The Chinese Inland-Coastal Inequality: The Role of Human Capital and the 2007–2008 Crisis Watershed

Abstract

We investigate the role of human capital (HC) in the Chinese inland-coastal inequality and, related to this, how the consequences of the 2007–2008 crisis may induce China to re-focus its development path on HC. We compare panel data analyses for two periods (1998–2008 and 2009–2017) for two diverging groups of provinces (the richer/coastal and the relatively poor/inland areas). In the first period, the economic strengths that influenced the Chinese take-off and the dualism are confirmed. However, the results show that an evolution in local economic endowments is taking place: first, HC has a more evident economic effect after the crisis only in the inland provinces; second, the development path of the inland area is changing, with an evolution towards more productive sectors which can favor higher returns to HC.

1 Introduction

Two forms of inequality affect the Chinese territory. The first form, the best known and studied, is the one between urban and rural areas (Lu and Chen 2006; Sicular et al. 2007): for instance, by 2017 in the rural areas the per capita disposable income was around 39% the one in urban areas (2017 data from National Bureau of Statistics of China). The second form of inequality, studied in this article, is between coastal and inland provinces (Yang 2002). In this case, we observe a per capita GDP (Gross Domestic Product) in the poorest provinces (Yunnan, Gansu) being less than 30% of the richest ones (Beijing, Shanghai) (2017 data from National Bureau of Statistics of China). These two types of inequality affect the national Gini index: by 2016, it was as high as 38.5 (World Bank data), positioning China above many Western countries and halfway between the BRICS countries.¹

With respect to these forms of inequality, there are at least two important qualifications. The first is that while urban–rural inequality appears to be quite stable, the coastal-inland one has risen strongly in recent years (Kanbur and Zhang 1999), thus deserving in-depth studies. The second aspect concerns the relationship between inequality and economic growth: this is indeed very controversial, with the literature being still inconclusive² (see for instance the literature review in Chen 2010).

Excessive inequality, among people or between areas, could be a limit to development due, respectively, to observed and perceived disparities inducing behaviors unfavorable to economic growth (see Barro 2000) or, as in this study, to the delay in the long-term development process of a part of the country (e.g., Capello 2016).

For the second type of inequality, we refer to the different local economic strengths that shape the diverging development paths and thus imply the inland-coastal dualism in the Chinese case. In this sense, a “positive” stimulus to change could result from a wider exogenous “negative” event.

In particular, considering the rapid evolution of the Chinese economy and the changes forced by the international crisis, we wonder if these events can lead to rethinking local development paths, fostering greater attention on the diffusion of (advanced) education. We expect that the consequences on HC may in turn have consequences on the dualism: how did these two macro areas respond to the 2007–2008 crisis, how the role of HC could have changed according to different development paths?

The focus on HC is due to the fact that the growing differences between inland and coastal areas can be explained, at least in part, by the different local endowments. In this respect, HC may play a major role and should be thoroughly investigated for at least three important aspects. First, HC is a crucial determinant of regional economic performance, in the present as well as in the past (e.g. Felice 2012). Second, a large part of the within-country inequality in income is influenced by education (Acemoglu and Dell 2010). Third, the economic contribution of HC could be more important in the less-developed areas of a country (Gitto and Mancuso 2015; Felice 2018), which means that in the case of China it should be an important lever for the “poorest”.

In addition, we expect that the response of the Chinese economy to the widespread of advanced education could still be controversial in the analyzed period. On the one hand, the rapid Chinese economic growth would presuppose the contribution of HC as a key economic strength; on the other hand, the China’s development level may not be yet able to fully exploit capabilities of educated workers as an economic strength. In fact, the diffusion of primary and middle school education was necessary for the rapid development of the Chinese economy until the early 2000s, while, more in general, the local labor force does not yet have the characteristics of those of high-income economies, characterized by high productivity and continuous innovation (Li et al. 2017a).

The focus on the crisis period is influenced by the fact that China has not suffered from the serious consequences of the 2007–2008 events observed elsewhere, despite a sharp decline in exports³ (Wen and Wu 2019). Nonetheless, we may expect a change in postcrisis economic strengths: for example, a stronger focus on a resource capable of creating innovation and value in the long run, such as HC, as opposed to an export-led strategy based on low wages. This change could favor provinces not uniformly, following differences in past and current investments in education and training.

