Spatial Distribution and Location Determinants of High-Tech Firms in Shenzhen, a Chinese National Innovative City

Abstract

The development of high-tech firms is a vital driver for the economic growth of a city but their distribution and location determinants at the intra-urban level are still unclear. We aim to deepen the understanding of location determinants of high-tech firms, so we construct an analytical framework and use GeoDetector to investigate high-tech firms in Shenzhen based on firms and POI open data in 2023. We find that high-tech firms are distributed in a spatial pattern of ‘one core and six clusters’ with high density in the western area despite industrial heterogeneity. Agglomeration economies and amenity-based factors play a significant role in the distribution of high-tech firms. Institutional factors and classical locational factors have more significant effects on the location of high-tech service and manufacturing firms, respectively. This study contributes to the literature on study spatial units, the influence of amenities, and industrial specificities. These findings highlight public policies on industrial park planning, transportation systems, and public services.

1. Introduction

In this age of the globalized knowledge economy, innovation has emerged as a primary driver of national and urban development. High-tech firms contribute to the sustained economic growth of a city by creating jobs and services, paying taxes toward the municipal budget, and increasing the level of human capital [1,2]. Moreover, developing high-tech industries facilitates optimizing the urban spatial and industrial structure because they are knowledge- and technology-intensive. However, high-tech firms do not evenly distribute themselves in a city; they are even more clustered than traditional producer service and manufacturing enterprises [3,4,5]. The high added value and positive externalities that these firms contribute to their specific locations have led to scholars and policymakers focusing on their spatial distribution and location determinants.

The location determinants of industrial activity have always been a core topic in economic geography. Neoclassical location theories, including Weber’s industrial location paradigm, Christaller’s central place theory, and Hotelling’s principle of market competition, explained firm location choice solely using the least cost or demand maximization approach [6]. Agglomeration economies have been regarded as important for both manufacturing and producer service firms. Marshall argued that plants in similar industries can benefit through sharing, matching, and learning mechanisms when they locate close to each other [7,8]. Furthermore, Marshallian specialization and Jacobian diversification have been proven necessary for industries where tacit knowledge and face-to-face communication are crucially important for innovation [5,7]. The institutional environment, culture, norms, and other local non-economic characteristics matter for sophisticated knowledge-intensive firms. Recently, the human capital theory and orthodox location theory argued that the urban living environment, especially urban amenities, has a relationship with local industries because amenities can attract skilled workers who are vital for high-tech firms [4]. All of these theories attempt to explain the location determinants of high-tech firms, but which of them has more explanatory power compared to other theories?

China has always emphasized high-tech industry development and Shenzhen has become one of the national innovation hubs and a global city attracting high-quality talents. In 2023, China was the economy with the most science and technology clusters ranked among the top 100 and Shenzhen–Hong Kong–Guangzhou became the second highest-performing cluster in the global ranking [9]. China has launched a series of policy experiments and Shenzhen has developed a ‘national innovation pilot city’ since the implementation of the ‘Torch Program’. Investigating the location of high-tech enterprises within Shenzhen and clarifying different location determinants can provide a heuristic for other cities in Chinese context. Therefore, the research questions for this study are as follows: (1) What is the spatial distribution pattern of high-tech firms in Shenzhen? (2) What are the roles of economic, institutional, and amenity-based factors in high-tech firm location decisions? (3) Do these factors influencing high-tech firm distribution differ depending on firm characteristics?

To address the above issues, we developed an analytical framework by integrating different theoretical dimensions, then explored the spatial pattern and location determinants of high-tech firms. This study aims to deepen the understanding of location determinants of high-tech firms by filling three gaps. We identify subjects more precisely by using open data from a national certified high-tech enterprises database. We use sub-district as the spatial unit at the intra-urban level to analyze firm distribution. We consider firm characteristics and interactive effects between different factors in detail.

The rest of the paper is organized as follows: In Section 2, we review the literature and construct our analytical framework. Section 3 describes the study area, research data, spatial analysis, and regression methodology. Section 4 is an analysis of the main results on spatial distribution and location determinants. Section 5 discusses government planning, industrial heterogeneity, and the role of amenities. Section 6 ends the paper with conclusions.

2. Literature Review

2.1. Intra-Urban Location Patterns of High-Tech Firms

Because a large volume of firm-level information has become available, studies on the location patterns of high-tech business have begun to focus on individual firms instead of industries and using spatial units smaller than county and municipal scales. Barnes and Hutton (2009) conducted a case study of the emergence of the ‘new economy’ such as computer system design, software publishing, and information services in Vancouver’s inner city [10]. They found that Yaletown had become a favored site for firms and their employees because of its spatial proximity to the central business district (CBD), attractive built environment heritage, policies facilitating land use, and high-quality public and private amenities. Frenkel (2012) analyzed interview data from 117 managers of high-tech firms located in four high-tech parks within the Tel Aviv metropolitan region in Israel and found that high-tech firms in their early stages tend to locate in industrial parks with support services and potential innovation facilities [1]. Although these qualitative studies shed light on the early emergence and location choice of particular high-tech firms and analyzed micro-mechanisms, they did not capture the spatial pattern of high-tech firms within cities owing to data limitations.

