Location Determinants and Patterns of Foreign Logistics Services in Shanghai, China JUNJIE HONG

Nested logit models are introduced and applied to an investigation of location determinants and patterns of foreign logistics services within a Chinese city. The estimation results based on recent census data show that foreign logistics establishments favoured the sites near the central business district. They were attracted by good transport infrastructure and vigorous market demand. High population density deterred foreign logistics establishments, while the availability of supporting services attracted them. We also observed that the central city area’s attractiveness to foreign logistics services had decreased since the mid-1990s. Finally, some implications for policy-makers are suggested.

INTRODUCTION

The global logistics industry has entered a stage of rapid growth. The Ministry of Commerce of China estimated that in 2004 the cost of logistics constituted about 21 per cent of GDP in China. For many commodities, logistics costs are 40 to 50 per cent higher than they would be in the United States [Bolton and Wei, 2003]. Despite this weakness, logistics businesses have developed rapidly in China. Between 2001 and 2003, 70 per cent of logistics service providers experienced a 30 per cent annual increase in the country. The logistics industry reported an annual growth rate of 31 per cent in 1999, 35 per cent in 2000, 55 per cent in 2001, and is expected to continue to expand quickly in the near future [Bolton and Wei, 2003]. Promising market potential has attracted many foreign logistics entrants into China [Hong et al., 2004]. Foreign investments in logistics-related sectors1 increased from US$7.0 billion in 1996 to US$14.8 billion in 2001.2 Since the late 1970s, the logistics industry has been transformed from an industry dominated by a few large state-owned enterprises to one consisting of many private and foreign logistics providers. With China’s entry into the World Trade Organization (WTO) in 2001 most restrictions on the provision of logistics services by foreign firms have been removed gradually, which will encourage more foreign firms to enter the market.

Dr. Junjie Hong is associate professor at the School of International Trade and Economics, University of International Business and Economics, Beijing, China, 100029. Email: [email protected] The Service Industries Journal, Vol.27, No.4, June 2007, pp.339–354 ISSN 0264-2069 print/1743-9507 online DOI: 10.1080/02642060701346490 # 2007 Taylor & Francis

340

THE S ERVICE INDUS TRIES JOURNAL

This is a study of the location behaviour of foreign logistics services within a Chinese city – Shanghai. Shanghai is the centre of economy, finance and international trade in China. The government has approved a plan to build Shanghai into a logistics hub of East Asia. To accomplish this objective, a number of preferential tax and land use policies have been granted to foreign logistics firms. According to the Second Census on All Basic Units in China, approximately one-third of foreign logistics establishments in China clustered in Shanghai in 2001, suggesting that the city has been successful in attracting foreign logistics investment. Due to its advantages in market size and government support, Shanghai is likely to serve as a logistics hub in the Northeast Asian region [Oum and Park, 2004], and hence provides us a suitable context to study the intrametropolitan location3 of foreign logistics services. Although some progress has been made in understanding intrametropolitan locations of service firms, most previous studies [e.g., Coffey et al., 1996; Coffey and Shearmur, 2002; Shearmur and Alvergne, 2002; Aguilera, 2003; Wernerheim and Sharpe, 2003] focused on business services or producer services.4 Little research has been done concerning the location of logistics services within a city. Also, there is a lack of comprehensive understanding of service location in developing countries. The goal of this paper is to contribute to a better understanding of intrametropolitan location determinants and patterns of foreign logistics services in a developing economy. It also makes an attempt to use a nested logit model, which has rarely been used in the literature on service location, to study the location of logistics services within a metropolitan area. The remainder of this paper is organised as follows. The next section gives the literature review, followed by a description of methodology, data source and explanatory variables. We then report and discuss the empirical results. The paper concludes with a summary of the major findings and a discussion of policy implications.

LITERATURE REVIEW

This section reviews the literature, including intrametropolitan location determinants of service activities, location patterns of service activities, and discrete choice approach used in empirical location work. Location Determinants of Service Activities Little research has been done to investigate the location determinants of logistics activities. One exception is Oum and Park [2004], who analysed the location of distribution centres of multinational enterprises across Northeast Asian countries. They found the following location considerations: market size, transport conditions, labour considerations and input costs. According to the best of our knowledge, there appears to have been no attempt to study location determinants of logistics services within a metropolitan area. Therefore, we have to rely on some findings from the extant literature on intrametropolitan location determinants of various types of service activities. It is believed that there is some degree of commonality between different types of services.