Two comparisons can help us to answer our questions, considering the 20-years period 1998–2017. First, a comparison of panel data analysis on the richer and on the poorer provinces (the first-level administrative division) allows us to identify the differences in local economic strengths between these two contexts, by observing their effects on both the variation and on the absolute values of provincial GDP. We expect that a different role of HC (e.g., a greater effect of educated workers in the more advanced provinces) should help explain disparities in economic performance, although its effects may not be observable in the short term (Chang and Shi 2016). Second, the analysis is applied to two periods, before and after the 2007–2008 crisis (1998–2008 and 2009–2017 years). This comparison serves to studying whether or not recessionary events have affected the role of HC—proxied through enrollment in advanced education – at the local level.

It should be stressed that, thus far, the results of the literature on the relationship between local economic endowments and regional inequality in China are controversial. For example, concerning the 1982–2005 years, Hao and Wei (2009) found that physical capital intensity and total factor productivity are relevant in regional income inequality, while a small role is played by HC. On the contrary, Wang and Yao (2003) found a relevant economic role and a huge potential (Wang and Yao 2003: for the 1978–1999 period) and an increasing role in fostering economic development (Zhang and Zhuang 2011; 1997–2006 period) for HC.

So far, however, in the studies about the evolution of the Chinese regional dualism two key aspects are still missing: the role at the local level of other conditioning factors, those characterizing the inland/coastal divide, and the consequences of the 2007–2009 international crisis.

In this sense, our article contributes to the literature in proposing an “extended” framework of the policies to improve advanced HC—which are, to our knowledge, those more useful for economic development (those with expected higher returns, Barro and Lee 2013). Our measure of HC includes—in addition to the well-established indicator of enrollment in higher education (e.g., Asteriou and Agiomirgianakis 2001)—also the level of enrollment in senior secondary education (i.e., the second level of secondary education, after the end of compulsory school, see Bush et al. 1998).

In studying the Chinese economy (e.g., Tuan et al. 2009), the secondary and tertiary enrollment levels have been aggregated and tested as HC proxies. Our new proxy, however, represents an extended level of advanced HC. It could have an important role for a fast-growing economy, which is not yet able to take full advantage of the so-called knowledge workers, as the Western countries are.

Finally, we take into account the endogeneity issue that stems from the reverse causality between HC and GDP, and we also test for the presence of spatial effects among provinces.

The paper is organized as follows. In Sect. 2, we present an outline of China’s modern economic development, taking into account the events that could influence the differences between the coastal areas and the hinterland ones, with a focus on the role of higher education (Sect. 2.1). In Sect. 3, we explain the models used in the analysis and their application to two groups of provinces. The variables, proposed by the economic literature, are presented in Sect. 4, followed by the results in Sect. 5. In the concluding Section, we put forward some policy implications.

2 Literature Survey: Some Aspects of the Chinese Coastal-Inland Inequality

The study of inequality among Chinese provinces is usually associated with the analysis of China’s path of economic growth: namely, with the social and economic reforms beginning in 1978, that have initiated a period of strong economic growth, followed by an increase in inequality (Fan et al. 2011). The dualism between the coast and the hinterland can be observed as a classic case of the North–South problem (Williamson 1965), due to the development of a leading sector (Pred 1965) that experiences self-sustained economic growth by attracting resources and investments from “peripheral” areas, a scheme well-known in the literature (Friedmann 1966). Broadly speaking, the different endowments and structural conditions across the provinces, in addition to the different policies adopted, have contributed to the persistence of inequalities in the long run (Ho and Li 2008). In the economic history of modern China, we observe two major phases in this sense (e.g. Herrerías et al. 2011).

The initial phase of inequality was caused by the rise of the heavy industry sector during the 1950s, which has led to a sharp decline in public interest for the traditional and less productive agricultural sector, typical of the inland areas. Inequality between provinces increased strongly in those years (Kanbur and Zhang 2005). Lack of agricultural resources (the “Great Famine” at the end of the 1950s, see Chang and Wen 1997) led to reforms in the 1970s in order to increase productivity in this sector: they allowed more freedom for farmers, whose incomes increased (resulting into a temporary decrease in inequality between provinces, Wroblowský and Yin 2016). Higher productivity reduced labor demand and, at the same time, higher incomes increased the aggregate demand (Fan et al. 2011).

As known, China has left a planned economy toward a market-oriented economy with the reforms started in 1978, which created the conditions for a period of strong economic growth. The political and economic focus on maximizing trade after the 1980s, and the subsequent adaptation to market laws during the 1990s have brought benefits to coastal provinces and caused an increase of inequality with respect to the internal areas (Fan et al. 2011). This happened, despite the implementation of policies aimed at boosting the inland provinces and exploiting the different local endowments (Chen 2010). Arguably, this increase is in line with the first part of the Kuznets curve observed in the long-run pattern of regional inequality by Williamson (1965) for the early phases of economic development, rather than with a new rise of regional inequality observed, instead, for more advanced economies (Amos 1988).

