Introduction

Since its transition from a planned economy to a socialist market economy in 1978, China has experienced rapid economic growth through increased integration into the world market. Though economic globalization has greatly improved Chinese people’s living standards and conditions, it has also caused extreme imbalance among regions, urban–rural areas, and industries (Han 2020). In particular, provinces and municipalities have disparately experienced the benefits of economic globalization: for example, between 1978 and 1998, real gross domestic product (GDP) in Fujian province increased by 13.9% while that of Gansu grew by only 6.7% (Q. Zhang and Zou 2012). In 2006, Beijing, Shanghai, and Tianjin were the richest areas with per capita GDP nearly 0.5 to 2.5 higher than the national average, while regions in the west, such as Gansu and Qinghai, had per capita GDP far below the national average (Zhang, 2012).

Such widening disparities both across and within provinces in China had an immense impact on the existing structures of welfare and social security provision. Implicit lifetime employment had been replaced by massive layoffs, widespread unemployment, forced early retirements, and frequent failure to provide social security benefits (Giles et al. 2006). In order to better distribute the costs of economic restructuring, the Chinese government actively advanced welfare institutions particularly since the 1990s when the state sector underwent massive labor force reduction (Frazier 2006). Efforts to provide state-led solutions had once again gained momentum since the 2008 global financial crisis, which exposed the weaknesses of the Chinese system in protecting its citizens under adverse global and domestic economic conditions (H. X. Zhang 2012). During this period, government spending on social safety net and employment effort has increased from 5447.16 million yuan in 2007 to 19,018.69 million yuan in 2015 (K. Guo and N’Diaye 2009).

However, despite China’s unprecedented expansion of welfare programs over the past decade, the provision of welfare benefits remains largely uneven across provinces. For instance, coverage rates of the urban pension system varied from approximately 20% in Guizhou to 79% in Shanghai in 2010 (HelpAge International 2013). Regions that share similar levels of economic growth in the south-east coast area also display varying levels of coverage rates: in 2010, actual coverage rates of the Basic Old-Age Insurance Scheme for Enterprise Employees ranged from 72% in Guangdong to 46% in Jiangsu (Quan 2012). Why do provinces with similar levels of economic openness provide different levels of social protection? How does increased integration to the global economy differentially affect patterns of social provision at the provinces?

Existing theories on the effects of globalization on welfare state development argue that governments may either expand welfare provision to compensate for the social costs of globalization (Cameron 1978; Rodrik, 1997) or cut costs on social spending in order to secure competitiveness in the international market (Kaufman and Segura-Ubiergo 2001; Rudra 2002). However, previous mechanisms that had been used to explain why national governments expand or retrench welfare provision as a response to globalization do not fully address the varying impact of economic openness on local economies and their subsequent provision of welfare benefits. Moreover, the compensation hypothesis tends to assume that the threat of social instability, which pressures governments to compensate via welfare expansion, is uniform within a nation. In other words, how local governments experience and respond to different levels of social instability that may also be unassociated with the social costs of globalization is largely undertheorized in the literature. For example, the threat of social unrest that local governments experience in China tends to vary widely between provinces. A record of the annual number of protests at the prefecture-level city between July 2013 and June 2016 shows that several cities (e.g., Beijing, Shenzhen, Chongqing, Chengdu, Shanghai) had experienced between 365 and 577 protests a year, which is significantly higher than other localities (e.g., Ningxia, Qinghai) that experienced 10 to 15 protests a year (Göbel 2016). The regional variation of social unrest may be attributed to various factors other than economic openness, such as internet penetration and censorship (Göbel 2016; Steinhardt 2015; King et al., 2013), local courts’ decisions on protest-supported cases (van Rooiji et al. 2012; He 2014), grassroots officials’ ability to “buy stability” via their respective stability maintenance funds (C. K. Lee and Zhang 2013), or the power structure of local leaders at different administrative levels (C. Y. Tsai 2021; O’Brien 1996).

To further our understanding on the relationship between international economic integration and regional patterns of social insurance provision in China, this paper argues that the effect of globalization on welfare provision in local economies is conditioned by the threat of social unrest. By analyzing a provincial-level dataset in China from 2001 to 2015, I find that the subnational variation of social insurance coverage depends on the interaction between provincial trade openness and the threat of social unrest. This suggests that as economic openness shapes the size of welfare provision conditional on the threat to social stability, local governments will expand social insurance coverage in order to preempt collective action or mass incidents that may arise from increased exposure to the international market. Likewise, highly economically open localities that experience low threat of social unrest may not face similar pressures to appease public discontent via welfare, and therefore retrench or decrease welfare provision in order to pursue higher economic growth. I argue that the political incentives behind welfare expansion in open economies with higher threat of social unrest at the provincial level are twofold, both based on local officials’ career prospects: first, increased labor disputes create industrial relations problems, which may pose the threat of capital flight and endanger the goal of achieving local economic growth; second, higher frequency of labor disputes signal a threat to social stability, wherein stability maintenance is emphasized as the overriding principle of the Chinese Communist Party. Local officials who fail in securing both goals of economic growth and maintaining social order are unlikely to be able to advance their careers, which incentivizes them to preempt mass incidents that threaten regime stability by expanding social protection for those negatively affected by globalization.