Studies quantitatively analyze high-tech firm location at the intra-urban level mainly in Chinese cities [7]. Zhang et al. (2013) revealed that high-tech firms have evident agglomeration characteristics in Beijing and that the hotspot had shifted from the central city to suburban areas [11]. The distribution of information technology and research services in Shanghai is similar to that in Beijing [12]. Li et al. (2015) focused on electronic information service and computer and communication equipment manufacturing firms, and found that their spatial distribution presents a stronger agglomeration pattern than other industries and many of them locate in specialized industrial districts [13].

In western contexts, scholars find different intra-urban location patterns for high-tech firms in different socioeconomic environments. Spencer (2015) investigated ‘science-based’ and ‘creative’ industries within the three largest city-regions in Canada using firm-level data, and found that firms in science-based industries concentrated in low-density, single-use neighborhoods in the suburbs [14]. This is similar to Chinese science and industrial parks in suburbs. Romero De Ávila Serrano (2019) compared the knowledge-intensive business services in six city-regions in Europe and the U.S., and argued that these firms locate near the CBD or subcenter in a polycentric form for the sake of urbanization economies [15]. A similar spatial location pattern for professional knowledge-intensive business services was observed in Belgrade, Serbia [16]. In Warsaw, computer programming enterprises belonging to the ‘creative’ advanced business services cluster located along the transport corridor within secondary and tertiary business districts due to agglomeration benefits and relatively low land prices [17].

Due to deindustrialization in developed cities in the context of post-industrialization, scholars have paid more attention to high-tech services, especially information technology services, than to high-tech manufacturing activities. Characteristics of firms such as scale, sector, development stage, and ownership can also influence their location patterns in cities [1,18]. Although there are relatively few studies focusing on high-tech manufacturing firms, the existing literature has confirmed that high-tech services prefer clustering in secondary and tertiary business centers, while high-tech manufacturers tend to cluster in science parks and industrial parks in suburbs.

2.2. Location Determinants for High-Tech Firms

The previous literature has mainly considered four types of location determinants: classical locational factors, neoclassical locational factors, institutional factors, and talent-attractive factors. Classical locational factors are still important when it comes to high-tech firms because all high-tech activities need a specific place, capital, labor, and other production factors. However, there are some features that make them different from traditional manufacturing and producer services. Among classical location factors, the most studied are land price, human capital characteristics, and transport accessibility [11,19,20]. Agglomeration economies are probably the most studied neoclassical factors for high-tech firms [7,18]. Some studies confirm localization economies within high-tech manufacturing clusters [8,20]. The local pool of skilled labor and knowledge spillover indicated in localization economies have a facilitating influence on high-tech firms similar to the agglomeration of diverse institutions indicated in urbanization economies. Access to higher education institutions provides not only knowledge spillover but also skilled labor resources [11,21]. The infrastructures shared within a high-tech cluster include transportation infrastructures, telecommunication equipment, opening labs, and co-working offices, which implies that sharing mechanisms is about not only lower cost, but also attracting workers [7]. In terms of local supplier and consumer linkages, many studies extended the analysis to related and unrelated variety and found that high-tech start-up locations are affected by related variety, especially technologically related variety [5,8].

Institutional factors can play a vital role in the development of high-tech industries; therefore, it is common to investigate different institutional factors. Institutional factors include government actions such as taxes, environment regulations, land policies, and incentive programs [19]. High-tech firms typically face underexplored niche markets, high research and development investments, and scarce workforce resources, so they rely more on public support than other firms [22,23]. Formally planned districts for developing high-tech industries, such as science parks, industrial parks, innovation centers, ‘accelerators’, and ‘incubators’, are the main studied factors in the location of high-tech activities [24]. These places show a strong spatial correlation with the location of high-tech firms, particularly start-ups [1,11,22].