In more detail, major cities and coastal provinces exploited the positive effects of international trade, given a context of increasing globalization: they invested in productivity improvements, in various sectors, thus encouraging the location of firms and the spread of knowledge and spillovers. In the coast, not only income levels but also local endowments were diverging from the internal areas. According to Hao and Wei (2010, p. 204), “globalization, decentralization and marketization” were important in the interaction with the Chinese economic strengths. Liu and Li (2006) argued that differences in local endowments were referable, in turn, to the quality of HC, the allocation of physical capital, and the levels of available technology while Tsui (2007) focuses on the role of productivity differences at the local level; the persistence of inequalities can also be due to the fact that often the location of firms is related to the productive specializations already present in each local area (He et al. 2018).

The financial crisis that hit several Asian countries in the 1990s has prompted the central government to promote various forms of investments in the internal provinces, for several years (Yao 2009). However, obstacles to a new phase of convergence were posed by the central and local governments’ goal to maximize economic growth, with poor interest on resource redistribution in order to reduce inequality, as far as it could affect their main goal—although it is worth mentioning that, particularly in the Chinese case, there seems to be no negative effects of inequality reduction on economic growth, in the long run (Chen 2010).

2.1 The Role of Higher Education in China

In the broader context described above, the challenging role of HC in contributing to economic growth takes great relevance.

China has made considerable efforts to increase HC as a strategic economic resource. For example, Li et al. (2013)⁴ estimate an average increase of 6.72% of national HC for the years 1985–2008 (with acceleration after 1994), and this is mainly due to education and urbanization processes, affecting in particular the less developed regions (Fraumeni et al. 2019). Investment in education, therefore, has a strategic function that goes beyond economic growth: provided that it is possible to reduce the unequal access to education, especially higher education (Heckman and Yi 2012), it is possible to reduce inequalities between areas (Fleisher et al. 2010).

The effect of HC has progressively increased; it has grown together with the decisive role of international trade, thanks to economic and political reforms. Since the 1950s, economic growth has remarkably influenced the development of policies related to education; however, it is from the 1970s onwards that HC has shown the first noteworthy signs of contribution to economic development (Tsen 2006). In particular, in the following decade Chinese big firms, more autonomous, have improved and exploited the rate of return to HC; however, the average productivity level in China was still significantly lower than that of advanced economies (Meng and Kidd 1997). The same trend continued during the 1990s, though positive effects of educated workers were beginning to be recorded (Heckman 2005). Education-related productivity was increasing (Wang and Yao 2003) although the benefits of higher education were still modest in those years—a period when changes and reforms were undergoing also in the education system, which remained under the control of the state (e.g., Yang 2000; Song and Wang 2005; Ngok 2008). It is precisely at the end of the 1990s that the Chinese government decided to increase the spread of higher education, with positive effects already in the early 2000s (Che and Zhang 2018). Since then, higher education in China has developed through increasing internationalization, openness to foreign universities and between national academic institutions (Yang 2010).

The relationship between advanced education and economic growth was among the main reasons beyond these reforms, which resulted into enrollments for millions of students which, however, did not diminish the risk of perpetrating inequalities (based on the different economic conditions and opportunities among students, Wan 2006). In particular, China has regarded postsecondary education as a key factor for future development and has committed considerable resources to increase the diffusion and improve the quality of it (Altbach 2009). Since the 2000s, the new economic and market conditions have confirmed the expectations. The spread of education has increased the productivity of knowledge workers (e.g., Fleisher et al. 2011), confirming the spread of higher education as a major source of economic growth across Chinese provinces, albeit with conflicting results regarding better living conditions, equality and sustainability (see Niu et al. 2010; Whalley and Zhao 2013; Lao and Xue 2016; Zhu et al. 2018). However, these changes pose new challenges to China, such as an organization of higher education different than in the past (e.g., by including private institutions, Wu and Zheng 2008), and capable to address new problems, such as unemployment for some types of workers (Wang and Liu 2011).

3 Model

In order to assess the extent to which HC has an effect on economic growth, we focus on the analysis of HC following an approach derived from an augmented Solow model (Mankiw et al. 1992), and we estimate the following equation:

$$ \begin{aligned} GDP\_VAR_{it} & = \beta_{1} EDUCATION_{it - 1} + \beta_{2} lnGDP\_PC_{it - 1} + \beta_{3} TRADE_{it - 1} + \beta_{4} FIXED\_CAP_{it - 1} + \beta_{5} WAYS_{it - 1} \\ & \;\; + \beta_{6} GOVERN_{it - 1} + \beta_{7} SECTOR_{it - 1} + \mu_{i} + \tau_{t} + \varepsilon_{it} \\ \end{aligned} $$

(1)

where the dependent variable is the GDP growth rate, namely its annual variation at the provincial level (at the prices of the previous year, %); as a second step, we repeat the same analysis considering the value of GDP per capita (GDP_PC) as the dependent variable. EDUCATION is proxied through HC, and the other variables are the control ones used to test the robustness of the results—in more detail, the independent variables are described in Sect. 4 (Table 1). Furthermore, in the analysis of GDP variation, we consider GDP per capita as an independent variable. In the following analysis, when we consider GDP per capita as dependent variable, we include in the independent variables also one-lagged GDP per capita.