The remainder of the article is organized as follows. The second section provides a discussion of the compensation-efficiency debate that characterizes existing studies on globalization and the welfare state, and further examines previous scholarship on welfare development in China. Next, I provide a theoretical discussion that addresses the mechanisms of how provincial social insurance coverage may vary depending on the relationship between economic openness and the threat of social unrest. The remaining sections test the proposed hypotheses using provincial-level data from 2001 to 2015 and conclude with a discussion of the implications of the findings.

Economic Integration and Government Welfare Provision

The international integration of markets for goods and services has inspired insightful research on the mechanisms behind globalization and welfare development in both advanced industrialized economies and developing countries (Adserà and Boix 2002; Kaufman and Segura-Ubiergo 2001; Rodrik, 1997; Cameron 1978; Avelino et al., 2005; Rudra 2002, 2008). Various studies have also examined the intervening role of regime type on economic openness and social spending, highlighting the role of electoral competition (Avelino et al., 2005) or labor power (Rudra 2002).

Existing studies on the effects of economic integration on government welfare provision are divided as to how governments respond to globalization with social policies: by protecting people’s welfare against the vagaries of the international economy or by cutting costs for international competitiveness. Conventional perspectives of the compensation argument contend that governments use a variety of policy instruments to protect their economies from the competitive risks of the international economy (Cameron 1978). Increased exposure to the international market are likely to experience social instability and discontent, which creates electoral pressure to government officials to protect domestic interests by strengthening social insurance mechanisms (Avelino et al., 2005). The welfare state is thus a mechanism for compensating the social costs of globalization, wherein social spending acts as a risk-reducing instrument in open economies (Rodrik, 1997).

On the other hand, proponents of the efficiency hypothesis argue that governments will cut costs on social spending in order to secure competitiveness in the international market. High levels of social spending may be reflected in the costs of labor, which in turn will affect domestic firms’ ability to compete in the international market (Kaufman and Segura-Ubiergo 2001). Increased capital mobility adds on to the pressure to cut costs as international investors normally avoid investing in countries that spend beyond their means and also pose a higher risk of capital flight. Kaufman and Segura-Ubiergo (2001) find that social security transfers, mainly pensions, are more vulnerable to the negative effects of international economic integration, while Rudra (2002) argues that social welfare in developing countries highly endowed with low-skilled labor is also more likely to be adversely affected by globalization.

When evaluating the compensation-efficiency debate, studies have often examined the intervening role of regime type on social spending (Avelino, Brown, and Hunter 2005; Haggard and Kaufman 2008; Adserà and Boix 2002; Wibbels and Ahlquist 2011), which expect that economic integration is more likely to lead to public sector expansion in democracies that are subject to electoral competition and interest group pressures Avelino et al., 2005). In an open economy, authoritarian regimes were more likely to retrench on social programs compared to their democratic counterparts (Huber et al., 2008; Adserà and Boix 2002). Rudra and Haggard (2005) similarly find that trade openness has an adverse impact on social security and welfare spending of non-democracies.

However, neither the compensation nor the efficiency paradigm can sufficiently explain the subnational variation of social insurance coverage in China. The compensation hypothesis has been built on the OECD countries and focuses on the role of social organizations, partisan politics, and leftist institutions (Swank 2002; Rodrik 1997; Garrett 1998). According to these theories, more open economies tend to strengthen left-labor movements, which in turn result in larger demands for government transfers (Cameron 1978; Garrett 1998). When the labor sector suffers collective action problems and thus lacks bargaining power, developing countries are expected to contract social spending as they tend to emphasize competitiveness by taking advantage of surplus labor (Rudra 2002). However, China lacks a strong labor organization or party orientation comparable to that of Western Europe; though China has the highest union density in the developing world, labor has very little bargaining power (Chan and Senser 1997). Likewise, as competitive elections are not held in China, the role of electoral pressure (Cameron 1978) or interest groups (Avelino et al., 2005) cannot be comfortably applied to the Chinese case. Furthermore, both compensation and efficiency paradigms focus mostly on the effects of globalization on social welfare at the national level; the varying impact of globalization on local economies and the subsequent provision of welfare benefits has largely been underexamined.

Existing Explanations for Social Welfare Development in China

Various studies that examine the social welfare system in China have commonly identified and discussed issues of fragmentation and regional inequalities in the provision of various social policies. For example, X. Huang (2014) identifies the fragmented social health insurance system which serves the authoritarian leaders’ interests in maintaining regime stability through the consolidation of societal divisions and prevention of horizontal mobilization. Selective provision of old-age pension along the lines of urban versus rural residents, state-owned enterprises and non-state enterprises, and permanent versus migrant workers have also been examined (Freedman and Yanxia 2016; Frazier 2004).