An increasing number of studies explored the effects of amenities and quality-of-life factors on location decisions. An amenity is something physical that helps to provide comfort, convenience, or enjoyment for people; these are critical to attract and retain the creative class, professional and technical workers, scientists, and engineers [25,26]. The creative classes’ preferences for amenities shape their distribution and finally affect the distribution of high-tech firms through the labor market, because these workers help generate innovations and play key roles in high-tech firms [27,28]. Therefore, it is important to discuss the effect of amenities on the location of high-tech sectors. The scope of amenities can vary across studies [4,29]. Natural amenities imply attractive environments, including mountains, oceans, high-quality air, and a comfortable climate. These kinds of amenities are important at the inter-city level [4,27]. Other scholars proposed the concept of environmental amenities, which includes parks and other public spaces [10]. These seminatural sites can also attract industries, especially creative industries, and increase neighbor land quality [4]. Transportation amenities have different effects on high-tech firm location. Inter-city transportation infrastructures have a negative influence, while urban transportation amenities such as subway and bus stations have positive effects in China [11]. Service and consumer amenities include restaurants, theaters, shopping malls, health care resources, educational resources, and so on [30]. Chinese scholars classify health care and educational services as public consumer amenities, while western scholars regard all these as private amenities. There is an urgent need for contextualized investigation because the public and private sectors provide different supports for amenity attractiveness, and it is important for urban planning and design [27].

As for industrial heterogeneity, manufacturers are different from service firms and high-tech industry is different from traditional industry. Peng et al. (2021) found that emerging industry land use is more affected by the quality of the population and innovation-driving forces than traditional industry land use [31]. Many scholars investigated the influence of amenities on location decisions for creative firms and knowledge-based firms but paid less attention to high-tech firms [32].

2.3. Research Gap

Existing studies provided a broad understanding of the location determinants of high-tech firms in different contexts. However, there still remains a research gap which needs to be filled. First, previous discussions of firm distribution pattern and determinants at the intra-urban scale have either focused less on high-tech firms or confused high-tech firms with knowledge-based firms, information and communications technology (ICT) firms, and high-tech producer service firms. High technology defined on an industrial basis cannot precisely differentiate high-tech activities, because not all firms in high-tech industries are technology-intensive, and conversely, some firms in low-tech industries are technologically intensive [21]. In addition, the emergence of high-technology business services challenges the distinction between knowledge-intensive business services and high-technology-intensive manufacturing in the contemporary information era. We regard high-tech firms as the most knowledge- and technology-intensive components of both high-tech and traditional industries; therefore, a deeper understanding at the firm level is necessary.

Second, the factors influencing high-tech firm location decisions are comprehensive, and the same factors can be explained from different perspectives. But there are few studies systematically investigating traditional and new location determinants and considering heterogeneity. Previous studies confirmed the effect of amenities on firm location in creative industries. According to related theories, urban amenities also attract and retain other types of skilled workers, who are crucial for high-tech firms. However, how amenities affect the distribution of talents and further influence high-tech firm location is still unclear.

Third, the urban context should be considered when explaining and generalizing research results. Scholars have investigated location determinants of high-tech firms in several growing cities within different contexts, such as the spatial structure of the city, land policy, and cultural context. Previous studies at the intra-urban level were mainly conducted in Chinese cities, but few studies examined high-tech firms in Shenzhen [7]. Given that Shenzhen is a benchmark and pioneer of Chinese high-tech industry development, it is necessary to focus on location determinants of high-tech firms in Shenzhen.

Based on previous findings, and to fill these research gaps, this study proposed an analytical framework to comprehensively understand the location determinants of urban high-tech firms (Figure 1).

3. Data and Methodology

3.1. Study Area

The study area is Shenzhen, Guangdong, which is located in southern China and neighbors Hong Kong (Figure 2). Shenzhen has become the ‘Silicon Valley of China’, and the output value of its high-tech industries reached CNY 2387.2 billion in 2018 [33,34]. On average, Shenzhen had 8.5 state-level high-tech enterprises per square km in 2020, and 24.7 thousand state-level high-tech enterprises in 2023 [35]. Shenzhen Software Park ranked first out of 44 software and industrial parks in the 2023 China Torch Software Industrial Bases List. Shenzhen has ranked first among cities in China for PCT international patent applications for 20 consecutive years, and Huawei was ranked first in the world for many years [36]. The inclusion of these high-tech Shenzhen firms demonstrates that Shenzhen insists on building a modern industrial system led by scientific and technological innovation, continuously forming new momentum and advantages for development, and enhancing the international competitiveness of high-tech industries [20].

There are 10 districts, 74 towns and sub-districts, and 676 communities at three administrative levels in Shenzhen. With the decentralization reform in China, local governance at the township level has become crucial to some extent, because a town is enough to bear a specific function for the whole city. Considering the trade-off between the availability of data and a higher spatial resolution, we chose to work at the township level.

3.2. Data and Variables

This study uses two main datasets collected from open platforms. The first consists of information on high-tech enterprises. In Shenzhen, 24,737 such enterprises received governmental certification in 2023 as published on the Website of the Administration of Certification of High-tech Enterprises. These enterprises are truly technology-intensive because they engage in continuous research and development to form their core independent intellectual property rights, as well as carry out business activities based on these rights, according to the certification conditions. Additional firm information is derived from the open data platform qcc.com (accessed on 1 June 2024), where we matched information such as location, scale, and sector with certified firms’ names.

