Open Access

China’s spatial (dis)integration as a multiethnic paradox: what do the interprovincial data say?

China Finance and Economic Review20164:1

DOI: 10.1186/s40589-015-0025-4

Received: 15 July 2015

Accepted: 18 December 2015

Published: 6 January 2016

Abstract

Background

Compared with its surging foreign trade, China’s domestic trade growth from 2000 to 2010 had been less encouraging. Then, what are the driving forces behind the dynamic pattern of China’s domestic trade?

Methods

Using the gravity model of trade and China’s interprovincial panel data, this paper shows that the negative effect of distance-related transactions costs on interprovincial trade tends to rise from 2000 to 2010. After constructing China’s 56 ethnic groups into a single, interprovincial similarity index, I cannot find any evidence that supports the view that ethnic links should serve as a factor promoting bilateral trade.

Results

However, my estimated coefficients on 37 major ethnic groups show that both positive and negative ethnic influences on trade exist in China. Specifically, 14 ethnic groups (Lahu, Qiang, Jingpo, Tu, Mongol, Manchu, Hui, Zhuang, Dongxiang, Daur, Kirgiz, She, Maonan, and Tibetan) are found to contribute to China’s interprovincial trade, while five ethnic groups (Han, Va, Kazak, Dai, and Blang) tend to impede China’s interprovincial trade.

Conclusions

These findings will be useful for policy-makers to reappraise which of China’s ethnic groups are playing the most (least) important roles in, and to introduce the optimal informal institutions into, the promotion of interprovincial economic cooperation in China.

Keywords

Domestic trade Spatial (dis)integration Interprovincial ethnic linkage Province China

Background

The first decade of the twenty-first century was unusual to China. Promoted by its entry into the World Trade Organization (WTO) on December 11, 2001, China’s economic growth has significantly driven by its remarkable performance in foreign trade. WTO data shows that China’s exports and imports enjoyed the average annual growth rates of 18 % and 16 % from 2000 to 2010, respectively, much higher than the average annual growth rate of the global trade volume in the same period, which was only 3 %.1 In 2000, for example, China was the seventh leading exporter and eighth largest importer of merchandise trade. Since 2001, China has steadily increased its share of global manufactured exports. Notwithstanding the global reductions in trade, which resulted from the US financial crisis in 2008, China replaced Germany as the world’s largest exporting nation in 2009. In 2010, China continued to be the leading merchandise exporter (US$1.58 trillion, or 10.4 % of world exports), followed by the USA (8.4 % of world exports), Germany (8.3 % of world exports), and Japan (5.1 % of world exports).2

When looking inside China, however, one can only find less encouraging news. For example, compared with its surging foreign trade as mentioned above, which has increased by more than four (for exports) or three (for imports) times from 2000 to 2010, China’s domestic trade has only achieved a growth of 86.26 % during the same period (see Table 1). This means that China’s domestic trade—both intra-provincial and interprovincial—has only had an average annual growth rate of about 6 % from 2000 to 2010. Frankly speaking, this may not have been treated as a low figure in many other countries during that period of time. However, compared with its 16–18 % of annual foreign trade growth rate and 10 % of annual gross domestic product (GDP) growth rate from 2000 to 2010, China’s domestic trade performance can only but be labeled as “poor”.
Table 1

Changes of China’s domestic and interprovincial trade from 2000 to 2010

Provincial economy

Domestic trade (thousand tons)

Ratio of interprovincial trade (%)

2000

2010

Change (%)

2000

2010

Change (%)

Anhui

6087

12,092

98.65

56.84

47.93

−8.91

Beijing

2612

1571

−39.85

72.21

95.23

23.02

Chongqing

1613

2197

36.21

56.79

65.95

9.16

Fujian

2475

3704

49.66

46.22

53.48

7.26

Gansu

3236

6186

91.16

52.10

61.27

9.17

Guangdong

4521

7505

66.00

74.70

79.72

5.02

Guangxi

2815

6109

117.02

67.10

75.82

8.72

Guizhou

3585

7991

122.90

69.29

82.98

13.69

Hainan

311

542

74.28

NA

10.89

NA

Hebei

11,399

16,481

44.58

60.61

56.09

−4.52

Heilongjiang

12,701

16,888

32.97

54.47

49.19

−5.28

Henan

9655

13,374

38.52

78.92

71.53

−7.39

Hubei

3937

5698

44.73

62.81

66.57

3.75

Hunan

4668

5783

23.89

65.62

61.27

−4.35

Inner Mongolia

9171

37,698

311.06

69.55

77.00

7.46

Jiangsu

4076

6372

56.33

63.67

85.75

22.08

Jiangxi

2959

5376

81.68

57.92

51.95

−5.97

Jilin

5630

7674

36.31

60.55

69.53

8.98

Liaoning

12,520

18,118

44.71

34.03

29.94

−4.09

Ningxia

1782

4414

147.70

73.12

86.45

13.33

Qinghai

647

3096

378.52

81.14

61.66

−19.48

Shaanxi

3280

8836

169.39

65.95

69.09

3.15

Shandong

10,585

18,285

72.74

52.24

62.67

10.43

Shanghai

1054

959

−9.01

89.56

93.53

3.97

Shanxi

28,469

60,812

113.61

91.96

92.76

0.80

Sichuan

5516

7389

33.96

46.79

51.59

4.80

Tianjin

2004

7240

261.28

81.64

54.46

−27.18

Tibet

 

30

NA

 

100.00

NA

Xinjiang

3353

6775

102.06

73.67

74.05

0.39

Yunnan

2882

5209

80.74

59.51

67.92

8.41

Zhejiang

1929

3806

97.30

62.10

41.59

−20.51

All

165,472

308,210

86.26

65.53

69.06

5.39

(1) “Trade” only includes freight exchange via national railways. (2) NA = data are not available

Source: calculated by the author based on China Association of Communications and Transportation and the National Development and Reform Commission of the People’s Republic of China (2001, 2011)

Even worse news comes from China’s interprovincial trade performance. China’s official statistics on interprovincial trade (in terms of freight exchange via national railways) are puzzling. For example, except for China’s two peripheral territories (i.e., Hainan and Tibet) whose data are not available in 2000, the proportions of interprovincial trade to the total domestic trade have risen in only 17 provincial economies (i.e., Beijing, Jiangsu, Guizhou, Ningxia, Shandong, Gansu, Chongqing, Jilin, Guangxi, Yunnan, Inner Mongolia, Fujian, Guangdong, Sichuan, Shanghai, Hubei, and Shaanxi) from 2000 to 2010. By way of contrast, the proportions for the remaining provincial economies have either decreased (i.e., in Tianjin, Zhejiang, Qinghai, Anhui, Henan, Jiangxi, Heilongjiang, Hebei, Hunan, and Liaoning) or been kept almost unchanged (i.e., in Shanxi and Xinjiang) during the same period (see Table 1 for more details).3

Indeed, the above phenomenon is unusual, especially after the following facts are taken into account:
  1. (i)

    Since the 1990s, there has been a significant improvement of transport infrastructures (including, inter alia, the completion of various expressways and high-speed railways across the nation) in China

     
  2. (ii)

    Since 1999, the “Western Regional Development Policy” has been implemented by the Chinese central government in order to speed up the development of the western and central provinces by encouraging the economic cooperation between the East-West provinces

     
  3. (iii)

    Since 2008, and as a result of the global reductions in trade, which resulted from the US financial crisis, the Chinese government has made various efforts in order to stimulate China’s domestic consumption

     

Then, what are the driving forces behind the dynamic patterns of China’s domestic trade and how to explain its interprovincial trade puzzle?