FOREIGN LOGISTICS SERVICES IN SHANGHAI, CHINA

341

It is now well recognized that there is a need to study the impact of locationspecific attributes on service location [Wernerheim and Sharpe, 2003]. The role of some traditional location factors, such as market demand, has been identified by existing research [e.g., Coffey et al., 1996; Wernerheim and Sharpe, 2003]. Shukla and Waddell [1991] found that in the Dalls– Fort Worth metropolitan area, the central business district (CBD) had a strong centralizing influence on service location, and accessibility to the airport and major highways acted as a significant pull factor. A recent study by Aguilera [2003] identified the role played by the relationship between service providers and customers. If faceto-face contact with customers is important, service activities will locate near their market. The availability of supporting services, such as finance and hotels is another important consideration, because it can ease business operations [Ihlanfeldt and Raper, 1990; Wu, 2000]. The location choices of foreign investments may be influenced by government policies as well. In China, some special zones can offer preferential policies, and thus attract more foreign investments [Wu, 2000; He, 2002; 2003]. Location Patterns of Service Activities One of the central concerns in the literature is whether the central city area has lost its attractiveness to service activities. A number of previous studies [Cuadrado-Roura and Gomez, 1992; Baro and Soy, 1993; Gorden and Richardson, 1996; Harrington and Campbell, 1997] have observed that service activities tend to locate outside the traditional CBD. For instance, Aguilera [2003] found that in the Lyon metropolitan area, there was an 85 per cent increase in the number of business service firms in peripheral areas, but this figure for the central city area was only 63 per cent between 1982 and 1996. By investigating the location pattern of services activities in the Montreal metropolitan area, Coffey and Shearmur [2002] noted that the CBD contributed only 6 per cent of total service employment growth, compared with more than 40 per cent for other districts. The above evidence shows an absolute growth but relative decline for the CBD, and implies that the CBD is likely to lose its dominant position. Some possible centrifugal forces include: outward movement of some service clients and employees, and the development of edge cities [Shearmur and Alvergne, 2002]. Moreover, with the improvement of transport infrastructure in peripheral sites, the traditional centre is no longer most accessible [Aguilera, 2003]. The central city is normally associated with higher costs and suffers from congestion problems, which make it less attractive to new establishments. Some researchers argue that the growth in suburban areas does not necessarily lead to the demise of CBD [Anas et al., 1998]. Moulaert and Gallouj [1995] found that some business services continued to seek CBD areas, while employmentintensive producer services tended to locate in non-CBD areas. They identified the following location factors: the price of office space, the infrastructure and accessibility, the social – cultural environment, and the proximity to supporting services. Although the traditional central city area is normally associated with high land

342

THE S ERVICE INDUS TRIES JOURNAL

costs, it is still attractive to firms that intend to maximise their accessibility and facilitate face-to-face contact. A recent study by Shearmur and Alvergne [2002] found that location patterns of business services were a combination of concentration and dispersal. For instance, while small consumer-oriented branches diffused across the entire metropolitan area, ‘global’ head offices still preferred the CBD. Ozdemir [2002] analysed the distribution of foreign service investments in Istanbul, and found that about 79 per cent of foreign transport firms chose the traditional central city. Only a small proportion located at the peripheral area. Discrete Choice Approach Used in Empirical Location Work Increasing availability of data sets has driven many researchers to use econometric advances to analyse industrial locations. A commonly used econometric model is the discrete choice approach, which mainly includes conditional logit model and nested logit model. Carlton [1983] applied a conditional logit model to a spatial framework for the first time. This pioneering work suggested that location decisions could be modelled in a partial equilibrium setting, following a verifiable economic process that results from profit maximising behaviour. Taking firms’ location choices as the dependent variable and locational attributes as independent variables, the conditional logit model performed well in analysing industrial location. This approach has been widely used in the sequential studies on firm location [e.g., Shukla and Waddell, 1991; Guimaraes et al., 2000; Leitham et al., 2000; Wu, 2000]. For instance, Wu [2000] used a conditional logit model to study FDI location within metropolitan Guangzhou, and identified some important locational factors, such as market size, distance to the city centre and government policies. A troublesome assumption of conditional logit model is Independence of Irrelevant Alternatives (IIA assumption), which states that the ratio of the probabilities of choosing one alternative over another is unaffected by the presence or absence of any additional alternative in the choice set. This assumption may be violated in industrial locations since investors may view neighbouring districts as a closer substitute than more distant districts [Head et al., 1999; Figueiredo et al., 2002]. An alternative approach is the nested logit model proposed by McFadden [1978]. It could be used to circumvent the problem of IIA assumption. Some previous studies [e.g., Hansen, 1987; Ondrich and Wasylenko, 1993] have applied a nested logit model to study industrial location. For instance, Hansen [1987] classified the Sao Paulo region in Brazil into two areas. Firms were assumed to choose an area first, followed by the choice of a city in the area. This research successfully identified the following location determinants: market size, urban infrastructure and urbanisation economies. The above literature review reveals some possible service location determinants (such as market size, transport condition, input costs and government policies), which provide us guidance for selecting explanatory variables. There seems to be mixed evidence concerning whether the central city area has lost its attractiveness to service activities, suggesting a need for further study. This paper uses nested logit models to study location determinants of foreign logistics services in a Chinese city. We also attempt to analyse distribution patterns of logistics services within the city.