The second dataset is the statistical characteristics of towns and sub-districts. According to the analytical framework developed in Section 2, ten indicators are selected as independent variables (

Table 1. Variables.

Category	Factors	Definitions	Unit
Classic locational factors (A)	Land rent (A1)	the standardized price of residential land	10 thousand CNY/m2
	Human capital (A2)	percentage of population with junior college education and above	%
	Inter-city transportation accessibility (A3)	number of highway entrances and exits	number
Agglomeration economies factors (B)	The stock of manufacturers (B1)	number of existing manufacturers in a sub-district	number
	The stock of tech-based firms (B2)	number of existing tech-based firms in a sub-district	number
	Universities and science development institutions (B3)	number of universities and science development institutions	number
Institutional factors (C)	Industrial parks (C1)	accumulative level of industrial parks	number
Amenity-based factors (D)	Transportation amenities (D1)	number of subway stations	number
	Private consumer services (D2)	density of private consumer services facilities	number/km2
	Public services (D3)	density of public services facilities	number/km2
	Environment settings (D4)	density of city parks and squares	number/km2

). The classic locational factors are land rent (A1), human capital (A2), and inter-city transportation accessibility (A3). Agglomeration economies factors include the presence of manufacturers (B1), tech-based firms (B2), and universities and science development institutions (B3). Development zones (C1) represents institutional factors. The amenity-based factors are transportation amenities (D1), private consumer services (D2), public services (D3), and environment settings (D4).

We use the standardized price of residential land appraised and published by Shenzhen government departments to indicate A1 and collect data from the Planning and Natural Resources Bureau of Shenzhen Municipality. We use the number of highways entrances and exits within sub-districts to represent A3 rather than railway stations and airports, because the latter are only distributed in several sub-districts. For C1, we adopt a classification variable to indicate the existence of industrial parks in each sub-district, with the value 0 if there is no industrial park within the sub-district and the value 1, 2, or 3 for industrial parks at the corresponding level (municipal or lower, provincial, and national). National and provincial industrial park information is collected on the basis of the China Development Zone Audit Announcement List 2018 and three batches of Guangzhou’s Characteristic Industrial Parks List from 2021–2023, respectively, while data on municipal and lower-level industrial parks are collected from the open platform qianzhan.com (accessed on 5 June 2024). Private consumer services (D2) include convenience stores, restaurants, cafes, shopping malls, supermarkets, and tea houses provided by the private sector, while public services (D3) include public hospitals, libraries, museums, art galleries, exhibitions, and educational facilities in the Chinese context [30]. Environment settings (D4) include city parks and squares. We collected point of interest (POI) data from amap.com (accessed on 5 June 2024) to indicate all factors in categories B and D and factor A3.

3.3. Methodology

Mixed methods consist of both the methods analyzing the firms and influencing factors. The research methodology framework for this study is shown in Figure 3.

3.3.1. Spatial Distribution Analysis Methods

• Kernel density estimation

Kernel density estimation was used in this study to describe the spatial pattern of firm points. If one industry has a larger frequency at a specific distance, it indicates that this industry has a higher degree of density at this distance. The kernel estimator used in this study is shown in Formula (1), and was developed by Silverman [30,37]:

$$\left(d\right)={\displaystyle \frac{1}{n\left(n-1\right)h}}\sum _i^{n-1}\sum _{j=i+1}^{n}f\left({\displaystyle \frac{d-d_{i,j}}{H_{opt}}}\right).$$

(1)

where $n$ is the number of observations, $d_{i,j}$ is the distance between the observation $i$ and the observation $j$, and $H_{opt}$ is the smoothing parameter that is significant, whose formula is as follows:

$$H_{opt}={0.9An}^{-1/5}.$$

(2)

where $A=\mathrm{m}\mathrm{i}\mathrm{n}(\sigma ,{\displaystyle \frac{IQR}{1.34}})$, $IQR$ is the interquartile range, a robust estimate of the spread.

2.

Getis–Ord Gi* Statistic

The Getis–Ord Gi* statistic is used to identify cold and hot spots of high-tech firms in this study. This model combines the classical G-statistics and Moran’s I-statistics and identifies the statistically significant clusters of high or low values by calculating their ZG scores. The formula by which the statistic is calculated is as follows:

$$G={\displaystyle \frac{\sum _{i=1}^n\sum _{j=1}^n{w}_{i,j}{x}_i{x}_j}{\sum _{i=1}^n\sum _{j=1}^n{x}_i{x}_j}}.$$

(3)

where $x_i$ and $x_j$ are attribute values for sub-district $i$ and $j$, $w_{i,j}$ is the spatial weight between features $i$ and $j$, $n$ is the number of sub-districts, and $i\ne j$. The ZG score for the statistic is computed as follows:

$$ZG={\displaystyle \frac{G-E\left(G\right)}{\sqrt{V\left(G\right)}}},$$

(4)

where:

$$E\left(G\right)={\displaystyle \frac{\sum _{i=1}^n{\sum }_{j=1}^n{w}_{i,j}}{n(n-1)}},$$

(5)

and:

$$V\left(G\right)=E\left(G^2\right)-E^2\left(G\right).$$

(6)

The Getis–Ord Gi* statistic of each sub-district is its ZG score. For positive ZG scores with obvious statistical significance, a higher ZG score implies a closer grouping of high values. For negative ZG scores with obvious statistical significance, a lower ZG score indicates a closer grouping of low values. The Jenks natural breaks method is used in this study to divide the ZG scores into six grades with different confidence levels.