Methods

Literature review

Past studies of the determinants of spatial economic interdependence seem controversial, or at least incomplete. According to the Heckscher–Ohlin theorem, if the two factors of production are capital and labor, countries with dissimilar levels of per capita income (or, more precisely, dissimilar capital/labor ratios) will trade more than countries with similar levels (Heckscher 1919; Ohlin 1933). However, a number of empirical results indicate that the inclusion of income level as a determinant of trade contradicts the assumptions of traditional Heckscher–Ohlin theory (e.g., Linder 1961; Deardorff 1998, p. 15). In order to fill up this gap, economists have put forward new theories that base international trade on, among others, economies of scale, market imperfections, and cross-national differences in technology (e.g., Markusen 1986; Helpman 1987; Krugman 1995).

However, past studies have raised more questions than they have answered. For example, the effects of geographical proximity on trade have not been shown to fall over time. Rather, these effects have been shown to strengthen over time for 1950–1988 (Boisso and Ferrantino 1997) and 1965–1992 (Frankel et al. 1997a). Similarly, Rauch (1999) provides no evidence that, as a result of technological innovation, declining distance-related transactions costs should have led to increased trade flows. One possibility is that these analyses exclude important explanatory variables, thereby biasing the estimates.4 To clarify related issues, it is necessary to isolate the influences of all distance-related variables on trade. In particular, the inclusion of some relevant cultural variables might allow us to gain a better understanding of the black box containing the distance-related transactions costs that affect spatial economic activities.

China has officially identified, except other unknown ethnic groups and foreigners with Chinese citizenship, 56 ethnic groups. Although the majority of China’s population is of the Han nationality (which accounts for more than 90 % of China’s total population), the non-Han ethnic groups have a population of more than 100 million. Thanks to the easing migration policy that has been implemented since the 1980s, China’s interprovincial labor flows have increased dramatically. It is noteworthy that these flows have also been conducted by people coming from the inland, ethnic-minority, areas and moving into the coastal, Han-majority areas. Consequently, China’s interprovincial ethnic networks have been enhanced. As of 2010 when the Sixth National Population Census of the People’s Republic of China was conducted, each of China’s 31 provinces has become home to almost all ethnic groups. How have these growing ethnic networks contributed to (or impeded) China’s interprovincial economic cooperation and integration?

There is a widely held view that easily observable impediments, such as transportation costs, do not adequately capture transactions costs in international trade. Trade is also reduced by hidden transaction costs associated with unobserved trade barriers.5 In addition, some studies use international panel data and find that cultural distance or dissimilarity—as proxied by, among other things, the ethnic/linguistic and religious differences across national populations—is a robust determinant of the volume of international trade (see, for example, Rauch and Trindade 2002; Noland 2005; Guiso et al. 2006; and Guo 2009, pp. 77–102).

Since the 1990s, numerous quantitative studies have examined the role that cultural factors play in international trade (e.g., Havrylyshyn and Pritchett 1991; Foroutan and Pritchett 1993; Frankel and Wei 1995; Frankel et al. 1997a; Rauch 1999; Guo 2007; Melitz 2008; Felbermayr et al. 2010). These studies used linguistic or/and religious links as one or more explanatory variables. The estimated results suggest that countries which are similar to one another have been more likely to trade with each other in the postwar period. In other words, there is evidence of cultural barriers to trade.

Indeed, trade and economic cooperation may be affected by cultural dissimilarities, as it is easier and more efficient for people with the same cultural identity (ethnicity, language, religion, or any other cultural elements) to trust and communicate each other than for those with different cultural identities. In this paper, our particular interest is to test how ethnic differences have influenced China’s interprovincial trade and economic cooperation. Even though language is an effective tool of communication and that religion can provide insights into the characteristics of a culture, we would rather select ethnicity as the explanatory variable. The rationale is that most, if not all, of China’s ethnic groups are identified in terms of either linguistic or religious traditions. Another reason lies in the fact that, in China, it is more difficult, if not impossible, to collect interprovincial panel data on linguistic and religious groups than those on ethnic groups.

The gravity model is most commonly used by international and regional economists to study trade. The classic early application of the model was by Linnemann (1966), who continued work first reported in Tinbergen (1962) and then in Pöyhönen (1963).6 Generally, a gravity model assumes that the volume of trade between any two economies will be directly proportional to the product of their economic masses (measured by GDP or GNP) and inversely proportional to the distance between them. Per capita incomes (measured by product of per capita GDPs or GNPs) have become a standard covariate in the gravity models of, for example, Eaton and Tamura (1994); Frankel et al. (1997a, b) and Rauch (1999).

The model

In this paper, our particular interest is to test how various ethnic groups have influenced China’s interprovincial trade and economic cooperation. Recent gravity equations, compared to the earlier ones, have included multilateral trade resistance (MTR) terms (e.g., Feenstra 2004; Baldwin and Taglioni 2006; Head and Mayer 2013). The MTR terms arise because in a general equilibrium model, trade flows between any two provinces not only depend on trade costs between the two provinces under consideration, but also on trade costs between all other trading pairs. However, in this paper, we intend to simplify the MTR terms. The rationale is that China’s domestic MTR terms, if they do exist, are much weaker than the international ones. To this end, we only add country-fixed effects to the gravity models. As noted by Adam and Cobham (2007), these effects can be thought of theoretically as approximations to MTR terms. The basic form of the gravity model to be used in our empirical analysis of interprovincial trade is as the following7:
$$ \begin{array}{l} \ln \left(\mathrm{TRAD}{\mathrm{E}}_{ij}\right)={\upalpha}_0+{\upalpha}_1 \ln \left(\mathrm{G}\mathrm{D}{\mathrm{P}}_i\mathrm{G}\mathrm{D}{\mathrm{P}}_j\right)+{\upalpha}_2\mathrm{lnDISTANC}{\mathrm{E}}_{ij}+{\upalpha}_3\mathrm{ADJACEN}{\mathrm{T}}_{ij}\\ {}\kern10.5em +{\upalpha}_4 \ln \left(\mathrm{GD}\mathrm{PP}{\mathrm{C}}_i\mathrm{GD}\mathrm{PP}{\mathrm{C}}_j\right)+{\upalpha}_5\mathrm{P}\mathrm{D} + \upbeta \mathrm{ETHNIC}5{6}_{ij}\end{array} $$
(1)

In Eq. (1), “ln” represents the natural logarithm; TRADE ij , measured in thousand tons, is the total freight exchange between provinces i and j. GDP i GDP j is the product of GDP (in Chinese currency) of the ith and jth provinces. DISTANCE ij represents the distance between the geographical centers of gravity of the ith and jth provinces (in kilometers).8 ADJACENT ij is a dummy variable, which takes the values of “1” for provinces i and j to have a common border and of “0” otherwise. GDPPC i GDPPC j is the product of GDP per capita (in Chinese currency) of the ith and jth provinces. PD denotes a province dummy, which takes the values of “1” for provinces to be either a mountain-dominated or an island province (Chongqing, Guizhou, Hainan, Qinghai, Sichuan, Tibet, or Yunnan—we include these provinces since they have China’s most complicated geographical conditions and therefore have the most difficulties in transportation) and of “0” otherwise.

ETHNIC56 ij represents the extent to which the ith and jth provinces are ethnically linked each other (details about the measurement will be discussed in Eq. (6) in “The data” sub-section). Note that since ETHNIC56 is a comprehensive index for all of China’s 56 ethnic groups, it can only be used to derive a general pattern of correlation between interprovincial trade and ethnic linkage. If one wants to examine the role that each ethnic group plays, the gravity model can be now written as the following:
$$ \begin{array}{l} \ln \left(\mathrm{TRAD}{\mathrm{E}}_{ij}\right)={\upalpha}_0+{\upalpha}_1 \ln \left(\mathrm{G}\mathrm{D}{\mathrm{P}}_i\mathrm{G}\mathrm{D}{\mathrm{P}}_j\right)+{\upalpha}_2\mathrm{lnDISTANC}{\mathrm{E}}_{ij}+{\upalpha}_3\mathrm{ADJACEN}{\mathrm{T}}_{ij}\\ {}\kern10.5em +{\upalpha}_4 \ln \left(\mathrm{GD}\mathrm{PP}{\mathrm{C}}_i\mathrm{GD}\mathrm{PP}{\mathrm{C}}_j\right) + {\upalpha}_5\mathrm{P}\mathrm{D} + {\displaystyle {\sum}_{\mathrm{k}=1}^{37}{\beta}_k\mathrm{ETHNI}{\mathrm{C}}_{ijk}}\end{array} $$
(2)

In Eq. (2), ETHNIC ijk represents the extent to which the kth ethnic group is linked between the ith and jth provinces (details about the measurement will be discussed in Eq. (5) in “The data” section). Only 37 ethnic groups—each with a population of less than 100,000 persons (see Appendix for more detailed information)—are included in this equation.