FOREIGN LOGISTICS SERVICES IN SHANGHAI, CHINA

343

METHODOLOGY

Location choice is an important decision since some factors (such as market size and transport conditions) that can influence a firm’s profit are different across districts in a city. The representative logistics service provider encounters a number of districts as location alternatives. It is plausible that the firm will locate in the district in which it could maximise its profit. Formally, the district j is chosen by firm i if and only if Pij ¼ max {Pij0 ; j0 ¼ 1, . . . . . . N}

(1)

Where, Pij0 denotes the profit of firm i provided it locates in district j0 . N represents the number of choice alternatives (i.e., districts). We assume that the profit function of firm i is composed of a deterministic (Uij) and stochastic term (1ij): Pij ¼ Uji þ 1ij

(2)

Following previous studies [e.g., Guimaraes et al., 2000; Leitham et al., 2000], we assume that the deterministic term is a linear combination of observed district characteristics (X1j, X2j,. . ., Xnj). Then the profit of firm i provided it locates at district j can be expressed as: Pij ¼ P(X1j , X2j , . . . . . . , Xnj ) þ 1ij

(3)

If the error terms are independently and identically distributed as Weibull density function, the probability of firm i choosing district j (Pij) can be described by: pij ¼ exp Pij (X1j , . . . . . . , Xnj )=

N X

exp Pij0 (X1j0 , . . . . . . , Xnj0 )

(4)

j0 ¼1

The above conditional logit model has been used by a number of previous studies [e.g., Carlton, 1983; Wu, 2000]. However, there is a troublesome assumption (i.e., IIA assumption). An alternative approach that can circumvent the problem is a nested logit model. Assuming that there are M areas. Logistics firms first choose an area to locate, followed by the choice of a specific district within the area. We obtain a nested logit model as follows: pijjm ¼

exp Pijm J P exp Pij0 m

(5)

exp (Pim þ a0 vm ) M P exp (Pim0 þ a0 vm0 )

(6)

j0 ¼1

pim ¼

m0 ¼1

344

THE S ERVICE INDUS TRIES JOURNAL

vm ¼ log

J X

exp (P j0 )

(7)

j0 ¼1

Where, pijjm is the probability of firm i choosing district j conditional on choosing area m, P ijm is the profit of firm i if it locates in district j, which is under area m, Pimdenotes the profit of firm i if it locates in area m, pim represents the probability of firm i choosing area m, vm is the inclusive value of area m, which contains J districts, and a0 is a measure of the similarity of districts under area m. The dependent variable in the above nested logit model is a dummy variable, which is equal to 1 when a district is chosen and 0 otherwise. The location-specific variables (such as market demand and transport conditions) are taken as independent variables. The nested logit model is employed since it has the following advantages: (1) it is consistent with firms’ behaviour of profit maximisation; (2) it has a simple mathematical structure and is easy to estimate; and (3) it can circumvent the problem of IIA assumption. Some previous studies [e.g., Ondrich and Wasylenko, 1993] have used this model to study industrial location successfully. However, to date, no research has used the methodology to study service location. This paper makes such an attempt for the first time.

DATA SOURCE AND EXPLANATORY VARIABLES

The data used for this research could be classified into two groups: firm-level data and district-level data. The firm-level data were drawn from ‘The Second Census on All Basic Units in China’, which was conducted by the government at the end of 2001. We obtained the data from the State Statistical Bureau of China. ‘Foreign logistics firms’ in this research are defined as all types of foreign firms that provide logistics services for other manufacturing or commercial companies and have their own independent accounting system. Logistics services include various modes of transportation, public warehousing, some value-added services, arrangement of freight (or cargo) transportation and export brokers, etc. Because the land area of some districts in Shanghai varied significantly in 1993 and has remained almost unchanged since then, we chose foreign logistics firms established between 1993 and 2001 as the observations. The geographical distribution of foreign logistics providers in the city is reported in Table 1. It indicates that nearly 51 per cent of the firms cluster in Pudong district. In the regression, all districts but ‘Chongming’ were taken as alternative choices. The district of ‘Chongming’ was excluded because it received no foreign logistics investment during the period. Moreover, ‘Chongming’ is an island, quite far away from Shanghai mainland. The 18 districts included were classified into 2 areas according to their geographical location. These are central city area (CENTRE) and peripheral area (OTHER), as shown in Table 1. We assume that investors make location decisions sequentially. They first choose an area between CENTRE and OTHER, followed by the choice of a specific district in the area. The dependent variable in our econometric model is a dummy variable (LOCATION), which equals 1 if a district is chosen and 0 otherwise (see Table 2).

345

FOREIGN LOGISTICS SERVICES IN SHANGHAI, CHINA TABLE 1 DISTRIBUTION OF FOREIGN LOGISTICS FIRMS IN SHANGHAI

Areas (abbreviation) Central city area (CENTRE)

Peripheral area (OTHER)

Districts

Number

%

Huangpu Luwan Xuhui Changning Jin’an Putuo Zhabei Hongkou Yangpu Pudong Minhang Baoshan Jiading Jinshan Songjiang Qinpu Nanhui Fenxian Chongming Total

36 15 20 50 15 14 4 73 6 305 8 38 8 0 5 1 1 1 0 600

6.00 2.50 3.33 8.33 2.50 2.33 0.67 12.17 1.00 50.83 1.33 6.33 1.33 0 0.83 0.17 0.17 0.17 0 100