Table 1 Data description (

Data refer to the 31 Chinese provinces (i) for two periods (t) 1998–2008 and 2009–2017. ${\beta }_{1}, \dots , {\beta }_{7}$ are the parameters estimated in Sect. 5. The source for all variables is the National Bureau of Statistics of China (our elaborations). The regression also includes a full set of time dummies ${\tau }_{t}$ which represent time-specific factors that can affect the provinces growth dynamics (i.e. macroeconomic shocks) and provincial time-invariant characteristics ${\upmu }_{\mathrm{i}}$, while ${\upvarepsilon }_{\mathrm{i},\mathrm{t}}$ is the idiosyncratic error term.

In order to perform the panel data analysis, we consider a fixed effects (FE) model where differences among the provinces are considered parametric shifts in the regression equation: in our case, this leads to more reliable results given that all the provinces are included (Wooldridge 2012). FE allows for unobserved heterogeneity when there is a small population size and the entire population is included in the analysis (Johnston and Di Nardo 1997), limiting the problems of collinearity (Baltagi 2008).

Fixed effects are commonly used in regional studies. For example, FE is used by Fleisher et al. (2010) to investigate the role of HC, among other factors, on GDP in the Chinese provinces, and by Zheng et al. (2013) to study the determinants of regional allocation of infrastructure investments among the Chinese provinces.

However, we are aware of the fact that the problem of conceptual endogeneity of many variables with respect to GDP can affect these kind of models—discussions on the endogeneity problem and possible solutions are common in the relevant literature (Barro and Lee 1994; Durlauf 2009). In our case, endogeneity is due to the fact that the GDP growth rate influences the opportunity for education, and this relationship is arguably strong in China, as stated among the others by Heckman and Yi (2012, p. 7, referring to “the current practice of having tertiary education financed mainly by families”); on the other side, a strong relationship is known between education and income, insofar as a higher level of education should guarantee a higher income, identifiable by the so-called educational returns. For this reason, in order to avoid inconsistent estimates caused by the reverse causality between provincial economic growth and the level of HC, we adopt an instrumental variable approach—two-stage least squares (2SLS) procedure—aimed at obtaining an exogenous variation in the individuals with secondary and tertiary education level by means of the first and second lags of the same variables.

Yet, the instruments need to be justified in terms of relevance and orthogonality to the error terms in Eq. (1). First, the relevance of the instruments has been verified by implementing the under-identification and weak identification tests. The former is a Lagrange multiplier (LM) test of whether the equation is identified, i.e., that the excluded instruments are “relevant”, meaning correlated with the endogenous regressors. However, when the i.i.d. assumption is dropped, as in our case where the errors are clustered by province, the LM and Wald versions of the Kleibergen and Paap (2006) rk statistics are reported instead of the Anderson LM and Cragg-Donald Wald statistics which are no longer valid. The weak identification test, instead, allows us to verify if the excluded instruments are weakly correlated with the endogenous regressors. In particular, we rely on the framework provided by Stock and Yogo (2005) who have identified critical values with which the Kleibergen-Paap Wald rk (KP) F statistic has to be compared. Finally, since the model is over-identified, we use the Hansen’s J test in order to verify the validity of the instruments.

The analysis must be applied on homogeneous groups of provinces, in particular we must define a “richer group” (high socioeconomic development and income) and a group of relatively poorer provinces representing the “inland” context. For the grouping of the provinces, Herrerias and Ordoñez (2012) found the presence of five convergence clubs based on income; if instead the productivity level is considered, two clubs can be found; alternatively, if capital intensity is considered as discriminant five other clubs are observed. These groupings, however, do not seem suitable for the study of the dualism we are expected to carry out in this work, as we are going to explain. An alternative possible partition is the “three economic belts” scheme, proposed by Li and Wei, which according to them is based on the Seventh Five-Year Plan (1986–1990) and “commonly used to analyze regional inequality in China” (Li and Wei 2010, p. 305); in this partition, the eastern group coincide with the coastal area as represented in many studies, while the central and western groups coincide with the inland area. We consider this well-established division into two groups,⁵ which has been used, among the others, by Kanbur and Zhang (2005)⁶ in their inland-coastal inequality analysis.