Theoretically, cultural dissimilarity can result in both social transactions costs (a factor directly impeding trade) and “economic complementarity” (an important factor directly facilitating trade) simultaneously. As a result, the relationship between trade and cultural similarity may follow a non-linear pattern (Guo 2004, 2009, pp. 96–101). Our interest now goes to the clarification of specific cultural groups which may have different types of influences on trade with provinces differing in income levels. To go further, this paper employ a new explanatory variable: ln(GDPPC i GDPPC j )ETHNIC56 ij . Consequently, a modified form of gravity model is written as:
$$ \begin{array}{l} \ln \left(\mathrm{TRAD}{\mathrm{E}}_{ij}\right)={\upalpha}_0+{\upalpha}_1 \ln \left(\mathrm{G}\mathrm{D}{\mathrm{P}}_i\mathrm{G}\mathrm{D}{\mathrm{P}}_j\right)+{\upalpha}_2\mathrm{lnDISTANC}{\mathrm{E}}_{ij}+{\upalpha}_3\mathrm{ADJACEN}{\mathrm{T}}_{ij}\\ {}\kern10.5em +{\upalpha}_4 \ln \left(\mathrm{GD}\mathrm{PP}{\mathrm{C}}_i\mathrm{GD}\mathrm{PP}{\mathrm{C}}_j\right)\kern0.5em \mathrm{ETHNIC}5{6}_{ij}+{\upalpha}_5\mathrm{P}\mathrm{D} + {\displaystyle {\sum}_{k=1}^{37}{\upbeta}_k\mathrm{ETHNI}{\mathrm{C}}_{ijk}}\end{array} $$
(3)

In Eq. (3), the ethnicity variable is now entered into the gravity model linearly and also as a product with the natural log of per capita GDPs. Thus, the effect of an ethnic group on interprovincial trade is now allowed to exist separately and to depend on the income levels of trading partners, measured by the natural log of their per capita GDPs. As a matter of fact, since ETHNIC56 ij can be written as Ethnic ij1 + Ethnic ij2 + … + Ethnic ijk  + … + Ethnic ij56, the non-linear effects of some, if not all ethnic variables on interprovincial trade may be derived from Eq. (3).

Specifically, as for the kth ethnic group (k = 1, 2, …, 37), if the estimated coefficients on Ethnic ijk (that is, β k ) and on ln(GDPPC i GDPPC j )ETHNIC56 ij (that is, α4) have different signs (such as α4 < 0 and β k  > 0; or α4 > 0 and β k  < 0) and are statistically significant in Eq. (3), one obtains a threshold value (ln(GDPPC i GDPPC j )* k ) by letting the first-order differential of the dependent variable (ln(TRADE ij ) with respect to Ethnic ijk be zero, which is:
$$ \ln \Big(\mathrm{GDPP}{\mathrm{C}}_i\mathrm{GDPP}{\mathrm{C}}_j{\Big)^{*}}_k = \hbox{-} {\upbeta}_k/{\upalpha}_4>0\ \left(\mathrm{with}\kern0.5em {\upalpha}_4<0\ \mathrm{and}\kern0.5em {\beta}_k>0;\ \mathrm{or}\kern0.5em {\upalpha}_4>0\ \mathrm{and}\kern0.5em {\beta}_k<0\right). $$
(4)
As for the case of α4 < 0 and β k  > 0:
  1. (i)

    If ln(GDPPC i GDPPC j ) k is smaller than ln(GDPPC i GDPPC j )* k , the kth ethnic group has a positive effect on the trade between the ith and jth provinces

     
  2. (ii)

    If ln(GDPPC i GDPPC j ) k is larger than ln(GDPPC i GDPPC j )* k , the kth ethnic group has a negative effect on the trade between the ith and jth provinces.

     

The data

The major task of this paper is to quantitatively investigate the sources for changes in China’s spatial economic integration over time. Thus, the use of the cross-sectional data from China’ s provincial economies in different years enables that the estimated results are not artifact of any particular time period and to allow for changes in coefficients. Generally, a decade-long period is appropriate for this kind of research because analysis for a shorter period would not reflect relevant social and economic changes, while significant changes in transportation and communication technologies would have to be accounted for if a longer one is used. Of course, a longer period is still more helpful if three or more sets of cross-sectional data are available. However, this would increase inevitably the costs in data collection. In this paper, after taking into account data availability, we select 2 years—2000 and 2010.

The data on interprovincial trade are cited from China Communications Yearbooks (2000 and 2010).9 China’s 2000 and 2010 provincial GDP and per capita GDP data are from China Statistical Yearbooks (NBS 2001, 2011). Unlike those of many Western democratic economies, China’s provincial capitals usually are also the economic centers of their respective provinces. To this end, the following terms are used to express China’s interprovincial geographical proximity: “distance between capitals” and “interprovincial adjacency”. Distance between capitals is represented by the distance (in kilometers) between two provinces’ capitals via national railway. The data on distance between capitals are calculated by the author based on the data released by the Ministry of Railways of the People’s Republic of China.

In this paper, a comprehensive method is used to construct interprovincial ethnic linkages. Suppose that there are k ethnic groups in both provinces i and j. If the ith and jth provinces’ population shares for the kth ethnic group are expressed by x k (it ranges between 0 and 1) and y k (it ranges between 0 and 1), respectively, the kth ethnic group’s linkage index between provinces i and j can be measured by the following formula:
$$ \mathrm{ETHNI}{\mathrm{C}}_{ijk}= \min \left({x}_k,{y}_k\right),\ \mathrm{where}\kern0.5em {x}_k\in \left(0,\ 1\right)\kern0.5em \mathrm{and}\kern0.5em {y}_k\in \left(0,\ 1\right). $$
(5)

In Eq. (5), min (•) denotes the minimization of the variables within parentheses. The data on the population shares (x k and y k ) are calculated by the author based on the data released by the Fifth and Sixth National Population Census of the People’s Republic of China (conducted at 0:00 a.m. on November 1 of 2000 and 2010, respectively).