3.

Bivariable spatial autocorrelation

The bivariable spatial autocorrelation can describe the spatial correlation and dependency of two geographical elements practically and effectively. We explore the spatial correlation of high-tech manufacturing and service enterprises using global bivariable Moran’s I; the equation is as follows [38]:

$$I_{ms}={\displaystyle \frac{n{\sum }_i^n{\sum }_{j\ne i}^n{w}_{ij}{z}_i^m{z}_j^s}{(n-1){\sum }_i^n{\sum }_{j\ne i}^n{w}_{ij}}}.$$

(7)

where $I_{ms}$ is the global bivariable Moran’s I between high-tech manufacturing and service enterprises. $n$ is the count of analysis units, a 1 km × 1 km grid in this study, and $i,j$ is the unit $i,$ and unit $j$. $z_i^m$ is the count of high-tech manufacturing firms in unit $i,$ and $z_j^s$ is the count of high-tech service firms in unit $j$. $w_{ij}$ is the weight matrix.

The global bivariate Moran’s I index can only provide a global assessment of the spatial correlation as a whole, but it ignores the instability of spatial process and cannot judge the local spatial agglomeration characteristics. Therefore, we use the local bivariable Moran’s I; the equation is as follows [39]:

$$I_i=d_i{\sum }_{j=1}^n{w}_{ij}{d}_j.$$

(8)

where $I_i$ is the local bivariable Moran’s I between high-tech manufacturing firms in unit $i$ and high-tech service firms in unit $j$. $d_i$ and $d_j$ are the variance standardization of high-tech manufacturing firms in unit $i$ and high-tech service firms in unit $j$. $w_{ij}$ is the weight matrix.

3.3.2. GeoDetector Model

The GeoDetector model is mainly used to detect the spatial differences of geographical elements and analyze the drivers of these differences, including factor detection, interaction detection, and ecological detection [40]. This study used GeoDetector software to quantitatively analyze the impact of every influencing factor and the interactive effects between different factors on high-tech firm location. The formula is as follows [41]:

$$q=1-{\displaystyle \frac{\sum _{h=1}^L{N}_h{{\sigma }^2}_h}{N{\sigma }^2}}.$$

(9)

where $q$ is the statistic measuring the strength of the effect of an influencing factor and takes a value between 0 and 1. The closer to 1, the more this factor contributes to the dependent variable. $h=\mathrm{1,2},\dots ,L$ is a given class of an influence factor and $L$ is the number of classes. $N_h$ and $N$ are the numbers of samples in class $h$ and entire study area, respectively; and ${\sigma }_h$ and $\sigma$ are the variance of the dependent variable in class $h$ and the entire study area, respectively. A $p$-statistic, an indicator of statistical significance for each explanatory variable, is also calculated by a non-central F-distribution [42]:

$$p\left(q<x\right)=p\left(F<{\displaystyle \frac{N-L}{L-1}}{\displaystyle \frac{x}{1-x}}\right)=1-a.$$

(10)

where $a$ is the probability of q being higher than or equal to $x$. In a 95% confidence interval, a factor with a $p$ value greater than 0.05 is considered to have a statistically insignificant relationship with the dependent variable and can be eliminated from the model.

Moreover, by estimating the value of the $q$-statistic corresponding to the interaction of two explanatory variables, GeoDetector can also quantify the degree of the interactive impact of each pair of factors on high-tech firm location.

4. Empirical Results

4.1. Spatial Distribution of High-Tech Firms in Shenzhen

The Kernel density pattern for all high-tech firms is shown in Figure 4. High-tech firms are distributed in a spatial pattern of ‘one core and six clusters’, with higher density in the west than in the east. In the western districts such as Nanshan District, Bao’an District, Longhua District, and Futian District, many high-tech firms locate along the coastal line, and areas with low firm density are largely mountain land in ecological preservation zones. Specifically, the west side of Shenzhen has the Yuehai–Nanshan core cluster, Shatou cluster, Fuhai cluster, and Dalang center. The southeastern districts have a mountainous coastline and other large mountainous areas, especially Yantian District and Dapeng New District, so firms on the east side are mostly located in northeastern districts for ecological conservation reasons. Specifically, there is one Baolong cluster in Longgang District.

