong, Guangzhou, and Dongguan are top-ranked cities in terms of economic complexity, whereas Yunfu, Heyuan, and Jiangmen represent the bottom three. In terms of the variation in economic complexity Shenzhen, Zhuhai, Foshan, Zhongshan, and Yangjiang were found to have the highest standard deviation across this period, showing the highest volatility of structural change in the economy. It was found that the distribution pattern of product specialisation of a city, defined as the number of product sectors in the city with revealed comparative advantage, has been changed over time. City pairs with the closest distribution of product sector specialisation are Yangjiang–Shanwei, Yunfu–Meizhou, and Zhongshan–Foshan. Hong Kong’s leadership in the finance sector correlates to the growth of the surrounding cities, demonstrating a possible linkage of this sector in the PRD city cluster. It was also found that during the rapid growth phase of PRD cities, the agricultural industry was found to be the least economically sustainable sector, whereas finance, real estate, and building and construction were the most economically sustainable sectors.