In order to identify the main causes behind such fragmentation of China’s social welfare system, recent studies have focused on the economic and political incentives of local leaders. Under the slogan of “cooking in separate kitchens (fenzao chifan)” in the 1980s, the central state increasingly decentralized fiscal policy to local governments, which further led to the transference of the primary responsibility of public goods provision such as education, health care, infrastructure, and social security programs (Teets 2013). Because the central leaders have higher information costs and less expertise to distribute benefits to various social groups (X. Huang 2015), the discretionary power to design and implement social welfare policies has been delegated to local state agents (L. L. Tsai 2007). Local leaders are thus the main providers and sponsors of social welfare who redefine and carry out different forms of innovative social insurance projects (X. Huang 2015; Mok and Xiao Fang, 2013). Due to this decentralized nature of public goods provision, identifying local officials’ incentives and motivations behind the implementation of social policies is therefore key to understanding the subnational variation in welfare provision in China.

One such explanation of local leaders’ incentives concentrates on the economic calculus of welfare allocation. Frazier (2010) examines the historical background behind the development of public pensions as a response to uneven growth and examines local governments’ revenue-raising and expenditure assignments as key in understanding the development of pensions. Y. Huang (2004) finds that rural leaders in developing industrial sectors are more likely to provide better health care, while Solinger and Hu (2012) argue that wealthier locales provide more assistance to those viewed as unsightly or incompetent to incentivize them to stay at home, keeping them away from the regular labor market and thereby achieving a modern city appearance.

Contrary to the economic incentives regarding welfare provision, other studies have focused on the political incentives of local officials. Reflecting the main component of the sustainable development ideology set forth by the Party (Zuo 2015), the expansion of social welfare benefits and social spending has been found to affect local officials’ career trajectories (Lü and Liu 2013). Provision and expansion of welfare have also been examined from the approach of securing regime stability. Wallace (2014) attributes China’s successful maintenance of stability to redistributive policies targeted toward the countryside, along with the household registration system that effectively manages urbanization. Chen et al., (2016) find that local officials were most responsive to public requests regarding the Minimum Livelihood Guarantee (dibao) that were made with the threat of collective action. X. Huang (2015) examines the interaction between central and local incentives, in which central leaders, due to their emphasis on regime survival and stability, delegate discretionary authority to design the coverage and generosity of social health insurance to local leaders.

Based on this recent scholarship on local incentives to expand social welfare, this paper contributes to the literature on globalization and welfare provision in the context of China in the following aspects: first, this study emphasizes the interaction of both economic and political incentives that local leaders face in their jurisdictions. Previous studies have focused on either local officials’ political or economic interests, or have examined them separately from the perspective of the central versus local government. Second, contrary to extant studies that examine the role of economic integration as a secondary effect that needs to be controlled for, I treat the effects of globalization as the main explanatory variable to understand the varying levels of social provision in China. China has experienced a commodification of labor with its market reform that brought about rapid economic growth; this commodification process led to increases in labor unrests as well as demands for social protection against increased market competition (C. K. Lee 2007). Lastly, this study theorizes the interplay between globalization and the threat of social unrest on the expansion of social insurance coverage. I argue that the threat to regime stability acts as a conditioning factor between economic openness and welfare expansion. As the threat of social unrest is not always met with government concessions via increased welfare, varying levels of potential social unrest should be examined as an intervening variable in which the effect of economic openness is dependent on.

Theoretical Discussion: The Political Incentives for Expansion of Welfare Provision

Threat of Social Unrest

In this section, I present a theoretical mechanism to integrate the compensation and efficiency hypotheses that explain government responses to economic openness via welfare provision. This mechanism relies on the political survival and career advancement of local officials as the main motivations behind their provision of social protection. I argue that increased economic integration into international markets affects local officials’ career prospects in two ways: first, increased threat of social unrest in highly open economies may obstruct local officials in maintaining stability in their jurisdictions; second, high potential of social unrest may also pose the threat of capital flight, which would negatively affect local officials’ pursuit of economic growth. Failure to achieve the two prominent goals of stability maintenance and economic growth will disrupt local officials’ career advancement.

Previous studies on the compensation hypothesis have relied on the implicit effect of the threat of social instability to explain governments that expand commitments to social spending. Countries that increase exposure to international markets are likely to experience social dislocations, uncertainty, and unequal distributive effects (Kaufman and Segura-Ubiergo 2001). This in turn creates a potential for political instability and may ultimately endanger politicians’ positions in government (Kaufman and Segura-Ubiergo 2001; Avelino et al., 2005). Therefore, in order to keep instability at bay and dampen potential insurgencies, governments will be more willing to provide social protection for dislocated and aggrieved sectors from globalization. Research on authoritarian redistribution has also emphasized the relationship between the threat of social stability and expansion of welfare benefits in authoritarian regimes; autocrats or ruling parties in non-democracies often redistribute in order to secure political survival (Bueno de Mesquita et al. 2004; Albertus 2015; Gandhi and Przeworski 2006; Magaloni 2006).