Since there are 56 ethnic groups in China, the overall ethnic linkage between provinces i and j can be measured by the following formula:
$$ \mathrm{ETHNI}\mathrm{C}5{6}_{ij} = \mathrm{ETHNI}{{\mathrm{C}}_{ij}}_1 + \mathrm{ETHNI}{{\mathrm{C}}_{ij}}_2+ \dots + \mathrm{ETHNI}{{\mathrm{C}}_{ij}}_{56}={\displaystyle {\sum}_{k=1}^{56} \min \left({x}_k,{y}_k\right)} $$
(6)

In Equation (6), min (•) denotes the minimization of the variables within parentheses. For all k, x k (0, 1), y k (0, 1), and ∑x k  = ∑y k  = 1. Consequently, ETHNIC56 ij ranges between 0 and 1. In the extreme cases, when ETHNIC56 ij =1, provinces i and j have a common ethnic structure (i.e., for all k, x k  = y k ); when ETHNIC56 ij =0, the two provinces do not have any ethnic linkage with each other (i.e., for all k, x k (or y k ) = 0). In other words, greater values of ETHNIC56 ij indicate greater degrees of ethnic linkages between the two provinces. This formula has been used in Guo (2004; 2009, p. 89) and Noland (2005). 10

A brief statistical description of selected variables included in Eqs. (1), (2), and (3) is given in Table 2 (for 2000) and Table 3 (for 2010). The largest numbers of observations (i.e., interprovincial samples) are 465 for both 2000 and 2010. However, because the data on interprovincial trade are unavailable for the two provinces of Hainan and Tibet (including 59 province pairs) from 2000 as well as for 13 province pairs (i.e., Beijing-Hainan, Guizhou-Tibet, Hainan-Heilongjiang, Hainan-Jilin, Hainan-Liaoning, Hainan-Ningxia, Hainan-Shanghai, Hainan-Tianjin, Hainan-Tibet, Hainan-Xinjiang, Jilin-Tibet, Ningxia-Tibet, and Tibet-Yunnan) from 2010, the total numbers of observations that can be actually used for our regressions are reduced to 406 (i.e., 465−59 = 406) for 2000 and 452 (i.e., 465−13 = 452) for 2010 accordingly.
Table 2

Descriptive statistics for the data on selected variables, 2000

Variable

N

Minimum

Maximum

Mean

SD

ln(TRADE ij )

406

2.99573

11.59910

6.84499

1.36932

ln(GDP i GDP j )

465

5.73532

13.62832

10.70744

1.43231

ln(GDPPC i GDPPC j )

465

16.19674

20.00507

17.70224

0.69716

ln(DISTANCE ij )

465

4.91998

8.75037

7.47203

0.64053

ETHNIC56 ij

465

0.06160

0.99930

0.75563

0.24827

k = 1. Bai

465

0.00001

0.00532

0.00004

0.00028

k = 2. Blang

465

0.00000

0.00001

0.00000

0.00000

k = 3. Buyi

465

0.00001

0.00129

0.00005

0.00008

k = 4. Dai

465

0.00000

0.00008

0.00000

0.00001

k = 5. Daur

465

0.00000

0.00120

0.00001

0.00006

k = 6. Dong

465

0.00001

0.01331

0.00011

0.00077

k = 7. Dongxiang

465

0.00000

0.00303

0.00001

0.00015

k = 8. Gelao

465

0.00000

0.00009

0.00001

0.00001

k = 9. Han

465

0.06061

0.99682

0.74690

0.25065

k = 10. Hani

465

0.00000

0.00003

0.00001

0.00001

k = 11. Hui

465

0.00025

0.15621

0.00371

0.00904

k = 12. Jingpo

465

0.00000

0.00001

0.00000

0.00000

k = 13. Kazak

465

0.00000

0.00012

0.00000

0.00001

k = 14. Kirgiz

465

0.00000

0.00004

0.00000

0.00000

k = 15. Korean

465

0.00002

0.01072

0.00013

0.00064

k = 16. Lahu

465

0.00000

0.00002

0.00000

0.00000

k = 17. Li

465

0.00000

0.00159

0.00001

0.00007

k = 18. Lisu

465

0.00000

0.00023

0.00000

0.00001

k = 19. Manchu

465

0.00006

0.03705

0.00118

0.00443

k = 20. Maonan

465

0.00000

0.00089

0.00000

0.00004

k = 21. Miao

465

0.00004

0.03037

0.00069

0.00280

k = 22. Mongol

465

0.00008

0.01789

0.00064

0.00164

k = 23. Mulao

465

0.00000

0.00081

0.00000

0.00004

k = 24. Naxi

465

0.00000

0.00047

0.00000

0.00002

k = 25. Qiang

465

0.00000

0.00004

0.00000

0.00000

k = 26. Salar

465

0.00000

0.00047

0.00000

0.00003

k = 27. She

465

0.00000

0.00372

0.00004

0.00024

k = 28. Shui

465

0.00000

0.00035

0.00001

0.00003

k = 29. Tibetan

465

0.00002

0.22530

0.00074

0.01057

k = 30. Tu

465

0.00000

0.00121

0.00002

0.00006

k = 31. Tujia

465

0.00003

0.04172

0.00066

0.00438

k = 32. Uygur

465

0.00002

0.00027

0.00004

0.00002

k = 33. Va

465

0.00000

0.00005

0.00000

0.00000

k = 34. Xibe

465

0.00000

0.00187

0.00001

0.00009

k = 35. Yao

465

0.00000

0.01114

0.00009

0.00064

k = 36. Yi

465

0.00002

0.02577

0.00022

0.00196

k = 37. Zhuang

465

0.00005

0.02701

0.00025

0.00142

Definitions of the variables shown in this table are given in the text

N number of observations, and SD standard deviation

Table 3

Descriptive statistics for the data on selected variables, 2010

Variable

N

Minimum

Maximum

Mean

SD

ln(TRADE ij )

452

2.30259

12.62984

6.95446

1.76181

ln(GDP i GDP j )

465

8.83243

16.76316

13.75672

1.38476

ln(GDPPC i GDPPC j )

465

19.15152

22.40275

20.60043

0.62203

ln(DISTANCE ij )

465

4.91998

8.75037

7.47203

0.64053

ETHNIC56 ij

465

0.08280

0.99870

0.75979

0.24414

k = 1. Bai

465

0.00000

0.00526

0.00005

0.00027

k = 2. Blang

465

0.00000

0.00016

0.00000

0.00001

k = 3. Buyi

465

0.00001

0.00231

0.00007

0.00016

k = 4. Dai

465

0.00000

0.00011

0.00001

0.00002

k = 5. Daur

465

0.00000

0.00105

0.00001

0.00005

k = 6. Dong

465

0.00001

0.01301

0.00013

0.00076

k = 7. Dongxiang

465

0.00000

0.00282

0.00002

0.00015

k = 8. Gelao

465

0.00000

0.00033

0.00001

0.00003

k = 9. Han

465

0.08176

0.99660

0.75070

0.24660

k = 10. Hani

465

0.00000

0.00014

0.00001

0.00001

k = 11. Hui

465

0.00020

0.14827

0.00356

0.00865

k = 12. Jingpo

465

0.00000

0.00002

0.00000

0.00000

k = 13. Kazak

465

0.00000

0.00071

0.00002

0.00004

k = 14. Kirgiz

465

0.00000

0.00089

0.00000

0.00004

k = 15. Korean

465

0.00001

0.00856

0.00013

0.00058

k = 16. Lahu

465

0.00000

0.00003

0.00000

0.00000

k = 17. Li

465

0.00001

0.00396

0.00003

0.00018

k = 18. Lisu

465

0.00000

0.00026

0.00001

0.00001

k = 19. Manchu

465

0.00011

0.03156

0.00111

0.00378

k = 20. Maonan

465

0.00000

0.00080

0.00000

0.00004

k = 21. Miao

465

0.00005

0.03136

0.00088

0.00296

k = 22. Mongol

465

0.00005

0.01774

0.00056

0.00157

k = 23. Mulao

465

0.00000

0.00073

0.00001

0.00003

k = 24. Naxi

465

0.00000

0.00038

0.00000

0.00002

k = 25. Qiang

465

0.00000

0.00005

0.00001

0.00001

k = 26. Salar

465

0.00000

0.00053

0.00001

0.00003

k = 27. She

465

0.00000

0.00306

0.00004

0.00021

k = 28. Shui

465

0.00000

0.00029

0.00001

0.00003

k = 29. Tibetan

465

0.00002

0.24438

0.00080

0.01149

k = 30. Tu

465

0.00000

0.00120

0.00003

0.00007

k = 31. Tujia

465

0.00006

0.04210

0.00079

0.00438

k = 32. Uygur

465

0.00001

0.00036

0.00004

0.00003

k = 33. Va

465

0.00000

0.00006

0.00001

0.00001

k = 34. Xibe

465

0.00000

0.00158

0.00001

0.00008

k = 35. Yao

465

0.00001

0.01086

0.00011

0.00064

k = 36. Yi

465

0.00002

0.03288

0.00027

0.00220

k = 37. Zhuang

465

0.00005

0.02644

0.00031

0.00141

Definitions of the variables shown in this table are given in the text

N number of observations, and SD standard deviation

Results and discussion

The gravity models constructed in “Methods” section can be tested by using the data described in “The data” section. We first run Eq. (1) by using both the data of 2000 and 2010 (the estimated results are shown in the second and third columns of Table 4, respectively) and the pooled data (the estimated results are shown in Table 5 in which the year-fixed effects on trade are also included in regression shown in the third column).
Table 4