The district-level data, which were used as independent variables, were mainly drawn from the Statistical Yearbook of Shanghai [Shanghai Statistical Bureau, 1994 –2002]. Some distance data were computed from the digital map of Shanghai. Table 2 gives the definitions of these variables. Location could affect the profit of foreign logistics firms because market size, input costs, transport conditions, government policies and the availability of supporting services vary spatially across districts in an urban area. We now proceed to explain the explanatory variables. In this research, market demand for logistics services is measured by the total number of employees in industrial and commercial companies in a district (LALLEMP). Compared to state-owned companies, privately owned firms are more likely to use external logistics services [Bolton and Wei, 2003]. Therefore, the percentage of private employment over state employment (LPRI) was used as a proxy for logistics outsourcing probability. Market size is expected to be a positive factor in attracting logistics investment. Land is an important input for logistics firms. Normally, land price in a city is negatively related to the distance to CBD [Li, 1997]. Therefore, we used the distance to CBD5 (DISCBD) as a proxy for land price. It should be noted that DISCBD might capture some other effects (such as agglomeration economies and facilitation of faceto-face contact), which makes the CBD attractive to service firms. Land price is also affected by land availability and development costs [Yeh and Wu, 1996], which normally vary with population density. Population density in a district (LPOPDEN) was thus used as another proxy for land price.

346

THE S ERVICE INDUS TRIES JOURNAL TABLE 2 DEFINITIONS AND DESCRIPTIONS OF VARIABLES

Symbol Dependent variable LOCATION Independent variables LALLEMP LPRI DISCBD LPOPDEN SEADUM DISPORT ROADUM RAIDUM DISPDAIR DISHQAIR NTZ HOTEL FINANCE

Description

Data source

Dummy variable equals to one if a district is chosen as firm location, zero otherwise

The second census on all basic units, China

Natural logarithm of total number of employees in industrial and commercial companies Natural logarithm of percentage of private employment over that in state-owned enterprises Linear distance from district centre to the central business district (CBD) Natural logarithm of population density Dummy variable equals to one if the district has at least one sea berth, zero otherwise Linear distance from district centre to the Shanghai Seaport Dummy variable equals to one if the district has state-level roadway, zero otherwise Dummy variable equals to one if the district has railway transport, zero otherwise Linear distance from district centre to the Pudong International Airport Linear distance from district centre to the Hongqiao International Airport Dummy variable equals to one if the district has new economic and technological zones (NTZ) or free trade zones (FTZ), zero otherwise Number of ‘foreign concern’ hotels Number of foreign financial companies

SYSH, 1994–2002 SYSH, 1995–1997 Shanghai Map 2001a SYSH,b 1994– 2002 Shanghai Map 2001 Shanghai Map 2001 Shanghai Map 2001 Shanghai Map 2001 Shanghai Map 2001 Shanghai Map 2001 Author’s database SYSH, 1994–2000 SYSH, 1994–2000

Notes: aDistance computed from digital map of Shanghai, 2001. b SYSH denotes the Statistical Yearbook of Shanghai.

Three variables, DISPDAIR, DISHQAIR and DISPORT, were used to measure the distance to Shanghai Pudong International Airport, Shanghai Hongqiao International Airport and Shanghai Seaport respectively. In addition, three dummy variables, SEADUM, ROADUM and RAIDUM were used to differentiate the districts that have, respectively, deep-water berths, state-level highway and railway from the others. Foreign logistics firms are expected to value good transport linkage. In China, the government establishes new economic and technological zones (NTZ) and free trade zones (FTZ) to attract foreign investors. Therefore, a dummy variable (NTZ) was used to differentiate the districts that have NTZ or FTZ from the others. Foreign logistics service providers are expected to favour these special zones. Finally, the availability of supporting services, such as finance and hotels may be relevant to logistics location, because it can ease business operations. Therefore, the number of ‘foreign concern’ hotels (HOTEL) and the number of foreignowned financial companies (FINANCE) in a district were used to measure the availability of supporting services.

FOREIGN LOGISTICS SERVICES IN SHANGHAI, CHINA

347

ESTIMATION RESULTS AND DISCUSSION

Some correlation coefficients among independent variables are found to be high. To avoid the presence of multicolinearity, the pairs with high correlation coefficients will not be included in the same equation. The estimation proceeds by maximising the log-likelihood of observations. The specifications with reduced sets of explanatory variables, which have satisfactory goodness of fit and display robustness, are selected and reported. We now proceed to discuss the estimation results. Nested Logit Estimation Results The nested logit estimation results are given in Table 3. Coefficient estimates and the associated asymptotic t-values are reported. It shows that most coefficient estimates are statistically significant and get expected signs. The adjusted likelihood ratio index in the range of 0.344 and 0.379 indicates that the estimation results are a satisfactory fit. The negative coefficient estimate for DISCBD indicates that foreign logistics firms are less likely to choose more peripheral districts. In this research, DISCBD was used as a proxy for land price, but it may capture some other effects (such as agglomeration economies and facilitation of face-to-face contact), which makes the CBD attractive to service firms. The estimation results show that foreign logistics establishments favour districts close to the CBD. It implies that although land cost in the central city area is higher than that in more peripheral locations, many logistics firms are still willing to bear this cost in order to take advantage of special benefits in the CBD. Another important reason is that locating near the CBD may maximise the possibility of forward linkages (or access to markets) and backward linkages [Bodenam, 1998]. This result is quite consistent with that of a recent study [Ozdemir, 2002], which found that most foreign transport firms located in the traditional central city area in Istanbul. The results also support the proposition that logistics firms are sensitive to transport conditions. Six variables (SEADUM, ROADUM, RAIDUM, DISPORT, DISPDAIR, DISHQAIR) were used to measure transport linkage in a district. Table 3 shows that accessibility to transportation infrastructure is a significant attraction to foreign logistics investors. The coefficient estimate of SEADUM ranges from 0.99 to 2.57, suggesting that proximity to a seaport is an important consideration when foreign logistics firms make their location choices. Accessibility to an airport also attracts logistics establishments, as indicated by the coefficient estimates for DISPDAIR and DISHQAIR. The above estimation results are consistent with our observation that many logistics establishments cluster near transport nodes. The effect of land availability and cost is captured by LPOPDEN, the natural logarithm of population density in each district. The variable is found to have a negative and statistically significant impact. This result is consistent with our proposition that higher population density is normally associated with higher land cost, which deters the entry of foreign logistics investment. Another variable, LALLEMP, was used to measure the market demand for logistics services in a district. As expected, it is a significant attraction to foreign logistics investors, with the elasticity ranging