Despite the clear effects of the potential of social instability that trigger the need to preemptively redistribute in non-democracies, studies on globalization and welfare state development have often excluded the potential of social unrest from empirical analysis. By doing so, proponents of the compensation hypothesis have assumed that a sizeable public sector is a routine and functional requirement of a free trade regime (Adserà and Boix 2002), and/or that political instability is automatically linked with trade openness. However, regions that experience increased economic integration do not face a uniform threat of social unrest, or respond similarly to provide increased social protection for those aggrieved sectors from globalization. Excluding the effects of potential political instability from the model fails to effectively explain the subnational variation in the generosity and coverage of social insurance schemes. Provinces or states may experience varying levels of the threat of social instability, which may in turn lead to differing levels of social welfare benefits within a country. This paper argues that the effect of globalization on welfare expansion depends on the threat of potential social instability in China. In other words, openness to trade shapes the size of local governments’ provision of welfare conditional on the threat of political instability.

Stability Maintenance and Local Officials’ Career Prospects

Increased threat of social unrest may incentivize local officials to provide social insurance because of the cadre evaluation system. Various studies have examined the impact of the cadre evaluation system on local government decisions for redistribution (Chou 2005; Edin 2003; X. Huang 2013; C. K. Lee and Zhang 2013). The cadre evaluation system is one of the most important components of government personnel management in China, as its results influence decisions about career appointment, promotion, transfer, and removal (Wang 2013). Among the various performance tasks emphasized by the central government, maintaining stability and strengthening political and legal work in order to safeguard social stability have been the most prioritized political task for the CCP since the Tiananmen Crisis (Yang 2017). The tendency to localize responsibility for stability maintenance and strengthen the implementation of an integrated public security management has become increasingly codified by the 2000s (C. K. Lee and Zhang 2013). As is the case, maintaining social order has become one of the priority targets of the “Target Responsibility System” that affects local leaders’ political careers (Chou 2005; Edin 2003). Local officials who fail to meet priority targets, such as family planning and social order, are likely to be vetoed in future appointments regardless of their work performance (X. Huang 2013, 2015; Chou 2005; Edin 2003).

Another interesting aspect of the evaluation system is that its results reflect not only the scores for performance targets (e.g., maintaining social stability), but also that of public opinion (minyi diaocha) (Zuo 2015). One of the most interesting developments of the cadre evaluation system beginning in the late 1990s was the introduction of citizen participation in performance measurement. The rate of resident satisfaction was listed in the evaluation framework and its weight was increased; for example, in Qingdao city, the relative weight of public satisfaction rate in performance measurement increased from 8% in 2006 to 14% in 2007 (Burns and Zhiren 2010).

The nature of the cadre evaluation system thus leads local officials to achieve government-mandated targets without compromising public opinion. The increasing weight of public opinion surveys on evaluation, coupled with the imperative need to maintain social stability, pressures local cadres to use non-repressive methods to mitigate social tensions. In such circumstances, selectively providing social insurance to aggrieved workers negatively affected by exposure to international competition emerges as an attractive tool to not only dampen potential insurgencies, but also secure favorable resident opinion.

Previous studies have also examined how the threat of social unrest may induce non-repressive government concessions, including redistribution and social welfare (Cai 2002; C. K. Lee and Zhang 2013; O’Brien 1996, 2005). Protests and petitioning enabled individuals to enter into direct negotiations with the state (Liebman 2013), and different attributes of collective demands that signal a threat to social stability have been found to prompt government concessions in the form of welfare benefits (Hwang 2019). As is the case, the expansion of social protection to those who were formerly not eligible may work as a powerful tool for leaders to retain social control (C. Li 2013). Local officials who wish to survive under the top-down evaluation system will aim to prevent social unrest from breaking out in their jurisdictions by proactively designing and providing social welfare benefits (X. Huang 2015).

Threat of Social Unrest and Economic Performance

Another important soft target for cadre evaluation is economic performance. Empirical studies on political turnover find that economic and fiscal performance enhances the promotional prospects of local leaders (G. Guo 2007; H. Li and Zhou 2005). Economic performance has been assigned heavy importance in the target responsibility system (Tsui and Wang 2004).