Basic regressions—56 ethnic groups as a single variable, 2000 and 2010

Explanatory variable

Coefficient in 2000

Coefficient in 2010

Constant

14.461a (1.462)

29.057a (2.447)

ln(GDP i GDP j )

0.647a (0.046)

0.759a (0.063)

ln(GDPPC i GDPPC j )

−0.253a (0.072)

−1.041a (0.108)

ln(DISTANCE ij )

−1.212a (0.097)

−1.377a (0.127)

ADJACENT ij

0.504a (0.149)

0.579a (0.196)

ETHNIC56 ij

−1.560a (0.324)

−1.393a (0.367)

PD

−0.350a (0.099)

−0.521a (0.134)

Coefficient of correlation (R 2)

0.644

0.586

Standard error of regression

0.823

1.141

F statistic

120.556

104.786

Sig. of regression

0.000

0.000

Number of observations

405

451

Dependent variable is the natural log of interprovincial trade. Figures within parentheses are standard errors. The variance inflation factors (VIFs) are all less than 2.5, which are not reported in the table

a”Denotes statistically significant at greater than the 1 % level

Table 5

Regressions based on pooled data—56 ethnic groups as a single variable

Explanatory variable

Coefficient (excl. year-fixed effects)

Coefficient (incl. year-fixed effects)

Constant

30.733a (1.840)

30.764a (1.839)

ln(GDP i GDP j )

0.865a (0.047)

0.862a (0.047)

ln(GDPPC i GDPPC j )

−1.131a (0.081)

−1.129a (0.081)

ln(DISTANCE ij )

−1.486a (0.095)

−1.450a (0.097)

ln(DISTANCE ij ) in 2010

−0.064b (0.031)

ADJACENT ij

0.429a (0.148)

0.466a (0.190)

ADJACENT ij in 2010

−0.065 (0.236)

ETHNIC56 ij

−1.694a (0.274)

−1.939a (0.313)

ETHNIC56 ij in 2010

0.487c (0.298)

PD

−0.587a (0.102)

−0.587a (0.101)

Coefficient of correlation (R2)

0.598

0.600

Standard error of regression

1.229

1.228

F statistic

225.620

151.267

Sig. of regression

0.000

0.000

Number of observations

856

856

Dependent variable is the natural log of interprovincial trade. The variance inflation factors (VIFs) are all less than 2.5, which are not reported in the table. Figures within parentheses are standard errors

a,b, andc”Denote statistically significant at greater than the 1, 5, and 10 % levels, respectively

Early comparative studies, using the international panel data of China and East Asia, show that geographical influence on international trade was reduced from the 1980s to 1990s (Guo 2007; 2013, p. 210). One of the major driving forces contributing to this tendency might be technological advance in transportation and communications. Intuitively, wide application of E-commerce and the declining of distance-related transactions costs have increasingly contributed to the growth of international trade. However, in this paper, the negative effect of “distance” on interprovincial trade is found to rise from 2000 to 2010. Obviously, this does not reflect China’s improvement of transport infrastructures; neither does it conform to the general pattern of international trade. The main cause of China’s interprovincial trade barriers may be the market-segmenting behavior that the Chinese provinces carry on in order to maintain social stability and maximize fiscal incomes (Poncet 2005). Undoubtedly, our finding shows a sign of China’s spatial economic disintegration during the first decade of the twenty-first century.

The estimated coefficients on “ADJACENT”, which are statistically significant, slightly increase from 2000 to 2010 (see Table 4). However, the year-fixed effect derived from the regression based on the pooled data (see the third column of Table 5) does not show any statistical significance for this kind of increase.

The estimated coefficients on “ETHNIC56” are statistically significant for 2000 and 2010. However, they are negative, suggesting that the interprovincial links of 56 ethnic groups as a whole have only but impeded China’s interprovincial economic activities. Obviously, this provides no evidence that supports the widely recognizable view that ethnic linkage index tends to encourage trade between provinces that are multiethnically linked. In fact, since the partial correlation between the natural log of TRADE and the ETHNIC56 scores yields an inverted-U shape curve for 2010 (see Fig. 1), the above estimated coefficients on ETHNIC56 may be misleading (at least for 2010).
Fig. 1

Partial correlations between trade and ethnic linkage, 2000 and 2010. Sources: Table 6 for the estimated coefficients (2010) and Appendix for ethnic population

Let us now run Eq. (2) constructed in the “Methods” section. In China’s Fifth National Population Census conducted in November 2000, only permanent populations were counted (whereas in 2010, those who had stayed at their current residences for 6 months or longer were also counted). This could affect the final estimated coefficients in 2000 (remember that the “floating” populations may sometimes play more important roles in interprovincial marketing and trade than in permanent residents). To this end, we will not use the pooled data of 2000 and 2010; instead, we will only run regressions based on the data of 2000 and 2010, respectively. The estimated results shown in Table 6 are derived by excluding the variables whose variance inflation factors (VIFs) exceed 10 (a value that is often regarded as indicating multicollinearity). These results, compared with those shown in Tables 4 and 5, can help us to better explain the diverse ethnic influences on interprovincial trade:
Table 6

The estimated coefficients on the ethnic variables defined in Eq. (2)

Explanatory variable

2000

2010

Coefficient

Standard error

Coefficient

Standard error

1. Bai

  

−192.637

341.756

2. Blang

−11,4466.708

83,123.810

−36,909.300b

17,153.46

3. Buyi

−1989.870c

1044.715

767.807

753.124

4. Dai

−17,883.500

14,948.590

−16,349.312a

5046.068

5. Daur

1432.134b

656.009

2586.816a

951.629

6. Dong

48.212

69.223

68.102

91.598

7. Dongxiang

1284.739a

488.076

1515.314a

433.379

8. Gelao

27,798.940a

9578.905

−4593.947

4339.398

9. Han

−1.350a

0.319

−1.421a

0.359

10. Hani

−24,297.300b

11,902.110

−1843.340

8839.561

11. Hui

10.359b

4.765

13.374b

6.508

12. Jingpo

245,643.800a

75461.210

60,181.957a

33170.5

13. Kazak

−9719.600

11,494.670

−10010.414a

3264.211

14. Kirgiz

−17,486.600

19,645.340

5896.574b

2606.726

15. Korean

−44.290

70.080

−91.085

103.646

16. Lahu

1913.425

34,644.910

69,627.249a

27,110.080

17. Li

−25132.900a

9002.619

−249.689

262.76

18. Lisu

8126.747

6308.788

7117.415

7282.878

19. Manchu

52.555a

11.183

48.142a

17.541

20. Maonan

2042.121c

1088.550

4115.362b

1693.773

21. Miao

2.001

23.957

−7.926

34.356

22. Mongol

35.200

26.217

132.086a

36.313

23. Mulao

    

24. Naxi

  

−2385.250

8354.333

25. Qiang

4665.429

16,645.120

35,288.611a

12,540.510

26. Salar

    

27. She

58.482

152.286

452.580c

245.120

28. Shui

  