348

THE S ERVICE INDUS TRIES JOURNAL TABLE 3 NESTED LOGIT ESTIMATION RESULTS

Specification DISCBD SEADUM

(1)

(2)




20.13 (29.46)


0.99 (6.14)

(3)

20.11 (26.15)


1.47 (6.24)

ROADUM




RAIDUM

0.87 (3.37) 0.01 (0.33)

DISPDAIR

(4)






DISHQAIR






2.57 (9.61)

0.78 (2.5)



20.02 (22.14)
20.04 (21.76)

DISPORT



20.70 (22.25)


1.22 (6.37)

LPOPDEN LALLEMP

1.32 (5.8)






HOTEL






21.49 (26.15)


0.30 (12.11)

FINANCE

Observations  Alternatives Log-likelihood r 2 Adj. a Inclusive value: Centre Other

20.004 (20.02)









20.08 (25.04)






0.21 (6.25)

1.38 (5.33) 0.009 (0.06)


20.99 (23.80)


0.35 (10.78)

600  18 21108 0.360

0.75 (2.37)


0.71 (2.76)

600  18 21077 0.379


0.39 (2.60)


0.55 (2.80)

LPRI NTZ






1.29 (5.5)


2.87 (8.79) 20.32 (21.10)






0.02 (4.88) 600  18 21077 0.379


0.36 (2.60)


0.63 (2.93)

600  18 21136 0.344
0.37 (1.81)


0.83 (2.72)

Notes: Dependent variable: LOCATION. t-statistics are in parentheses.





, , and denote significance at the 10%, 5% and 1% level respectively. a The adjusted likelihood ratio index (r 2 Adj.) equals 1 2[L(M) 2 k]/L(0), where L(M) is the model log-likelihood value, k is the number of parameters, and L(0) is the log-likelihood value with all coefficients equal to zero in the model.

from 1.22 to 1.38. Thus, the denser the industrial and commercial activities are, the more likely foreign logistics investment is to occur. This result is consistent with most previous findings [e.g., Coffey et al., 1996; Aguilera, 2003; He, 2003]. We used the percentage of private employment over state employment (LPRI) as a proxy for logistics outsourcing probability. Table 3 shows that it has no significant impact on location decisions of logistics firms. Surprisingly, Table 3 shows that foreign logistics investors respond negatively to NTZ, which is contrary to previous research results [e.g., Wu, 2000]. This may be caused by high congestion costs in the special zones due to a mass inflow of foreign investments. The dummy variable cannot determine how foreign investors

349

FOREIGN LOGISTICS SERVICES IN SHANGHAI, CHINA

respond to some specific preferential treatments, such as tax rate or land use. Future research needs to introduce variables that describe specific policy incentives and to investigate their influences when data are available. We also used two variables to describe the availability of supporting services by the number of ‘foreign concern’ hotels (HOTEL) and the number of foreign-owned financial companies (FINANCE) in a district. As indicated by Table 3, both coefficient estimates are positive and statistically significant, confirming the influence of supporting services. This result is partially consistent with that of Wu [2000], which found that the availability of hotels was an important consideration when foreign investors chose a location in metropolitan Guangzhou.

Districts’ Attractiveness to Foreign Logistics Providers Attractiveness here is defined as the probability that a district is chosen by a representative foreign logistics service provider. All districts’ attractiveness in 2001 based on specification 46 is reported in Table 4. It indicates that Pudong district attracts about 57 per cent of all foreign logistics firms, followed by Hongkou, Changning and Huangpu district. The top four districts attract more than 80 per cent of all foreign logistics firms in the city. It implies strong agglomeration economies7 of foreign logistics services, and suggests that foreign logistics investors favour locations with existing concentrations of logistics activities. This result is consistent with that of a recent study [Hong, 2007]. It indicates that in spite of recent advances in telecommunication technologies, agglomeration economies still have an important impact on intrametropolitan location [Coffey and Shearmur, 2002]. The variation in the central and peripheral area’s attractiveness during the period of study is given in Figure 1. Although the estimation results in Table 3 reveal that foreign logistics firms are more likely to locate close to the CBD, Figure 1 indicates that the attractiveness of central area has decreased since 1994. Further analysis shows that the attractiveness of most districts in the central area has decreased since the mid-1990s, thereby agreeing with previous studies that observed faster growth of services in non-central area [e.g., Harrington and Campbell, 1997]. Possible forces behind the suburbanisation of logistics services include: improvements of transport linkage in peripheral areas, outward movement of some service clients, and the development of sub-centres [Shearmur and Alvergne, 2002; Aguilera, 2003]. TABLE 4 DISTRICTS’ ATTRACTIVENESS TO FOREIGN LOGISTICS FIRMS IN 2001