High levels of labor unrest, however, may be adversary for attracting and retaining FDI. Government stability as well as ensuring law and order has been found to be highly relevant for investment decisions of multinationals (Busse and Hefeker 2005). According to Busse and Hefeker (2005), the absence of internal conflict, measured as political violence within the country and its potential impact on governance, is also a significant determinant of foreign investment inflows. Considering that motives such as low-cost labor or being close to export markets largely drove the investment decisions of transnational corporations during the 1980s and 1990s (Busse 2004), China has long been a prime location for investment (Robertson and Teitelbaum 2011). However, the advantages of securing access to low-resources in countries with repressive regimes have to be weighed against the possibility of protests by activist groups, involving negative publicity and the costs of public relations and consumer protests in the form of boycotts (Busse 2004).

Previous studies have found that increased levels of social unrest not only threaten stability, but also pose the risk of capital flight (Lensink et al., 2000; Le and Zak 2001). One of the features of globalization, other than increased competition, is increased capital mobility (Avelino et al., 2005; Alesina and Tabellini 1989; Moller et al. 2003). International investors may be incentivized to shift their capital to other countries as an insurance against the risk of future taxation (Svensson 1998; Quinn 1997; Rudra 2002; Moller et al. 2003). However, capital flight may also occur under political instability due to increased levels of protest and collective action (Chen et al., 2016; Robertson and Teitelbaum 2011; Lensink et al., 2000). Industrial unrests as well as political uncertainty may lead to capital flight as industrial instability not only compromises productivity, but also, if escalated to the level that threatens regime stability, endangers the credibility of the commitment made with the incumbent government.

Increased integration into capital markets thereby provides exit opportunities for international investors, who usually perceive contentiousness and noisiness of worker grievances as negative for investment (Robertson and Teitelbaum 2011). Increased and prolonged episodes of labor disputes may signal potential social instability that may not only deter FDI, but also encourage capital flight, which may obstruct economic growth. The emphasis on economic performance for cadre evaluation creates a political incentive for local officials to focus on coopting and preempting potential social unrest through providing expanded social protection, especially for those exposed to greater economic volatility.

However, if these localities do not experience high threat of social unrest, local officials will not be pressured to expand social protection as they do not face the threat of either social unrest or subsequent capital flight. In this case, I expect that local officials will be incentivized to pursue higher economic growth by retrenching or cutting back on social protection to secure market competitiveness. Although market competitiveness can be enhanced through the productive welfare state model that prioritizes commodification by encouraging wage labor, productive welfare states ultimately rely on improving domestic firms’ international competitiveness through cost containment (Rudra 2007). Increases in trade and capital openness can pressure governments to race to the bottom in social spending and labor standards to increase competitiveness (Swank 2002; Scharpf Schmidt, 2000; Clayton and Pontusson 1998). The range of social policies in the productivist model is therefore limited: they may repress or provide minimal social rights usually to the white-collar workers (Rudra 2007), or be subordinate to the overriding policy objective of economic growth (Holliday 2000). Provinces that experience low levels of social unrest may also indicate low potential labor power or political leverage, which may further enable local officials to view retrenchment as a convenient option to secure firm competitiveness because workers are less capable of defending their welfare benefits (Rudra 2002). Therefore, I expect that in the context of an open economy, provinces with low levels of social instability will be more likely to cut back on social welfare provision as a strategic response to enhance market efficiency and secure competitiveness. Based on this discussion, I offer the following hypotheses:

  • Hypothesis 1: Provinces in open economies with higher threat of social unrest provide broader coverage in social insurance programs.

  • Hypothesis 2: Provinces in open economies with low threat of social unrest provide lower coverage in social insurance programs.

Data and Measures

Dependent Variable

The dependent variable in the analysis is the coverage of urban employee basic pension insurance, basic health insurance, and unemployment insurance. The data is drawn from statistical yearbooks of 31 provinces in China, covering 15 years from 2001 to 2015. The outcome of interest is measured as the percentage of the number of people participated in the insurance programs to the total urban population of the province.

Studies on the effects of globalization on governmental welfare provision have relied on government social spending as the outcome of interest (Rudra 2008; Avelino et al., 2005; Rodrik, 1997). However, all types of spending measures are imperfect proxies for the actual payoffs that citizens receive (Kaufman and Segura-Ubiergo 2001). Changes in social spending may be an effect of general changes in government expenditure, as trends in central government spending may have an impact on welfare spending for lower-level governments (Kaufman and Segura-Ubiergo 2001). Moreover, the effects of a universal social policy, such as pensions, on a spending measure are realized over a long span of time; a significant policy change as an outcome of political struggles or economic conditions does not lead to an immediate increase in the spending measure (Lee et al., 2011).

I therefore rely on coverage of social insurance programs as it reflects the share of the population that actually has access to the benefits provided (X. Huang 2015). Since the late 1990s, the central government initiated social insurance programs for urban workers, including unemployment insurance, basic pension, and basic health insurance schemes (Qian and Mok 2016). The main social insurance programs in China can be distinguished by coverage: inclusive benefits (pensions, health insurance) that are broad in enrollment, and selective benefits (unemployment benefits) that usually require long-term formal employment and thereby exclude considerable segments of the population such as migrant workers or peasants (X. Huang and Gao 2018). In order to better understand the expansion of social policies as a qualitative outcome, this study focuses on the coverage of both inclusive and selective social insurance programs that aim to protect incumbent workers against the vagaries of the international economy.