−4178.755

2932.771

29. Tibetan

45.738

35.168

9.441b

4.825

30. Tu

−415.235

1047.778

3469.553c

2133.557

30. Tujia

−8.852

11.024

−8.700

14.283

31. Uygur

1081.944

2407.569

1386.841

1977.202

32. Va

12,915.170

13,274.480

−42,682.902a

11,427.990

33. Xibe

−65.376

411.421

−194.667

660.486

34. Yao

21.566

75.760

80.423

101.235

35. Yi

9.209

36.782

−4.697

40.519

36. Zhuang

45.252

33.230

98.107b

42.746

Coefficient of correlation (R 2)

0.745

0.645

Standard error of regression

0.725

1.095

F statistic

29.877

20.337

Sig. of regression

0.000

0.000

Number of observations

405

451

Only the coefficients on ethnic variables are included in this table. The ethnic variables whose variance inflation factors (VIFs) are larger than 10 are omitted from regressions

a,b, and “c”Denote statistically significant at greater than the 1, 5, and 10 % levels, respectively

  • As for 2000, seven ethnic groups (Jingpo, Gelao, Manchu, Hui, Dongxiang, Daur, and Maonan) have positive effects; four ethnic groups (Han, Li, Hani, and Buyi) have negative effects on interprovincial trade; and 26 ethnic variables (Bai, Blang, Dai, Dong, Kazak, Kirgiz, Korean, Lahu, Lisu, Miao, Mongol, Mulao, Naxi, Qiang, Salar, She, Shui, Tibetan, Tu, Tujia, Uygur, Va, Xibe, Yao, Yi, and Zhuang) are not found to have any significant influences on trade.

  • As for 2010, 14 ethnic groups (Lahu, Qiang, Jingpo, Tu, Mongol, Manchu, Hui, Zhuang, Dongxiang, Daur, Kirgiz, She, Maonan, and Tibetan) have positive effects on interprovincial trade; five ethnic groups (Han, Va, Kazak, Dai, and Blang) have negative effects; and 18 ethnic variables (Bai, Buyi, Dong, Gelao, Hani, Korean, Li, Lisu, Miao, Mulao, Naxi, Salar, Shui, Tujia, Uygur, Xibe, Yao, and Yi) are not found to have any significant influences on trade.

Remember that there is a negative relationship between China’s interprovincial ethnic links and trade, which can be witnessed by the negative coefficients on ETHNIC56 in Tables 4 and 5. Then why there are fewer ethnic groups with negative influences on interprovincial trade than those with positive influences? This may plausibly stem from the very fact that the Han majority whose estimated coefficients are negative for both 2000 and 2010 (see Table 6) has a much larger weight than any other ethnic minorities.

Using the estimated coefficients reported in Table 6, one may calculate each ethnic group’s contribution to interprovincial trade (the results are reported in Table 7). Here, take the Hui ethnic group as an example. As shown in Table 3, the minimum, maximum, and mean values of interprovincial ethnic links—represented by ETHNIC ij11 in Eq. (2)—are 0.00020 (i.e., the one for Jiangxi and Zhejiang), 0.14827 (i.e., the one for Ningxia and Qinghai), and 0.00356, respectively, in 2010. Given that the estimated coefficient on ETHNIC ij11 is 13.374 (shown in Table 6), the Hui’s contribution to interprovincial trade in 2010 ranges from 0.268 (that is, exp(0.00020 × 13.374) × 100–100) percent to 626.423 (that is, exp(0.14827 × 13.374) × 100–100) percent, with the mean value being 4.876 (that is, exp(0.00356 × 13.374) × 100–100) percent.
Table 7

Quantifying the ethnic groups’ contributions to interprovincial trade (%)

Ethnic group

2000

2010

Minimum value

Maximum value

Mean value

Minimum value

Maximum value

Mean value

1. Bai

      

2. Blang

   

0.000

−99.734

−2.643

3. Buyi

−2.057

−98.996

−12.583

   

4. Dai

   

−3.191

−83.380

−18.414

5. Daur

0.085

350.646

1.170

0.154

1417.056

2.123

6. Dong

      

7. Dongxiang

0.047

3668.286

2.256

0.056

7127.996

2.666

8. Gelao

6.748

882,465.122

51.102

   

9. Han

−10.450

−73.957

−63.703

−10.968

−75.736

−65.587

10. Hani

−6.241

−96.533

−22.619

   

11. Hui

0.207

364.588

3.762

0.268

626.423

4.876

12. Jingpo

0.000

6032.297

53.249

0.000

174.133

11.025

13. Kazak

   

−4.219

−99.922

−22.057

14. Kirgiz

   

0.035

19,224.296

1.813

15. Korean

      

16. Lahu

   

2.019

737.508

28.358

17. Li

−19.574

−10.000

−57.647

   

18. Lisu

      

19. Manchu

0.584

425.169

5.992

0.535

356.895

5.475

20. Maonan

0.027

412.995

0.697

0.054

2598.034

1.410

21. Miao

      

22. Mongol

   

0.626

941.393

7.657

23. Mulao

      

24. Naxi

      

25. Qiang

   

3.181

425.512

28.270

26. Salar

      

27. She

   

0.096

298.641

1.680

28. Shui

      

29. Tibetan

   

0.015

904.629

0.759

30. Tu

   

0.919

6409.243

10.314

30. Tujia

      

31. Uygur

      

32. Va

   

−6.882

−91.864

−37.559

33. Xibe

      

34. Yao

      

35. Yi

      

36. Zhuang

   

0.531

1238.236

3.037

Figures in each row, which are calculated based on Tables 2 and 6, represent percentages by which provinces that are linked by the left-side ethnic group would increase (or decrease if the figures are negative) bilateral trade as opposed to those that are not linked by the same ethnic group. Blank denotes unavailability since either the ethnic variables are omitted from the regressions or the estimated coefficients are statistically insignificant in Table 6

It must be noted that the estimated coefficients on some important ethnic minorities—such as Miao, Tibet, Uygur, Xibe, Yao, Yi and Zhuang—are statistically insignificant in either 2000 or 2010 (see Table 6). Technically, if an ethnic group is not found to exert any significant influences on China’s interprovincial trade, it may have, subject to different economic conditions, both positive and negative effects on the trade of different groups of provinces. In order to test this kind of non-linear effects, let us run Eq. (3). Since, as mentioned earlier, the quality of 2000's ethnic data is less reliable than that of 2010's, we only test the regression by using 2010's data. Unfortunately, we cannot derive more encouraging results (the estimated coefficients are not reported here).

Nevertheless, we do find that some ethnic groups (such as the Hui, the Mancu, the Mongol, and the Zhuang) have some non-linear influences on interprovincial trade. Specifically, the above regression yields not only positive coefficients on these ethnic minorities (i.e., β k  > 0) but also a negative coefficient (i.e., α4 < 0) on the interaction of ETHNIC56 and the income levels of trading partners (measured by the natural log of their per capita GDPs). However, since the threshold value—i.e., ln(GDPPC i GDPPC j )* k defined in Eq. (4)—is extremely large, the fundamental changes of these ethnic groups’ positive effects on interprovincial trade will not occur in the near future.

Conclusions

During the past decades, along with the gradual reform in the decentralization of authority (that is, transferring economic management and decision making from the central government to provincial and local governments), China’s interprovincial relations have been transformed accordingly. Naturally, the examination of the driving forces to the causes and consequences of interprovincial economic (dis)integration in China is an important taskforce not only to the economists but also to the policymakers who have concerns about their internal spatial economic efficiencies.

It has been found that ethnic Chinese (mainly encompassing the Han ethnic Chinese) networks play an important role in international trade. Rauch and Trindade (2002), for example, find that ethnic Chinese networks have a quantitatively important impact on bilateral trade through the mechanisms of market information and matching and referral services, in addition to their effect through community enforcement of sanctions that deter opportunistic behavior. Their estimated results show that for trade between countries with ethnic Chinese population shares at the levels prevailing in Southeast Asia, the smallest estimated average increase in bilateral trade in differentiated products attributable to ethnic Chinese networks is nearly 60 %.11

However, in this paper, we have not found any evidence that supports that the Han majority has played positive roles in China’s interprovincial trade. Although it seems that more in-depth investigation is still needed, we believe that our small and negative coefficients on the Han (see Table 6) stem from the very fact that the Han majority accounts for more than 90 % of China’s total population (more than 1.3 billion). A large population per se also implies a great degree of diversity or dissimilarity of its members (Alesina and Spolaore 1997, p. 1029). As a result, a common standard cannot be fully realized among different provinces’ Han people in China.