Spatial agglomeration patterns differ between high-tech manufacturing and service industries. Figure 5 shows the result of Getis–Ord Gi* statistic estimation. First, the number of high-tech manufacturing enterprises (8857) is significantly less than the number of high-tech service enterprises (15,880) in Shenzhen. The kernel density of manufacturing enterprises is also much lower than that of service enterprises (Figure 4b,c), which is consistent with findings that in the process of industrial transformation and upgrading in Shenzhen, information technology service enterprises have replaced manufacturers as the driving force of innovation [43]. Second, the high-tech service firms show an extremely strong single-center agglomeration pattern, forming a hot spot expanding outward from the Shenzhen Hi-tech Industrial Park (SHIP) in Nanshan District. The Kernel density value of this core area is many times higher than those of other clusters such as Shatou and Dalang. Three high-tech service clusters form from the expansion of SHIP, the transformation of the old industrial area (Futian CBD), and the development of a leading firms agglomeration area in the Dalang sub-district, respectively [34]. On the other hand, high-tech manufacturing firms form a contiguous hot spot agglomeration in the northwestern part of the city (Figure 5b).

The global bivariate Moran’s I is 0.359, which indicates that there is a significant positive spatial co-aggregation relationship between high-tech manufacturing firms and high-tech service firms. As shown in Figure 6, we illustrate local bivariate Moran’s I in a BiLISA cluster map containing four types of correlation of service firm–manufacturing firm. H–H clusters are mainly distributed in the western city and re always surrounded by an L–H cluster. The result shows that the two kinds of firms are co-located at the city level; however, high-tech service firms prefer core sites in a city and high-tech manufacturing firms locate surrounding these core sites.

4.2. Regression Analysis of Location Determinants for High-Tech Firms

4.2.1. Single-Factor Detection

The GeoDetector results are shown in

Table 2. Single-factor detection results.

Factors	All High-Tech Firms	High-Tech Manufacturing Firms	High-Tech Service Firms
Classic locational factors (A)	A1	0.050	0.225 **
	A2	0.103	0.293 **
	A3	0.191	0.405 **
Agglomeration economies factors (B)	B1	0.231 **	0.242 **
	B2	0.572 **	0.297 **
	B3	0.241	0.034
Institutional factors (C)	C1	0.178 **	0.112 **
Amenity-based factors (D)	D1	0.162 *	0.083
	D2	0.186 **	0.237 **
	D3	0.035	0.155 **
	D4	0.138 **	0.249 **

* and ** refer to statistical significance at 10% and 5%.

. In general, agglomeration economies, institutional factors, and amenity-based factors have more significant effects on the location of high-tech firms in Shenzhen than classical locational factors. Location determinants of high-tech firms follow the general rules of business location selection, but these firms show particular preferences regarding institutions and high-quality urban amenities.

First, the results indicate that the dependence of high-tech firms on agglomeration economies still exists, and localization and urbanization agglomeration play a significant role in the distribution of high-tech firms. Sub-districts with high-density manufacturing firms not only provide a rich source of suppliers and customers along the supply chain, but also enable the emergence of new high-tech firms through knowledge spillovers and learning effects. In terms of urbanization agglomeration economies, sub-districts hosting many high-tech firms in different industries can attract and nurture other high-tech firms by meeting multiple business needs efficiently, lowering search costs owing to spatial proximity. In addition, a clustered sub-district of high-tech firms provides a diversified but related industrial background for innovation and creativity.

Second, institutions and planning also have an obvious role in the distribution of high-tech enterprises. The establishment of industrial parks is a factor in location choice for all high-tech enterprises, because industrial parks provide structured clustering places and tend to implement the government’s policy orientation and industrial support programs. Especially for some strategic industries, industrial parks often act as incubators and boosters for start-ups [23].

Third, the effects of amenity-based factors have replaced those of traditional location factors for high-tech enterprises. Our results show that the traditional location factors have no statistically significant effects at a 90% confidence level on the location of technology enterprises. The amenity-based factors sorted from greatest to least effect are private consumer services (D2) > environmental settings (D4) > transportation amenities (D1) > public services (D3). This can be explained by the fact that private services are more accessible than public transportation facilities and services. For example, the service scope of large public hospitals and secondary schools covers the city. Private consumer services have a wide distribution but not subway stations which show an obvious spatial imbalance. Sub-districts where these facilities are concentrated are also ‘central places’ according to Walter Christaller, and naturally become places where enterprises are concentrated. In addition, both the private consumer services and the natural or artificial open space conditions surrounding the firms influence the extent to which employees’ daily consumption, working, and hedonic needs are met. Thus, different kinds of amenities directly and indirectly influence the distribution of high-tech firms.