This division showed in Fig. 1 allows us to compare the eastern provinces (the object of specific policies and favored in trade) with an area composed by the central ones (predominantly agricultural) and the western ones (less developed but with natural resources), with the aim of analyzing the different degree of economic development (Fan 1997; Li and Wei 2010).

Fig. 1

Source: Authors’ elaborations

Clusters of the Chinese provinces.

In order to illustrate economic inequalities and differences over time, in Fig. 2 we show the values of GDP per capita values for the Chinese provinces, for the first and the last year of our analysis.

Fig. 2

Source: Authors’ elaborations on National Bureau of Statistics of China data

GDP per capita in the Chinese provinces (yuan/person), selected years.

According to the National Bureau of Statistics of China data, in 2016 the East group (12 provinces), i.e. the richest provinces in our analysis, accounts for 45.1% of the Chinese population and 57.8% of the GDP; the inland group (19 provinces) represents 54.9% of the population and the 42.2% of the national GDP.

4 Data

As discussed above, studies of regional inequality usually refer to GDP and its growth rate. The annual variation of the provincial GDP (expressed at the prices of the previous year, see GDP_VAR in Table 1) is a well-established measure of regional performance (see the literature review in Greco et al. 2018). In addition, we consider the GDP per capita (GDP_PC) both to include a sort of control dependent variable and to observe traces of convergence among the provinces (e.g., Chatterji and Dewhurst 1996).

The independent variables represent HC, but we also control for other factors affecting inland-coastal inequalities according to the literature on the Chinese case. The relevance of HC deserves insights in China because it involves very different categories of workers (e.g., rural and urban workers, see Li et al. 2017a) that often characterize distinct economic areas, and because HC (as an economic resource) is present –therefore contributing to local development—in a highly uneven manner among the Chinese provinces (Fraumeni et al. 2019). In general, education is an important economic strength at the local level (Gennaioli et al. 2013; Crespo Cuaresma et al. 2014) and can contribute to reduce disparities (Sylwester 2002), and these facts represent a justification for spending on higher education (De Gregorio and Lee 2002). Given the relationship between enrollment at various level of education and economic growth (Asteriou and Agiomirgianakis 2001), we consider the ratio between the sum of enrollment rates in senior secondary and higher education, on the one side, and the total population, on the other (EDUCATION). In our view, such a measure of education represents a commitment to favor advanced HC, although it contributes to the rise of territorial differences due to its uneven concentration (Chi 2008). On the other hand, as anticipated in Sect. 1, this resource could display greater positive effects where it is relatively more scarce.

We show the spatial distributions of our index EDUCATION and the differences between the two periods in Fig. 3. We observe that, while in the first period a more marked polarization is observed towards the coast, in the second period the concentration shifts more towards the central area (in addition to greater heterogeneity in general).

Fig. 3

Source: Authors’ elaborations on National Bureau of Statistics of China data

Spatial distribution of the HC indicator for the averages of the two periods, precrisis (a) and postcrisis (b) (darker colors correspond to higher values).

For the other variables, we may refer to Fan et al. (2011, p. 48), who state:

“The evolution of regional inequalities in China since the revolution has been influenced by the policy stances taken by the authorities […] – the share of heavy industry in gross value of total output (a measure of the bias against agriculture and China’s comparative advantage), the ratio of trade volume to total GDP and effective tariff rate (a measure of the degree of openness), and the ratio of local government expenditure to total government expenditure (a measure of fiscal decentralization). We argue below that there is a close association between these policies and regional disparities.”

In this statement are already introduced three control variables. Concerning the first one, the share of heavy industry on total GDP as a measure of the bias against agriculture and China’s comparative advantage, this indeed, in our days, would not yield a proper representation of the richest provinces, given the declining importance of heavy industries in advanced economies. For this reason, we recalculate this variable (SECTOR) as the (inverse of the) ratio between the value added of the primary sector and the value added of industry, thus considering the two sectors more important for the economic dualism (Zhao and Tang 2018a). We expect this variable to illustrate the relative weight, or “economic role”, between these two sectors that are still crucial in many provinces. Of course, a relative growth in the importance of the primary sector would be indicative of an economy still characterized by low added value.

For the second proxy variable, a measure of the degree of openness (TRADE), we have calculated the ratio of the sum of import and export on GDP, which is usually considered to be representative of trade openness and economic strength in international markets (see Frankel and Romer 1999). While representing one of the major Chinese economic strengths, trade, and the export in particular, contributes in a highly uneven way to the wealth of the Chinese provinces (Perkins 1997).