To develop this argument further, let us assume that China’s domestic trade can be divided into interprovincial and intra-provincial trade, on the one hand, and interethnic and intra-ethnic trade, on the other hand. Obviously, according to the cost of transactions, intra-provincial trade is always preferable to interprovincial trade and intra-ethnic trade is preferable to interethnic trade. Therefore, ceteris paribus, the intra-provincial and intra-ethnic trade is always more profitable than the interprovincial and interethnic trade. If the population of an ethnic group is very small, then the intra-provincial and intra-ethnic trade is not able meet the demand of economic growth. As a result, the interprovincial and intra-ethnic and intra-provincial and interethnic trade may still be needed. In this case, the interprovincial and intra-ethnic trade is preferable to the intra-provincial and interethnic trade if the interprovincial transactions cost is lower than interethnic transactions cost, and vice versa. Figure 2, in which smaller ethnic groups have greater (in both positive and negative directions) effects on interprovincial trade, has provided evidence that small ethnic groups are always more important in promoting interprovincial trade than large ethnic groups.
Fig. 2

Ethnic influence on interprovincial trade decreases with the size of ethnic population

We need to clarify the inherent forces and narratives behind the differing influences of all the ethnic minorities on China’s spatial economic (dis)integration. For example, the Han-Uyghur unrest in and outside Xinjiang would have been responsible for the Uyghurs’ insignificant effects on interprovincial trade; and Tibetans’ positive influences on interprovincial trade in 2010 has benefited from the Qinghai–Tibet railway which went into operation in 2006. With the operation of the Qinghai–Tibet railway, the costs of transportation of both passengers and goods should be greatly reduced, allowing for an increase in volume—the costs per ton-kilometer will be reduced from 0.38 yuan to 0.12 yuan (Cnradio, 10 November 2006). As a result, more commodities will be carried to and from Tibet by the railway. However, it seems that I am not able to clarify all the influences of each and all of China’s 56 ethnic groups in a single paper.

Since China adopted different approaches when conducting the population census in 2000 and 2010, our ethnic data may not be comparable from 2000 to 2010. Therefore, cares should be taken when the changes of ethnic influences on interprovincial trade from 2000 to 2010 are to be clarified. But our estimated coefficients for 2010 seem to be more reliable than those for 2000 (shown in Table 6). If the 2010’s results shown in Table 7 are correct, we may conclude that 14 ethnic groups (Lahu, Qiang, Jingpo, Tu, Mongol, Manchu, Hui, Zhuang, Dongxiang, Daur, Kirgiz, She, Maonan, and Tibetan) tend to contribute to China’s interprovincial trade, that five ethnic groups (Han, Va, Kazak, Dai, and Blang) tend to retard China’s interprovincial trade. These findings will be useful for policy makers to reappraise which of China’s ethnic groups are playing the most (least) important roles in, and to introduce the optimal informal institutions into, the promotion of interprovincial economic cooperation in China.

It must be noted that interprovincial trade may also foster the interprovincial migration of ethnic groups in China, raising an issue of potential endogeneity in the analysis of ethnic influences on interprovincial trade in this paper. However, since our ethnic data only include permanent populations and that most, if not all, seasonable, short-stay migrants have only been officially defined as the “floating populations” (liudong renkou), this kind of potential endogeneity problem does not render the estimated results biased and inconsistent.

In order to overcome the problems with multicollinearity, we have omitted a number of ethnic groups from our regressions. The general rule of thumb is that variance inflation factor (endogeneity) exceeding 4 warrant further investigation, while those exceeding 10 are signs of serious multicollinearity requiring correction (Simon 2004). In weaker models, especially in those that are not supported by large sets of data, VIF above 2.5 may also merit further investigation (Berry and Feldman 1985, p. 49; Arceneaux and Huber 2007). In this paper, we have re-run all the regressions in Table 6 by omitting the explanatory variables with VIF exceeding 4 (the estimated results are not reported here). But we have found that the estimated results are quite stable after the variables with the VIF exceeding 10 are omitted from the regressions. Thus, even though the variables with VIF exceeding 4 are included, the estimated coefficients reported in this paper are not affected by multicollinearity.

Footnotes
1

Calculated by author based on the data released by the World Trade Organization (Available at http://www.wto.org/english/thewto_e/countries_e/china_e.htm. Accessed 2013-5-20).

 
2

Data sources: Rumbaugh and Blancher (2004) and WTO (2011).

 
3

Note that Beijing and Shanghai’s reductions of domestic trade from 2000 to 2010 are mainly due to their removals of large industrial, pollution-making plants during the above period.

 
4

For a more detailed analysis of the distance puzzle, see Head and Disdier (2008).

 
5

These trade barriers take a number of forms including legal and institutional differences (Anderson and Marcouiller 2002; Linders et al. 2005; Combes et al. 2005; and Guiso et al. 2006), ethnic/linguistic fractionalization (Rauch 2001; Rauch and Trindade 2002; Melitz 2008; and Felbermayr et al. 2010), and linguistic and religious dissimilarities (Guo 2004; 2007).

 
6

The earliest application of the gravity model can be traced back to the 1940s (see, e.g., Zipf 1946; Stewart 1948). More recent summaries can be found in Baldwin and Taglioni (2006) and Head and Mayer (2013).

 
7

Since GDPPC equals GDP/POP (where POP is population), Eq. (1) can be written as ln(TRADE ij ) = α0 + (α1 + α4)ln(GDP i GDP j ) + α2lnDISTANCE ij  + α3ADJACENT ij 4ln(POP i POP j ) + α5PD + βETHNIC56 ij . However, we will not use this equation since the inclusion of GDP and POP—unlike that of GDP and GDPPC—can easily result in multicollinearity problems.

 
8

A direct measure of transport costs (instead of distance) has been suggested as a proxy for trade costs, especially for within-country studies (Combes et al. 2005). However, since we only consider the trade via national railways in our research and that the per-kilometer costs of transportation via national railways are almost fixed throughout China, these two measures are not different from each other.

 
9

They are compiled by China Association of Communications and Transportation and the National Development and Reform Commission of the PRC and published by China Communications Yearbook Press in 2001 and 2011, respectively.

 
10

Several other methods can also be used. Boisso and Ferrantino (1997), for example, use ∑x k y k as the construct of similarity index. However, Eq. (6) can prevent the index of interprovincial ethnic linkages from further reduction when the values of x k and y k are small.

 
11

More recently, there is an empirical work in which the Chinese network is found to lead to a modest amount of trade creation of only about 15 % (Felbermayr et al. 2010).