There are also clear differences in the explanatory power of different factors for firm distribution in the high-tech service and manufacturing sectors. The model results show that classical locational factors only have an effect on the location of high-tech manufacturing enterprises, while universities and science development institutions only have an effect on the location of high-tech service enterprises (Table 2). Land rent, human capital, and inter-city transportation accessibility only influence location for manufacturers, because these enterprises have a need for production space and transportation of physical goods. In addition, there are differences between the two sectors in their sensitivity to different types of amenities. The model shows manufacturing firms are more influenced by public service facilities and environmental settings, while service firms are more influenced by the accessibility of intra-city transportation.

4.2.2. Factor Interaction Detection

The results of factor interaction detection for all high-tech firms are shown in Figure 7. The combined explanatory power of any two factors is stronger than that of a single factor, and enhancement types are mostly nonlinear. Among all 55 factor combinations, the three with the highest explanatory power are ‘D4∩B2’, ‘D3∩B2’, and ‘A3∩B2’, all of which have nonlinear enhancement. B2 is present in all three factor combinations. This shows that once high-tech firms locate in a specific sub-district with good amenities and easy access to external transportation, their agglomeration effect will add to the attractiveness of this area for similar firms, and further strengthen the concentration of high-tech firms. In addition, A1 and B3 also strengthened the explanatory power of the other factors. This is probably because, whether in the core or peripheral areas of the city, research institutes and higher rents are both concentrated in sub-districts with higher levels of other conditions than the surrounding ones, and play a mutually reinforcing role.

There are clear differences between the high-tech service and manufacturing sectors in the explanatory power of different factor combinations; interaction detection results distinguished by sector are shown in Figure 8. The factor combination with the most explanatory power for high-tech manufacturers’ location is ‘D3∩B1’. This result reflects the importance of localization agglomeration and public service facilities for high-tech manufacturing firms to attract skilled workers; the combination of the two factors tends to attract more high-tech manufacturing firms to locate in a sub-district. Moreover, the results show that amenity-based factors enhance the explanatory power of the other factors for the distribution of high-tech manufacturing firms to some extent but are less likely to enhance each other.

The three factor combinations with the most explanatory power for high-tech service firm location are ‘D3∩B2’, ‘B3∩B2’, and ‘D4∩B2’. B2 is the most frequent factor, as in the results for all high-tech firms. It is noteworthy that C1 significantly enhances the explanatory power of other factors for service firms but not manufacturing firms. Given the spatial distribution difference, high-tech service firms have a greater preference for industrial parks than manufacturing firms do when they expand from the core cluster.

5. Discussion

5.1. Government Planning and the Distribution of High-Tech Firms

The Shenzhen government plans the spatial structure for high-tech industry development in the form of industrial parks, which has a significant effect on the distribution of enterprises. Mature high-tech agglomerations come from industrial parks planned and constructed in the past, and most of the growing agglomerations are also located in currently planned parks. Since the reform and opening up, processing enterprises have emerged and clustered in the inner bay on the east bank of the Pearl River, which is the core of the Guangdong–Hong Kong–Macao Greater Bay Area now.

The plan for high-tech industry development in Shenzhen began when SHIP was established in 1985 in the present Yuehai sub-district. SHIP was upgraded to a state-level high-tech zone in 1996 and become an important pilot project for the construction of a world-class high-tech zone in the country. After 40 years’ sustained development and industrial upgrading, SHIP has become a high-tech service agglomeration core for the whole city, dominated by information services. Compared with other clusters in the former Special Economic Zone, it was these historical planning and location decisions that led the Yuehai sub-district to become such an enterprise agglomeration core [44]. In 2019, the spatial scope of the state-level high-tech zone expanded to 12 industrial parks with high potential, and SHIP was positioned as a comprehensive innovation core of Shenzhen (Figure 9). Moreover, the Baolong and Pingshan clusters in the northeastern city were included in the state-level high-tech zone.

Shenzhen supports the development of high-tech industrial by releasing high-quality industrial space for industrial parks, as well as through land supply, finance, and taxation policy. In the Chinese context, Shenzhen has always supported the development of high-tech manufacturing industry driven by innovation in order to enhance the security and stability of the whole economy, which is more proactive and effective than the industrial policies of ‘deindustrial cities’ and ‘postindustrial cities’ [45,46]. In 2001, the ‘Shenzhen High-tech Industrial Belt Planning and Development Outline’ was issued, which coordinated and optimized the spatial allocation of resources for innovation for the first time; industrial parks became the main sites for the implementation of industrial policies. Facing the needs of the new strategic high-tech industry, Shenzhen has narrowed the space for private industrial parks and supplied land for industrial parks from the government, implemented by state-owned enterprises. Moreover, new mixed-use and multi-story industrial building are provided through regeneration of old industrial districts to meet high-tech firms’ demands and improve industrial land use efficiency [47]. The role of active government enables high-tech enterprises to move into industrial parks and form agglomeration effects efficiently, thus bringing continuous innovation power to the city.