District Pudong Hongkou Changning Huangpu Xuhui Baoshan

Attractiveness

District

Attractiveness

District

Attractiveness

0.570486 0.108933 0.065426 0.06433 0.037905 0.031582

Minghang Jiading Luwan Jin’an Yangpu Zhabei

0.029552 0.028567 0.022418 0.014411 0.009173 0.005499

Qingpu Putuo Songjiang Nanhui Fenxian Jinshan

0.004353 0.003712 0.003287 0.000244 0.000119 0.000004

350

THE S ERVICE INDUS TRIES JOURNAL

FIGURE 1 VARIATION OF ATTRACTIVENESS TO FOREIGN LOGISTICS SERVICE PROVIDERS, 1993 – 2001

Table 5 gives the growth rate of the districts’ attractiveness to foreign logistics services from 1993 to 2001. It shows that the attractiveness of all districts in the central city area have decreased during the period. Interestingly, only five peripheral districts’ attractiveness (Pudong, Minghang, Jiading, Qingpu, Songjiang) has increased, while that of other peripheral districts has decreased. The growth rate of Minghang is 184.24 per cent, followed by Pudong district (85.13 per cent). This result partially supports Coffey and Shearmur [2002], which found that the resulting service decentralization had a form of polycentricity rather than of generalized dispersion, because agglomeration economies continued to play an important role in service location. Robustness Test The robustness of our specifications is tested by applying the specifications to subsamples of the dataset. The first test concerns the possible existence of different location determinants according to the type of business. In the original sample, we included all types of logistics service providers. One can expect that a firm which offers primarily warehousing or distribution services may have a greater propensity to locate outside the CBD due to land cost factors. For our first test of robustness, we selected firms that offer primarily warehousing or distribution services as the observations. As indicated by Table 6, the minus sign for DISCBD indicates that TABLE 5 GROWTH RATE OF DISTRICTS’ ATTRACTIVENESS TO FOREIGN LOGISTICS FIRMS, 1993 2 2001

District Pudong Hongkou Changning Huangpu Xuhui Baoshan

Growth rate (%)

District

Growth rate (%)

District

Growth rate (%)

85.13 218.19 23.72 219.53 258.89 282.52

Minghang Jiading Luwan Jin’an Yangpu Zhabei

184.24 17.53 231.62 246.06 260.86 220.15

Qingpu Putuo Songjiang Nanhui Fenxian Jinshan

30.61 243.70 20.98 258.17 232.90 246.69

351

FOREIGN LOGISTICS SERVICES IN SHANGHAI, CHINA TABLE 6 ESTIMATION RESULTS FOR FIRMS THAT OFFER PRIMARILY WAREHOUSING OR DISTRIBUTION SERVICES

Specification DISCBD SEADUM

(1)

(2)




20.10 (25.30)


1.10 (5.21)

(3)

20.10 (24.25)


1.58 (5.12)

ROADUM




RAIDUM

1.00 (3.05) 0.02 (0.8)

DISPDAIR

(4)






DISHQAIR






2.49 (7.3) 0.33 (0.87)

0.01 (0.72) 20.03 (21.06) 20.48 (21.28)


0.91 (3.20)

LPOPDEN




1.42 (4.59)

HOTEL






22.14 (26.34)


0.34 (10.68)

0.24 (5.79)

FINANCE

Observations  Alternatives Log-likelihood r2 Adj.a Inclusive value: Centre

252  18 2566 0.222 20.28 (20.43) 20.31 (20.49)

252  18 2531 0.269 0.05 (0.18) 0.03 (0.08)

Other




20.71 (21.94)









1.34 (3.79) 0.10 (0.46)


21.95 (25.03)


0.39 (9.19)

LPRI NTZ






20.09 (23.86)

DISPORT

LALLEMP






1.24 (3.96)


2.99 (6.24) 20.27 (20.66)






0.02 (2.81) 252  18 2541 0.256 0.02 (0.07) 20.02 (20.05)

252  18 2592 0.185 0.64 (1.46) 1.08 (1.36)

Notes: Dependent variable: LOCATION. t-statistics are in parentheses.