Independent Variables

Globalization

The key explanatory variable is globalization, which is measured by trade openness (Krishna et al., 2014; Avelino et al., 2005; Rudra 2002; Rodrik, 1997; Adserà and Boix 2002; Huber et al., 2008). Following conventional practices in most of the literature on globalization, trade openness is calculated as the ratio of the sum of imports and exports to Gross Regional Product (GRP). According to the Chinese Statistical Yearbook, the unit of measurement for the total value of imports and exports of operating units for each province was 1,000USD; for GRP, the unit of measurement was 100 million yuan. The total value of imports and exports were converted from USD to Chinese yuan for consistency.

Threat of Social Unrest

I expect that the effect of globalization on welfare expansion depends on the threat of social unrest in China. I use the number of labor disputes—the number of labor dispute cases that had been accepted every year, drawn from the China Labor Statistical Yearbook—as a proxy for the threat of social unrest. Labor disputes are defined as controversies between employers and employees, including collective negotiations, strikes, or even violent unrests (Y. Li 2014). Taking into account the lack of reliable data on the number of protests or petitions, labor disputes are the most important form of collective actions in China, listed as the “troika” for social unrest along with land disputes and protests over pollution (Y. Li 2014).Footnote 1 I assume these measures will largely capture the openness of the economy and the threat of social unrest at the province level. To correct for a highly skewed distribution, both measures were logged.

Control Variables

Several control variables will help isolate this main relationship and check for other influences on welfare coverage. I control for two factors of labor force composition: gender ratio and dependency ratio. Gender ratio is calculated as the ratio of male to female population of the province. It is expected that women may demand more protection than men in comparable jobs because of maternity leave (Estévez-Abe et al. 2001). Contrary to Estevez-Abe, Iversen, and Soskice (2001), however, I expect that gender ratio will affect coverage positively; provinces with more male population compared to female population will have higher levels of coverage, as the working population in China is still heavily male dominant. Dependency ratio is measured by the number of people over 60 years old (aged population) divided by the number of people aged 20 to 59 (working population) (X. Huang 2015; Rudra 2002). Contrary to X. Huang (2015)’s findings, I expect dependency ratio to have a negative effect on old-age pension coverage, as the aged population are considered to be recipients of pension, rather than participants who are currently covered by the pension system.

I also control for several economic factors as higher levels of economic development or stronger fiscal capacity may allow local governments with enough resources to expand welfare (Orloff and Skocpol 1984; Evans, Rueschemeyer, and Skocpol 1985). Economic development is measured by the logarithm of GRP per capita and government fiscal capacity by the ratio of budget expenditure to revenue (Lorentzen, Landry, and Yasuda 2014). Lastly, I control for urbanization, measured as the share of urban population of the total population in the province, which may either reduce or increase pension coverage. Some scholars argue that an increased urban population may not necessarily lead to increased levels of coverage due to the strict household registration system that prevents migrant workers without urban hukou from having access to social insurance (X. Huang 2015). Others argue that urban bias induces larger transfers to cities because the “squeaky wheel gets the grease”—the regime is more likely to hear complaints from areas of political strength (Wallace 2014).

A number of other control variables that may affect welfare provision are included in the regressions: provincial-level total population; migrant population, measured as the absolute value of the difference between provincial total population and local population (i.e., population with local hukou) to control for social risk and labor mobility (X. Huang 2015); level of natural resources, measured as the ensured reserves of coal at the province level, to control for the effect of natural resources on public goods provision (Hong 2018); and the level of foreign-funded enterprises to control for provincial dependency on foreign capital and the subsequent threat of capital flight.

Methods

To explore the relationship between globalization, social unrest, and welfare coverage, I constructed an unbalanced panel dataset of 31 Chinese provinces observed from 2001 to 2015 from the Chinese Statistical Yearbook and Chinese Labor Statistical Yearbook. This unbalanced, panel data structure poses the risk of correlated errors due to the presence of unmeasured unit-specific, time-invariant factors present in observations within the same unit (C.-S. Lee 2005; C.-S. Lee, Kim, and Shim 2011). Such unobserved components, if transferred to the error term, will generate heterogeneity bias and biased coefficients (Stimson 1985).

Random effects model (REM) and the fixed effects model (FEM) are the two approaches most widely used to correct for such unobserved, time-invariant effects. There are several reasons why the REM is preferred over the FEM. First, REM can accommodate time-invariant and slowly time-varying institutional factors by systematizing the random error terms (C.-S. Lee, Kim, and Shim 2011). The REM also accounts for respondent-specific characteristics that are unobserved and not captured by the independent variable (Western and Beckett 1999). As even slowly time-varying factors may lose their level of statistical significance in the FEM model, REM would provide a better context to incorporate time-invariant covariates such as geographic regions, political institutions, and regime type.