 

Declarations

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

(1)
Regional Science Association of China, Peking University

References

  1. Adam, C. and D. Cobham (2007). Modeling multilateral resistance in a gravity model with exchange rate regimes. The Centre for Dynamic Macroeconomic Analysis, University of St Andrews. Available at www.st-andrews.ac.uk/cdma. Accessed on 2014-4-8.
  2. Alesina A, Spolaore E (1997) On the number and size of nations. Q J Econ 112(Nov.):1027–1056View ArticleGoogle Scholar
  3. Anderson JE, Marcouiller D (2002) Insecurity and the pattern of trade: empirical investigation. Rev Econ Stat 84:342–352View ArticleGoogle Scholar
  4. Arceneaux K, Huber GA (2007) What to do (and not to do) with multicollinearity in state politics research. State Polit Policy Q 7(1):81–101View ArticleGoogle Scholar
  5. Baldwin R, Taglioni D (2006) Gravity for dummies and dummies for gravity equations, NBER Working Paper No. 12516View ArticleGoogle Scholar
  6. Berry WD, Feldman S (1985) Multiple regression in practice. Sage Publications, LondonGoogle Scholar
  7. Boisso D, Ferrantino M (1997) Economic distance, cultural distance and openness in international trade: empirical puzzles. J Econ Integrat 12:456–484Google Scholar
  8. China Association of Communications and Transportation and the National Development and Reform Commission of the PRC (2001) China communications yearbook 2000. China Communications Yearbook Press, BeijingGoogle Scholar
  9. China Association of Communications and Transportation and the National Development and Reform Commission of the PRC (2011) China communications yearbook 2010. China Communications Yearbook Press, BeijingGoogle Scholar
  10. Cnradio, 10 November (2006) Qingzang railway transported 0.73M passengers, boosts Tibet economy (in Chinese). China National Radio (Cnradio), Beijing. Available at www.cnradio.com.cn/2004news/internal/200611/t20061110_504324840.html. Accessed 2013/4/15.
  11. Combes P-P, Lafourcade M, Mayer T (2005) The trade-creating effects of business and social networks: evidence from France. J Int Econ 66(1):1–29View ArticleGoogle Scholar
  12. Deardorff AV (1998) Determinants of bilateral trade: does gravity work in a neoclassical world? In: Frankel JA (ed) The Regionalization of the World Economy. University of Chicago Press, Chicago, pp 7–22Google Scholar
  13. Eaton J, Tamura A (1994) Bilateralism and regionalism in Japanese and U.S. trade and direct foreign investment patterns. J Japanese Int Econ 8:478–510View ArticleGoogle Scholar
  14. Feenstra RC (2004) Advanced international trade: theory and evidence. Princeton University Press, Princeton, NJGoogle Scholar
  15. Felbermayr G, Jung B, Toubal F (2010) Ethnic networks, information, and international trade: revisiting the evidence. Annales d’Economie et de Statistique 97–98:41–70Google Scholar
  16. Foroutan F, Pritchett L (1993) Intra-Sub-Saharan African trade: is it too little? J Afr Econ 2(May):74–105Google Scholar
  17. Frankel JA, Wei S-J (1995) European integration and the regionalization of world trade and currencies: the economics and the politics. In: Eichengreen B, Frieden F, von Hagen J (eds) Monetary and fiscal policy in an integrated Europe. Springer, New York, pp 23–51Google Scholar
  18. Frankel JA, Stein E, Wei S-J (1997a) Trading blocs and the Americas: the natural, the unnatural, and the super-natural. J Dev Econ 47(1):61–95View ArticleGoogle Scholar
  19. Frankel JA, Stein E, Wei S-J (1997b) Regional trading blocs in the world economic system. Institute for International Economics, Washington, DCView ArticleGoogle Scholar
  20. Guiso L, Sapienza P, Zingales L (2006) Does culture affect economic outcomes? J Econ Perspect 20:23–48View ArticleGoogle Scholar
  21. Guo R (2004) How culture influences foreign trade: evidence from the U.S. and China. J Socio-Econ 33:785–812View ArticleGoogle Scholar
  22. Guo R (2007) Linguistic and religious influences on foreign trade: evidence from East Asia. Asian Econ J 21(1):100–121View ArticleGoogle Scholar
  23. Guo R (2009) Intercultural economic analysis: theory and method. Springer, New YorkView ArticleGoogle Scholar
  24. Guo R (2013) Understanding the Chinese economies. Academic Press/Elsevier, OxfordGoogle Scholar
  25. Havrylyshyn O, Pritchett L (1991) European trade patterns after the transition, Policy, Research and External Affairs Working Paper No. 74. The World Bank, Washington, DCGoogle Scholar
  26. Head K, Disdier A (2008) The puzzling persistence of the distance effect on bilateral trade. Rev Econ Stat 90(1):37–48View ArticleGoogle Scholar
  27. Head K, Mayer T (2013) Gravity equations: workhorse, toolkit, and cookbook, CEPR Discussion Paper 9322Google Scholar
  28. Heckscher EF (1919) The effect of foreign trade on the distribution of income. Ekonomisk Tidskrift 21:497–512Google Scholar
  29. Helpman E (1987) Imperfect competition and international trade: evidence from fourteen industrial countries. J Jpn Int Econ 1(Mar.):62–81View ArticleGoogle Scholar
  30. Krugman PR (1995) Growing world trade: causes and consequences. Brookings Pap Econ Activ 1:327–362View ArticleGoogle Scholar
  31. Linder SB (1961) An essay on trade and transformation. John Wiley and Sons, New YorkGoogle Scholar
  32. Linders G-JM, Slangen A, de Groot H, Beugelsdijk S (2005) Cultural and institutional determinants of bilateral trade flows, Tinbergen Institute Discussion Paper No. 074/3Google Scholar
  33. Linnemann H (1966) An econometric study of international trade theory. North-Holland, AmsterdamGoogle Scholar
  34. Markusen J (1986) Explaining the volume of trade: an eclectic approach. Am Econ Rev 76:1002–1011Google Scholar
  35. Melitz J (2008) Language and foreign trade. Eur Econ Rev 52(4):667–699View ArticleGoogle Scholar
  36. NBS (National Bureau of Statistics) (2001) China statistical yearbook 2000. China Statistics Publishing House, BeijingGoogle Scholar
  37. NBS (National Bureau of Statistics) (2011) China statistical yearbook 2010. China Statistics Publishing House, BeijingGoogle Scholar
  38. Noland M (2005) Affinity and international trade. Working Paper Series WP 05-3. Institute for International Economics, Washington, DCGoogle Scholar
  39. Ohlin B (1933) Interregional and international trade. Harvard University Press, Cambridge, MAGoogle Scholar
  40. Poncet S (2005) A fragmented China: measure and determinants of Chinese domestic market disintegration. Rev Int Econ 13(3):409–430View ArticleGoogle Scholar
  41. Pöyhönen P (1963) A tentative model for the volume of trade between countries. Weltwirtschaftliches Archiv 90(1):93–99Google Scholar
  42. Rauch JE (1999) Networks versus markets in international trade. J Int Econ 48(1):7–35View ArticleGoogle Scholar
  43. Rauch JE (2001) Business and social networks in international trade. J Econ Lit 39(4):1177–1203View ArticleGoogle Scholar
  44. Rauch JE, Trindade V (2002) Ethnic Chinese networks in international trade. Rev Econ Stat 84(1):116–130View ArticleGoogle Scholar
  45. Rumbaugh T, Blancher N (2004) China: international trade and WTO accession, IMF Working Paper WP/04/36. International Monetary Fund (IMF), Washington, DCGoogle Scholar
  46. Simon LJ (2004) Detecting multicollinearity using variance inflation factors. Penn State Department of Statistics, The Pennsylvania State University.
  47. Stewart JQ (1948) Demographic gravitation: evidence and applications. Sociometry 2:31–58View ArticleGoogle Scholar
  48. Tinbergen J (1962) An analysis of world trade flows, the Linder hypothesis, and exchange risk. In: Tinbergen J (ed) Shaping the world economy. The Twentieth Century Fund, New YorkGoogle Scholar
  49. WTO (2011) World Trade 2010, prospects for 2011 (Trade growth to ease in 2011 but despite 2010 record surge, crisis hangover persists), World Trade Organization (WTO) Press/628., 7 April, . Available at: http://www.wto.org/english/news_e/pres11_e/pr628_e.htm. Accessed 2013/4/5Google Scholar
  50. Zipf GK (1946) The P1P2/D hypothesis: on the intercity movement of persons. Am Sociol Rev 11(6):677–686View ArticleGoogle Scholar

Copyright

© Guo. 2016

Advertisement