5.2. Production Factors Demand Preferences and Industrial Heterogeneity

While both manufacturing and service firms are high-tech enterprises, their demand for different production factors leads to heterogenous distribution and location determinants. From the perspective of the core–periphery structure of Shenzhen, high-tech manufacturing firms are widely distributed in the urban periphery, such as the northwestern part of the city and the new industrial centers in the northeast. While high-tech service firms are clustered in the centers and have limited expansion towards the sub-centers. Technology widens the differences between the two sectors’ demands for land, human resources and transportation. In terms of land, manufacturing enterprises generally need larger areas for R&D, testing, warehousing, and production, and some advanced manufacturing industries require large-scale production to achieve input-output balance, while high-tech service industries’ land demand is smaller and mainly dominated by office space. In terms of transportation, ICT significantly reduces face-to-face communication between manufacturing enterprises throughout the production process, but raw materials, intermediates, and products have physical forms and must rely on cargo traffic. Most of the products and services of service enterprises are in non-physical forms and can be transmitted through communication networks; however, the face-to-face communication required by business is the core of service enterprises, which is why they are more sensitive to the movement of people within the city. In terms of human resources, manufacturing firms need a large number of skilled workers capable of learning new techniques and operating new equipment, which is closely related to the demographic educational background of the sub-districts where they locate. High-tech service firms need a large number of highly educated specialists, creators, and managers. These three differences make manufacturing firms more sensitive to the costs of land, inter-city transportation, and skilled workers, while service firms are more sensitive to attracting top talents.

Compared with the core area, the peripheral area has lower land rent and easier inter-city transportation, but is not well equipped with public services and has less high-quality open space. For high-tech manufacturing enterprises that need to attract skilled labor, it is especially important to choose a location in the peripheral area that is close to public service amenities and excellent environmental settings. Because high-tech service firms expand mainly for the purpose of proximity to the market and customers or talents, they choose locations with high transportation accessibility close to the city’s scientific research and other innovative institutions.

5.3. The Important Role of Amenities in High-Tech Firms’ Location

This study reveals that amenity-based factors, especially public services, can improve the effects of other location determinants on high-tech manufacturing firms. In previous studies, the effects of amenities on manufacturing firms have been less reported than effects on creative firms [4,21,30]. However, when the development of high-tech firms is driven by innovation, the considerations for firm location will change from the returns on traditional factor inputs to the margin between the benefits and the costs of innovation, which results in location determinants changing. On the one hand, manufacturing enterprises still have to be located in the periphery of the city with insufficient service supply, considering the cost of land and external cargo traffic. On the other hand, the living needs of high-tech firm employees are gradually increasing. Therefore, the influence of urban amenities on firm location in the periphery of the city is even greater than that in the core of the city (Figure 10). This study’s finding that public service amenities play a greater role for manufacturing firms than for service firms is consistent with other studies in Barcelona and Beijing [7,48]. We argue that analysis of the impact of urban amenities should not only focus on the creative industries.

6. Conclusions

This study investigates the distribution and location determinants of high-tech firms within the city of Shenzhen. We construct an analytical framework combining four perspectives to explain the relationship between various factors and the location of high-tech firms. The analytical framework also highlights the characteristics of firms and we further consider industrial heterogeneity. The main conclusions of this study are as follows: (1) High-tech firms in Shenzhen form a spatial pattern of ‘one core and six clusters’, with high density in the west. (2) Manufacturing firms are widely distributed in the urban periphery, while service firms cluster in the city centers. (3) Agglomeration economies, institutional factors, and amenity-based factors have more significant effects on the location of high-tech firms than classical locational factors. (4) The roles of classical locational factors and amenities in the distribution of manufacturing and service firms are different because of their different demands for production factors.

For policymakers, the above findings suggest that there is considerable room for public action to support high-tech firms in areas such as industrial park and transportation systems planning and public services allocation. According to findings about high-tech manufacturing firms, participatory planning with in-depth surveys of target sectors and firms about their specific demands helps improve industrial parks planning. A mix of residential, tertiary, and industrial buildings within a sub-district instead of highly segregated land use is more suitable for high-tech firms. Given agglomeration economies and the need for industrial collaboration, efficient organization of passenger and cargo traffic around high-tech clusters is important in attracting both manufacturing and service firms. The improvement of public services in peripheral areas of the city, especially in ‘industrial new towns’, is a competitive way to attract high-tech firms; local government should focus not only on basic production needs, but also on the daily living needs of workers.

There are some limitations in this study. This study is a cross-sectional analysis of data from 2023, lacking longitudinal changes in the influencing factors of the distribution of high-tech firms. Future research can collect panel data and analyze the evolution of the spatial pattern and location determinants over time. In addition, this study only considers differences between manufacturing and service industries; future research can explore heterogeneity in firm sizes, ownership structures, and firm life cycles.