Second, the FEM does not effectively address all between-unit variation present in the data (Nielsen and Alderson 1995). Though the FEM has the advantage of avoiding spurious relationships in cross-national data, this might lead to insignificant coefficients despite the theoretically explanatory power of the variables if substantial variation exists between units of observation (C.-S. Lee 2005). As the model in this study hypothesizes that levels of economic integration interacting with social unrest explain the different levels of coverage between provinces, my substantive logic primarily depends on between-unit variations rather than within-unit variations. As can be seen in the descriptive statistics below, the variation of the explanatory variables in the model is substantial across provinces rather than over time.Footnote 2

Results

Descriptive Results

Table 1 displays the summary statistics of the key variables. Coverage rates for pension insurance range from 13.3 to 75.9%, 17.3 to 78.6% for health insurance, and 11.6 to 57.7% for unemployment insurance for the panel dataset. When comparing the highest coverage rates for the three insurance programs, that of unemployment insurance is the lowest.

Table 1 Summary statistics of key variables
Full size table

Table 2 shows the descriptive statistics for the key variables by region in more detail. The north and northeast regions of China have maintained higher levels of coverage rates for pension, health insurance, and unemployment insurance. On the other hand, coverage rates for the three insurance programs have remained consistently low in the southwest region, which include Chongqing, Sichuan, Guizhou, Yunnan, and Tibet.

Table 2 Over-time changes in the average values of the key variables by 5 years
Full size table

It is interesting to note that despite having the highest coverage rates for pension insurance and health insurance, the northeast region of China does not display similarly high levels of trade openness. Likewise, regions that have low levels of insurance coverage rates do not necessarily have correspondingly low levels of trade openness or labor disputes. This suggests that levels of trade openness alone may not be sufficient to explain welfare provision, contrary to the expectations of previous studies (Iversen and Cusack 2000; Avelino, Brown, and Hunter 2005; Kaufman and Segura-Ubiergo 2001). In the next section, I test the effects of the interaction between trade openness and the threat of social unrest on the subnational variation in coverage rates.

Multivariate Results

Table 3 presents the regression estimates from regression models that test the effects of the threat of social unrest and globalization on coverage of pension, health insurance, and unemployment insurance. Model 1 shows the results of the baseline model testing the relationship between the interaction between trade openness and labor disputes on pension coverage. It is interesting to note that though the interaction term is not statistically significant in the baseline model, it has a positive and significant effect on pension coverage when including provincial-level controls (model 2) and year and region dummies (model 3). In the full model, gender ratio and the number of foreign-funded enterprises have a positive and significant effect on pension coverage. On the other hand, dependency ratio and the migrant population have a negative effect on pension coverage: as dependency ratio is measured by the size of aged population over working-age population, provinces with a relatively small working population may be associated with lower levels of coverage. Both GRP/capita and fiscal capacity are negatively associated with pension coverage. This may imply that provincial governments’ decisions to expand access to social insurance may be driven more by political, rather than economic factors.

Table 3 Regression results for pension, health insurance, and unemployment insurance (random effects)
Full size table

The effects of trade openness and labor disputes on pension coverage can be better understood through Fig. 1(1). According to this graph, provinces with low level of economic openness provide narrow coverage in social insurance regardless of the frequency of labor disputes. For pension coverage, the slope for provinces with low trade openness (0.03) is flat—coverage for pension insurance does not vary at any level of labor disputes. On the other hand, pension coverage increases with the number of labor disputes for provinces with higher levels of trade openness: pension coverage is most likely to be highest in provinces with both high levels of trade openness and labor disputes. At the lowest frequency of labor disputes, pension coverage decreases with more trade openness, which lends support to Hypothesis 2 that expects local officials to cut back on social protection to pursue higher economic growth in the absence of the threat of social unrest or capital flight.

Fig. 1
figure 1

The effects of trade openness and labor disputes on coverage rates (predictive margins with 95% confidence intervals)

Full size image

Models 4 to 6 present the effect of trade openness and labor disputes on health insurance coverage, while models 7 to 9 demonstrate the effect of the interaction term on unemployment insurance coverage. As can be seen in the full models for both health insurance and unemployment insurance, the interaction between trade openness and labor disputes has a positive and significant effect on coverage.