For the third variable, a measure of fiscal decentralization (GOVERN), we have calculated the ratio between “Local Governments General Budgetary Expenditure” and “Government Consumption Expenditure”, as a proxy of the strengths of two hierarchical levels of government. The latter variable refers to the fact that fiscal decentralization has forced local governments to cope with health, social and educational issues (Fan et al. 2011)—where also we observe strong inequalities (see Sect. 2).

Apart from these three control variables, we must consider that the infrastructure development is one of the main Chinese strengths (see the literature review and a case study on China by Bai and Qian 2010), which favors business activities and attracts foreign investment. In this framework, the presence of transport routes across China acquires great importance in economic studies (Banerjee et al. 2020). For example, the inadequacy of transport infrastructure has particularly affected the development of the western areas (He and Duchin 2009). Among the others, Fleisher et al. (2010) consider the ratio roads/area (km length per km²) as the key infrastructure in a study on HC and regional inequality in China. We consider a measure of the local transport infrastructures (WAYS), proxied by the sum of highways, railways, and waterways (calculated on the provincial area) to represent all local transports infrastructures.

Finally, we consider the well-established indicator “gross fixed capital formation” to GDP ratio (FIXED_CAP) (for the Chinese case see Ding and Knight 2011), as a measure of fixed asset investments (see the literature review by Zhang 2008).

In Table 1, we describe the variables with the definitions provided by the National Bureau of Statistics of China, source of the data. Provincial data cover the period 1998–2017 for the GDP growth rate and 1997–2016 for the regressors.

In Table 2, we present the descriptive statistics for the two groups of provinces and for China.

Table 2 Summary statistics

Table 2 suggests that despite a particularly higher average income in the coastal area, the average GDP growth rates by area are not very different, probably due to the postcrisis events that have hit the coastal provinces more exposed to the exogenous effects of the “imported” crisis. In the period considered, even the differences in enrollment are not very marked. On the contrary, the values of two important macroeconomic measure such as the GDP per capita and trade show evident differences, typical of a dualism.

5 Results

The following Tables 3, 4 offer detailed 2SLS results and tests for the two macro areas also separating the years in which we expect a change due to the post-2008 economic events.

Table 3 2SLS results for the precrisis period (1998–2008) (GDP growth rate)

Table 4 2SLS results for the postcrisis period (2009–2017) (GDP growth rate)

In Table 3, the precrisis results suggest some similarities between the two areas. Both areas show a positive effect of (our proxy of) HC and fixed assets investments. The higher level of economic development of the coastal area makes it possible to exploit the more the spread of advanced education. Similarly, this area obtains a greater benefit from the increase in fixed investments, confirming investment in physical assets as a well-established factor linked to Chinese economic development (e.g., Ding and Knight 2011). For the inland area, our results are in line with Arayama and Miyoshi (2004), which have found that HC has been a decisive factor for the growth of the central/internal area during the past decades, together with the increased productivity associated with it. However, differences in the strength of these effects seem to be in favor of increasing economic dualism.

Concerning the level of (our proxy of) income, higher values (only for the coast) in the previous period tend to decrease the variations of provincial GDP. This could indicate a slowdown (at least partially, related to one area) in the process of growing economic gap.

Other variables showing divergent effects between areas suggest a growth of the dualism in the 1998–2008 period. The growth of trade has positive effects for the coast, since trade openness and export in particular represent an economic strength especially for the provinces most open to international trade (Perkins 1997). The effect is unexpectedly negative for the inland provinces. This difference could be explained by the strongest influence of the outlet to the sea, which might also help explain the non-relevance of the availability of other transport ways. In this respect, Wang et al. (2021) explain the current level of entrepreneurship in some areas of China precisely with the historic presence (for many decades) of ports. Another cause of the non-relevance can be related to the excessive investment in infrastructure, which may be less useful due to the poor management (Ansar et al. 2016).

The last two control variables also mark a divergent development path in the precrisis period. The greater possibilities of exploiting at a local level (see Zhang 2020) resources, such as taxation, seem to be useful in the less developed provinces. The increase in the weight of the primary sector (in proportion to the secondary sector) would, as expected, lead to a reduction in GDP variations in the wealthiest regions, while this does not happen in the inland area. In some provinces of the latter, the primary sector maintains a greater economic role, and it is still influenced by the strong improvements recorded in the 1990s linked to market incentives and greater autonomy for state-owned companies and for local governments (Bosworth and Collins 2008).

Table 4 shows that China has changed some fundamental factors of development after the crisis, and some of the new strengths could be useful to foster development paths of inland-coastal desirable convergence, especially by favoring the relatively poor provinces. In general, we observe a reduced but still positive contribution of HC in China after the crisis, but with territorial differences (for example, see Zhu et al. 2018 on central China). Other aspects of economic development have changed.