, , and denote significance at the 10%, 5% and 1% level respectively. a The adjusted likelihood ratio index (r 2 Adj.) equals to 1 2 [L(M) 2 k]/L(0), where L(M) is the model log-likelihood value, k is the number of parameters, and L(0) is the log-likelihood value with all coefficients equal to zero in the model.

the selected logistics service providers still favour sites close to the CBD. One possible reason is that the variable of DISCBD may capture some other effects, such as good transport linkage and agglomeration economies, which makes districts near the CBD attractive to foreign logistics services. The estimation results also show that several variables (such as DISPDAIR, DISHQAIR) lose some significance. Further analysis shows that this is partially due to a smaller number of observations in the subsample. Another concern is that some logistics activities are not location independent. For instance, air travel (or waterway) service providers may have to locate close to an

352

THE S ERVICE INDUS TRIES JOURNAL

airport (or seaport). To deal with this concern, we constructed a subsample that excludes this type of services. Finally, we attempt to control for firm size by constructing a new subsample which consists only of firms with more than 30 employees.8 The regression results based on the above two subsamples9 indicate that coefficient estimates of explanatory variables are quite consistent with the original results in Table 3, although several variables lose their significance due to fewer observations. The consistency of the sign and statistical significance of most coefficients lead us to conclude that the estimate results are quite robust.

CONCLUSIONS AND IMPLICATIONS

This paper confirms the hypothesis that location decisions of logistics service providers within a metropolitan area are a function of locational variables that can impact firms’ profit. Based on a nested logit model, the empirical results showed that good transport linkage and big market size attracted foreign logistics services. Logistics establishments avoided districts with high population density but valued the availability of supporting services. The privatization of the local economy was found to be an unimportant factor. Quite surprisingly, foreign logistics firms avoided special zones, such as new economic and technological zones. The empirical evidence showed strong spatial concentration of logistics services. More than half of the foreign logistics firms in Shanghai clustered in Pudong district. The uneven distribution of services may result from agglomeration economies, and certain site factors that are not pervasive in all districts (such as transport conditions). We also observed that foreign logistics service providers were more likely to locate close to the CBD. However, the attractiveness of central city area had decreased while that of the peripheral area had increased steadily since the mid-1990s. The logistics policies of the municipal government currently focus on the construction of hardware infrastructure, such as transport. Our estimation results implied that foreign logistics providers considered not only hardware facilities, but also software factors (such as the availability of supporting services). Therefore, the importance of software factors cannot be ignored in order attracting foreign logistics firms and allocating them to designated logistics parks. Another interesting finding is that logistics firms are more likely to locate near a transport node (such as an airport or seaport). If this result is confirmed by future research, the government policy-makers should find it useful in predicting future location patterns and in analysing public policies intended to influence the future location of logistics activities. This finding supports government efforts to establish logistics parks near transport nodes as an approach to enhancing the attractiveness of logistics parks. Policy-makers should also note the increase of peripheral areas’ attractiveness to foreign logistics services. The clustering of foreign investments in non-central city areas has led to polycentric urban development in the city. For example, the convenient transport linkage has encouraged many foreign logistics firms to cluster near Shanghai seaport. This, together with the decentralized administrative structure in China, may contribute to a polycentric form of urban spatial structure.

FOREIGN LOGISTICS SERVICES IN SHANGHAI, CHINA

353

ACKNOWLEDGEMENTS The author would like to thank Dr. Ronald Goldsmith and two anonymous reviewers for making valuable suggestions and comments that have improved both the content and exposition of this paper. This research forms part of a larger project sponsored by ‘211 project’, the Ministry of Education of China, for which financial assistance is gratefully acknowledged.

NOTES 1. In China Statistical Yearbook, the ‘logistics-related’ sectors cover various modes of transportation, public warehousing, post and telecommunications. 2. US$1 ¼ 8.07 Chinese Yuan. The data are from China Statistical Yearbook (1997–2002). Inflation was controlled for. 3. The terms of ‘intrametropolitan location’ and ‘location within a city’ will be used interchangeably. In this research, both refer to location choices within the Shanghai metropolitan area, which covers the central city and peripheral areas (see Table 1). 4. For the definition and classification of producer services and business services, see Shearmur and Alvergne [2002], and Coffey and Shearmur [2002]. 5. The CBD is defined as the site where business activities cluster. The CBD of Shanghai is near the Renmin Square, Nanjing Road. 6. Specification 4 denotes the fourth model in Table 3, based on which we estimate the competitiveness of each district. In order to save space, districts’ competitiveness based on other specifications is not reported in the paper but is available upon request. 7. Agglomeration economies refer to the positive externalities resulting from the spatial concentration of existing economic activities. Logistics service providers favour the spatial concentration because it can generate scale economies in logistics intermediate inputs, the pooling of markets for skilled workers, and knowledge (or information) spillovers among logistics firms. 8. Firms that have no more than 30 employees are regarded as small firms in China. Another reason for choosing ‘30 employees’ as a split criterion is to include enough observations in the subsample. 9. In order to save space and ‘keep tables to a minimum’, the estimation results based on these two subsamples are not reported. However, both are available upon request.