Figures 1(2) and 1(3) provide a graphical representation of the interaction term of the full model based on the results reported in models 6 and 9. As can be seen in the case of health insurance coverage, among the provinces with the highest level of trade openness, those experiencing high levels of labor disputes are more likely to provide higher health insurance coverage. Likewise, the likelihood of health insurance coverage decreases significantly as labor disputes decrease among provinces with similarly high levels of trade openness. The effect of trade openness seems to turn negative to positive with statistical significance at levels higher than 0.53. Provinces with high levels of both trade openness and labor disputes tend to experience high levels of health insurance coverage, which lends support to the compensation hypothesis. On the other hand, provinces with high levels of trade openness but low frequency of labor disputes tend to have lower health insurance coverage. This may reflect the efficiency hypothesis where highly open localities that do not experience high labor dispute events may not be pressured to expand social protection as they do not face the threat of either social unrest or subsequent capital flight. In these cases, local officials will be incentivized to pursue higher economic growth by retrenching or cutting back on social protection to secure market competitiveness.

It is interesting to note that for health insurance coverage, the slope of low economic openness is negative: provinces with low trade openness provide lower coverage of health insurance with an increase in labor disputes. This may reflect cases where local officials prioritize stability maintenance over achieving economic growth, as the latter goal is difficult to pursue in contexts of low economic openness. The motivation to prioritize stability maintenance may lead to more repressive responses such as retrenchment, crackdown, or relational repression (Deng and O’Brien 2013) towards high frequency of labor disputes. The decreasing coverage of health insurance in provinces with low trade openness further lends support to previous studies that examine the relationship between local governments’ fiscal autonomy and health expenditures (Tan 2017; X. Huang 2015). Tan (2017) finds that in aggregate, 99% of government health expenditure is spent at the local levels and approximately 30% of the spending is financed by transfers from the central government. Health insurance benefits tend to be vulnerable to fiscal cutbacks, particularly when local governments are unable to make ends meet (X. Huang 2014). Because the availability of local government revenue is dependent on local economic development, among other factors, provinces with low economic openness but high social unrest may be more likely to face fiscal constraints for health insurance provision as the financial burden of maintaining domestic security rises (Xie 2013).

Lastly, Fig. 1(3) displays a similar pattern of the relationship between trade openness and labor disputes on unemployment insurance coverage. In provinces with low trade openness (0.03), there is no substantial difference in unemployment insurance coverage at different levels of labor disputes. On the other hand, for provinces with high levels of trade openness (above 0.53), coverage for unemployment insurance increases proportionately with the frequency of labor disputes. These results demonstrate how local governments may display patterns of either the compensation of efficiency hypothesis depending on the conditioning effects of the threat of social unrest.

Conclusion

This study aimed to examine the political economy behind unequal welfare expansion trends across provinces in China by focusing on two factors: provincial economic openness and social unrest. Building on previous explanations that focus on how cadre evaluation, which emphasizes stability maintenance and economic growth, incentivizes local officials to provide different levels of concessions to public discontent (X. Huang 2015; Hwang 2019; Liebman 2014; Wallace 2013; C. K. Lee and Zhang 2013; Chen et al., 2016), this paper aimed to offer a way in which the two competing theories on the relationship between economic openness and welfare provision could be integrated depending on the conditioning effect of the threat of social unrest.

In particular, I find that provinces with high economic openness and high threat of social unrest support the compensation hypothesis: they expand welfare coverage to proactively prevent and preempt the manifestation of the threat of social instability. On the other hand, those highly open provinces that face low threat of social unrest support the efficiency hypothesis: these localities may cut back on, or provide low levels of welfare coverage as a means to take advantage of international competitiveness and achieve economic growth. These findings demonstrate how the demands for state compensation and risk-sharing may vary for local officials, to whom maintaining social stability and achieving economic growth are imperative goals for political advancement. In short, as globalization has a differential impact on local economies, efforts to curb potential social unrest through expanded coverage will also vary according to each province’s economic and political conditions.

This study on the effects of globalization and the threat of social unrest on welfare provision suggest important directions for future research in the area of welfare state scholarship in non-democracies. By interacting labor disputes and economic openness, I aimed to disentangle the mechanism behind globalization and state compensation. More refined and comparable data of local economies, measures of the threat of social unrest, and cadre evaluations in China may provide new insight into the local industries or sectors that are more vulnerable to international competition than others, the different types of social instability threats that may consequently arise, and how this may affect the coverage or generosity of various social insurance schemes. Another avenue for future research concerns the role of firms and investors in welfare development. Though I have provided a state-centric approach to welfare provision, social policy as a result of strategic interactions between firms in China remains unexplored. Accounting for the responsibility of firms to partially contribute to employee’s social insurance schemes, a detailed analysis on the exposure to risk and its subsequent effect on firm incentives to invest in social policies will reveal new perspectives on the role of state, business, and labor in welfare development. Lastly, examining social insurance and assistance policies that target marginalized groups in China, such as informal workers, migrant workers, or the disabled would also provide a more comprehensive understanding of how local governments differently address the social costs of globalization experienced by various segments of the population. A more refined model that addresses the diverse state and societal actors of welfare provision may have important theoretical implications for the adaptability and resilience of the Chinese authoritarian regime.