A primary feature is international trade, that has lost part of its strength in some years of the second period⁷ as observed in the result for both areas. In this framework, several development opportunities have changed for all the provinces. First, EDUCATION shows a stable economic role for the inland area, while becoming non relevant for the development of the coast. We must consider that the positive effect of increasing levels of education tends to run out where education has developed earlier or faster (see for instance Di Liberto 2008, on the Italian case), that is, in the coastal area. This could indicate the presence of a currently “equal or even more productive” resource in the relatively poorer provinces that could help counter the effects of the crisis, and encourage convergence in the long run. Secondly, the role of investments in fixed capital persists as an economic strength only for the coast, while it becomes non-significant for the other provinces, as well as for China as a whole. However, the effect of transport routes begins to play a distinctive positive role for the economic growth of the internal provinces, probably because, even though the outlet to the sea retains its primary importance, specific plans for the expansion of transport ways and infrastructures connected with the inland are one of the government’s objectives (e.g., Qin 2016). In addition, we propose in “Appendix” (Tables 8 and 9) the results with the variable WAYS divided into its three components (see the definition in Table 1). We observe that the positive contribution for the inland provinces derives precisely from the development of the waterways (even in the first period) and from the highways. The coast enjoys the positive effect of the waterways only in the first period, while the development of the other two transport networks becomes negative after the crisis. These effects could confirm the usefulness of such infrastructures in the less equipped provinces, even though they are already on the path to economic growth.

Other important changes in the economic structure are evident. The division of power between hierarchical levels of government reverses the sign (with a very low coefficient) becoming almost irrelevant also in the internal provinces, which therefore now conform to other provinces for what concerns the opportunity to exploit public resources. The inland area finally begins to suffer negatively from an excessive weight of the low productive primary sector,⁸ also due to market forces which diminish the relevance of the agricultural sector, after the strong improvements during the 1990s (Zhao and Tang 2018b). Our finding gains in importance in light of the results of Arayama and Miyoshi (2004), who observed a substantial contribution of human capital in the western and central provinces in the years preceding ours. In this framework, they argued precisely the need to increase the role of education in central and western provinces with the aim of increasing the employment share of the secondary industry.

In order to strengthen our results, after analyzing the macroeconomic effects on GDP variations, we propose an analysis on GDP per capita in Tables 5 and 6.

Table 5 2SLS results for the precrisis period (1998–2008) (GDP_PC)

Table 6 2SLS results for the postcrisis period (2009–2017) (GDP_PC)

We observe that the value of GDP per capita is persistent over time, which means that its level of the previous period positively influences the current level. This effect is observed for all areas and is reduced in the postcrisis period. The effect of HC is similar to that recorded in Tables 3, 4 for the precrisis period, while it becomes more relevant in the second period, for all provinces. The observed effect is probably due to the growing role of productivity (via advanced education) in determining income (of which our dependent variable is a proxy).

About the control variables, a remarkable difference is observed in TRADE during the first period, as this variable positively influences the proxy of the Chinese average income in the inland area. Other control variables approximately confirm the differences between the two areas.

From the tests, we can observe that our estimates for the two areas are robust. First, the under-identification test indicates that the null hypothesis can be rejected, confirming that the excluded instruments are “relevant”, i.e. correlated with the endogenous regressor. Moreover, as the KP F-statistic values are higher than the critical values reported by Stock and Yogo (2005), we can argue that the instruments are strongly correlated with the endogenous regressors. The validity of moment conditions is confirmed by the Hansen’s J test as the null hypothesis cannot be rejected at 10% (i.e. the instruments are valid). In addition, in order to verify the cross-sectional dependence in fixed effects panel data models, we implement the Pesaran’s test (2004) which is a suitable statistic for balanced panels with small T and large N. Since the null hypothesis of cross-sectional independence cannot be rejected at 10%, we can argue that there is a substantial absence of spatial dependence in data. The tests are presented in the respective tables.

We have also investigated spatial dependence among Chinese provinces by estimating a spatial Durbin model (SDM) that includes both individual and time-period fixed effects (Table 7). The model is estimated through a Maximum Likelihood (ML; Yu et al. 2008) and a row-standardized weights matrix W is used to define geographical interactions.

Table 7 Spatial Durbin results for China (precrisis and postcrisis)

The main findings reported in Table 7 are substantially similar (both quantitatively and qualitatively) to the 2SLS results, confirming the validity and robustness of our estimates. Interesting and novel elements, instead, seem to emerge when we consider the advantages that stem from the integration among provinces; that is, the possibility of benefiting from the spatial proximity to tangible and intangible factors present in a neighbouring province in the form of growth spillovers, human capital externalities and infrastructure investment, as demonstrated by the positive and statistically significant coefficients associated to these variables.