REFERENCES Aguilera, A. (2003) Service relationship, market area and the intrametropolitan location of business services, The Service Industries Journal, 23(1), pp.43–58. Anas, A., Arnott, R. and Small, K. (1998) Urban spatial structure, Journal of Economic Literature, XXXVI, pp.1426–1464. Baro, E. and Soy, A. (1993) Business service location strategies in the Barcelona Metropolitan Region, The Service Industries Journal, 13(2), pp.103–18. Bodenam, J. (1998) The suburbanisation of the institutional investment advisory industry: metropolitan Philadelphia, 1983–1993, Professional Geographer, 50, pp.112–26. Bolton, J.M. and Wei, Y. (2003) Distribution and logistics in today’s China, China Business Review, 30(5), pp.8–17. Carlton, D.W. (1983) The location and employment choices of new firms: an econometric model with discrete and continuous endogenous variables, The Review of Economics and Statistics, Vol.65, pp.440–49. Coffey, W.J. and Shearmur, R.G. (2002) Agglomeration and dispersion of high-order service employment in the Montreal metropolitan region, 1981–96, Urban Studies, 39(3), pp.359–79. Coffey, W.J., Polese, M. and Drolet, R. (1996) Examining the thesis of central business district decline: evidence from the Montreal metropolitan area, Environment and Planning A, 28(10), pp.1795–814. Cuadrado-Roura, J.R. and Del Rio Gomez, C. (1992) Services and metropolitan centres: the expansion and location of business services, The Service Industries Journal, 12(1), pp.97–115. Figueiredo, O., Guimaraes, P. and Woodward, D. (2002) Modeling industrial location decision in US counties, Working Paper Series No. 18 of Universidade do Minho.

354

THE S ERVICE INDUS TRIES JOURNAL

Gorden, P. and Richardson, H. (1996) Beyond polycentricity: the dispersed metropolis, Los Angeles, 1970–1990, American Planning Association Journal, 62, pp.289–95. Guimaraes, P., Figueiredo, O. and Woodward, D. (2000) Agglomeration and the location of foreign direct investment in Portugal, Journal of Urban Economics, 47, pp.115– 35. Hansen, E.R. (1987) Industrial location choice in Sao Paulo, Brazil: a nested logit model, Regional Science and Urban Economics, 17, pp.89–108. Harrington, J.W. and Campbell, S.H. (1997) The suburbanisation of producer service employment, Growth and Change, 28, pp.335–59. He, C.F. (2002) Information costs, agglomeration economies and the location of foreign direct investment in China, Regional Studies, 36(9), pp.1029– 36. He, C.F. (2003) Location of foreign manufacturers in China: agglomeration economies and country of origin effects, Papers in Regional Science, 82, pp.351–72. Head, C.K., Ries, J.C. and Swenson, D.L. (1999) Attracting foreign manufacturing: investment promotion and agglomeration, Regional Science and Urban Economics, 29, pp.197–218. Hong, J.J. (2007) Firm-specific effects on location decisions of foreign direct investment in China’s Logistics Industry, Regional Studies, 41(5), pp.1–11. Hong, J.J., Chin, A. and Liu, B.L. (2004) Logistics outsourcing by manufacturers in China: a survey of the industry, Transportation Journal, 43(1), pp.17–25. Ihlanfeldt, K.R. and Raper, M.D. (1990) The intrametropolitan location of new office firms, Land Economics, 66(2), pp.182– 98. Leitham, S., McQuaid, R.W. and Nelson, J.D. (2000) The influence of transport on industrial location choice: a stated preference experiment, Transportation Research Part A, 34, pp.515–35. Li, L.H. (1997) The political economy of the privatization of the land market in Shanghai, Urban Studies, 34(2), pp.321–35. McFadden, D. (1978) Modeling the choice of residential location, in A. Karlgvist et al. (ed.), Spatial Interaction Theory and Residential Location, Amsterdam: North Holland. Moulaert, F. and Gallouj, C. (1995) Advanced producer services in the French space economy: decentralization at the highest level, Progress in Planning, 43, pp.139–54. Ondrich, J. and Wasylenko, M. (1993) Foreign Direct Investment in the United States, Kalamazoo, MI: W.E. Upjohn Institute for Employment Research. Oum, T.H. and Park, J.H. (2004) Multinational firms’ location preference for regional distribution centres: focus on the Northeast Asian region, Transportation Research Part E, 40, pp.101–21. Ozdemir, D. (2002) The distribution of foreign direct investments in the service sector in Istanbul, Cities, 19(4), pp.249–59. Shanghai Statistical Bureau (1994–2002) Statistical Yearbook of Shanghai, Shanghai: Shanghai Statistical Bureau Press. Shearmur, R. and Alvergne, C. (2002) Intrametropolitan patterns of high-order business service location: a comparative study of seventeen sectors in Ile-de-France, Urban Studies, 39(7), pp.1143–63. Shukla, V. and Waddell, P. (1991) Firm location and land use in discrete urban space, Regional Science and Urban Economics, 21, pp.225–53. Wernerheim, C.M. and Sharpe, C.A. (2003) High order producer services in metropolitan Canada: how footloose are they, Regional Studies, 37(5), pp.469–90. Wu, F. (2000) Modeling intrametropolitan location of foreign investment firms in a Chinese City, Urban Studies, 37(13), pp.2441– 64. Yeh, A.G.O. and Wu, F. (1996) New land development process and urban development in Chinese cities, International Journal of Urban and Regional Research, 20, pp.330–53.