6 Conclusions

Following the increased post-reform inequalities, one of the Chinese government’s objectives has been the «harmonious development» of all the provinces, with particular attention to the internal and rural areas (Fan et al. 2011). Our research suggests that, after the watershed of the international 2007–2008 crisis, important changes are taking place and could help reduce coastal-inland gaps. First, advanced education can actually be a factor of economic convergence between provinces, because after the crisis it shows positive effects where it is relatively more scarce, i.e. in the inland area. This opportunity means obtaining greater gains from education, where more possibilities of expanding this resource are present (see Zhu et al 2018 on central China). Second, we observe important changes in the economic characteristics of the less developed internal provinces: our findings suggest an evolution towards more productive economic sectors, as well as a greater ability to exploit both investments in fixed capital and the available infrastructures.

All these aspects related to a different possibility of exploiting HC—also in connection with the general economic development—lead us to assume a better reaction (i.e. resilience) to the crisis effects of the less developed provinces.

The comparison between the decade following the advent of the exogenous crisis and the previous decade illustrates how the weakening of some foundations of Chinese development (even if partial and temporary, e.g., exports) have damaged the most exposed provinces. On the contrary, the landlocked provinces have learned to exploit those resources that the other provinces already did (e.g., to focus on more productive economic sectors, transport infrastructure, HC), and even to do better after the economic shock.

The hinterland recovery can effectively be based on the development of the advanced and more productive HC, but policy makers must consider some limits to be addressed. We propose three goals for the proper development of HC.

First, our proxy of advanced HC confirms it as a key factor for regional growth in China (e.g. Fleisher et al. 2010), implying that obstacles to the access to advanced education as those of some rural areas (Heckman and Yi 2012), and for the offspring of underprivileged families (Qin et al. 2016), could hinder the inter-provincial and intra-provincial convergences.⁹

Second, despite the efforts already underway, in the future expanding education must pursued at all levels in China (Mohrman 2008). All levels of education are needed for long-term development, and especially higher education. In fact, as Li and Wang (2018, p. 309) write on the Chinese provinces: “[…] different types of human capital affect economic growth through different channels: Basic human capital contributes to growth via the “factor-accumulation channel” and advanced human capital via the “productivity channel”, both individually and simultaneously”.

Third, better targeting and “exploitation” of advanced HC is also useful for obtaining a higher correlation between the higher investments in advanced education and the growth in labor productivity.¹⁰ Public interventions must consider that spillovers of high skilled individuals (e.g., innovation) could take long periods to be effective, contrarily to those of individuals with lower education levels (see Chang and Shi 2016).

The reassessment of the role of advanced education and its appropriate exploitation can therefore be a strength for reducing the coastal-inland dualism after the crisis, reversing the previous results. This effect can also happen thanks to important industrial policy programs and to the broader structural reforms carried out in the first period (see Brondino 2019).

Finally, we propose a brief evolution of a scenario that Chinese policy makers might expect. A reduced importance of the agricultural sector could reduce urban–rural inequality and improve the socioeconomic condition in the internal area. The improvement in rural areas (mainly internal) should lead to the achievement of a better education among the resident population. The risk is that educated workers could choose to migrate towards more industrialized areas (Long et al. 2010) due to the high territorial polarization of innovation in the coastal provinces (Crescenzi and Rodríguez-Pose 2017), originating a so-called brain drain phenomenon. The outflow of educated workers (due to the best job opportunities in a more developed job market) could not only reduce the desirable contribution of HC development in the inland area, as highlighted in our analysis, but also influence the incentive in future enrollment levels (Suzuki and Suzuki 2016).

Against this scenario and its effects, it must be stressed, however, that the economic success of high skilled migrants can inspire the development of HC also in the places of origin, through for example the incentive to invest in education, the return migration, the creation of networks, with positive economic effects (Faggian et al. 2017).

This article presents an exploratory work; future research should consider different groups of provinces, such as the division into 3 areas (east-central-west) or the more specific division into 5 macro-regions, in order to better understand local socioeconomic characteristics and opportunities of resilience. A different grouping may also help explain the role of the HC proxy observed for the coastal provinces in the second period. Further research may also help overcome some limits to the present work, for instance by obtaining more data to test different proxies of HC, as well as by constructing a microeconomic framework that can better represent the education choices and chances.

Appendix

See Tables 8 and 9.

Table 8 2SLS results for the precrisis period (1998–2008) (GDP growth rate)

Table 9 2SLS results for the postcrisis period (2009–2017) (GDP growth rate)