Size and structure

In the welfare state literature, two main approaches are used to measure the size and structure of welfare states: the macro-level measure estimates the percentage of GDP (or total government expenditure) spent on social welfare based on official statistics, while the micro-level measure examines, from household survey data, the percentage of social benefit income in the household final income. The macro measure is more widely used in the western welfare state literature, partly because most government expenditure data are easily accessible via official statistics (and also compiled by the OECD) and partly because they are more clearly defined than household survey data and thus can be consistently estimated across countries. In recent years, more welfare state studies have adopted the micro-level measure, thanks to the availability of harmonised cross-national household survey data. The two measures do not always coincide, but taken together they often yield consistent conclusions about trends within a country and comparisons across countries (Garfinkel et al., Reference Garfinkel, Rainwater and Smeeding2010).

The micro measure has two unique advantages over the macro measure. First, the micro measure captures the actual income transfers and in-kind goods and services received by family members that can directly translate to material well-being, while the macro measure captures government spending at the aggregate level and does not always correspond to the actual benefits received by families. Second, because the micro measure captures social benefits at the household level along the income distribution and across various types of households, it enables scholars to estimate the redistributive effects of the welfare state, a goal unattainable by using the macro aggregate measure. In this article we will mainly rely on the micro measure based on the CHIP data, but will also supplement these main results using estimates from macro data on government welfare expenditures.

The size of the welfare state is measured as the amount of social benefits received by families as a percentage of their household final income. We adopt a comprehensive measure of total household per capita income (or final income), which includes market income, cash and in-kind social benefits, private transfers and other transfers less taxes and fees paid (Gao, Reference Gao2010; Gao and Riskin, Reference Gao, Riskin, Davis and Wang2008; Li et al., Reference Li, Sato and Sicular2013). Market income includes wage income, income from private enterprises/individual business, property income and rental value of owner-occupied housing. Social benefits include social insurance (mainly pensions and unemployment insurance), in-kind benefits (health care, housing and food) and public assistance (mainly minimum living standard assurance, or Dibao, the primary safety-net programme targeting the poor). The shares of the various social benefits in the total social benefit package measure the structure of the social benefit system.

We have a few important inclusions and exclusions in the measurement of social benefits that deserve some explanation. First, following the predominant tradition in the welfare state literature (Garfinkel et al., Reference Garfinkel, Rainwater and Smeeding2010; Hacker, Reference Hacker2004; Lampman, Reference Lampman1984), employer-provided benefits are included. This is particularly important in the Chinese case as the urban welfare system is heavily employment-based. These benefits are an integral part of the welfare state as they are mostly mandated by the government and the costs are shared by employers and the government. Including such benefits is also important for estimating the redistributive effects of the welfare state as these employment-based benefits tend to be more regressive than others.

Second, we include in-kind benefits – an important yet often omitted or undervalued component – in the social benefit package (Smeeding, Reference Smeeding1982). In-kind benefits are mostly estimated at their self-reported cash values. In all waves, participants were asked to estimate the rental value of the public housing that they occupied and the cash value of food items that they received. In 1995, 2002 and 2007, as well as in the 1988 rural surveys, participants were also asked to report the medical-care expenses covered by their employers, the government or the collective and estimate the cash value of in-kind health services, the sum of which became the value of total health benefits received by families. In the 1988 urban survey, however, the value of health benefits was not directly asked for and this is thus imputed using administrative data on provincial per capita public expenditure on employee health care. The administrative data differentiate public health expenditures on employees and retirees, respectively, by three types of employers – state (including public institutions), collective and other enterprises. Provincial per capita health expenditures on current or retired employees are obtained by dividing provincial total health spending by number of employees or retirees according to employer type. The administrative data are then imputed to individuals according to their employment status and type (Gao, Reference Gao2006). This imputation undermines the redistributive effect of health benefits for urban 1988, but is essential for the inclusion of this benefit for the entire study period.

Third, education, an important component of the welfare state, is not included in the calculation of the size of the Chinese welfare state using the CHIP data. Because the values of education benefits received were not directly asked about in the CHIP surveys, we would need to impute the values of education benefits using administrative data on government spending on education. This is doable but the results will underestimate the distribution of education resources and benefits across income groups and localities because of the lack of detailed education spending data at various levels of income and localities.

Figure 1 presents the share of social benefit income in household final income in urban and rural China respectively using CHIP 1988, 1995, 2002 and 2007 data and among the rural-to-urban migrants for 2002 and 2007. A few important findings are revealed by this figure.

Figure 1.

Shares of social benefit income in household final income

Source: Authors’ calculations using CHIP data.

First and foremost, it shows that there exists a huge gap between urban and rural social benefits, or, the urban and rural welfare states. In 2007, urban social benefits made up 20 per cent of household final income, while that share was only about 2 per cent for rural residents. If they are indeed treated as separate urban and rural welfare states, the size of the urban welfare state would be comparable to those of western industrialised countries (Garfinkel et al., Reference Garfinkel, Rainwater and Smeeding2010), while the size of the rural welfare state would be among the world's smallest and similar to those of the least developed countries. Empirical evidence on the welfare systems of developing countries is scarce. Grosh et al. (Reference Grosh, del Ninno, Tesliuc and Ouerghi2008) studied safety-net programmes in seventy-two developing countries. They found that government spending on safety-net programmes as a percentage of GDP averaged only 1.9 per cent, ranging between 0.2 per cent in Senegal and 7.7 per cent in Mauritius.

It is important to note that our measure – social benefits as a percentage of household final income – is a relative one. In fact, the sharp urban–rural gap revealed above is even greater when the actual levels of household income are considered. Figure 2 shows the levels of household per capita final income in annual yuan in urban and rural China and for rural-to-urban migrants (in 2007 value). Income levels rose significantly for all three groups during the study period, but the gap between the urban and rural areas remained large and became even larger in 2007. The urban-to-rural household final income ratio was 2.4 in 1988 and rose to 2.9 by 2007. In 2007, the ratio between the urban and rural levels of household per capita social benefits was 29, astonishingly high and much higher than the 20 per cent vs. 2 per cent difference when measured as a percentage of household final income as shown in Figure 1. However, our main analyses below still rely on the relative measure that we set out to use, as it is widely used in the international welfare state literature. Using this measure is also helpful for us to understand the different features of the urban and rural ‘welfare states’ within China. The income levels presented in Figure 2 can serve as a helpful contextual reference for understanding the remaining results.

Figure 2.

Household per capita final income (annual yuan in 2007 values)

Note: All income levels are converted to 2007 values using official urban and rural CPIs. Income levels for urban citizens and migrants are converted using the urban CPI, while those for rural citizens are converted using the rural CPI.

Source: Authors’ calculations using CHIP data.

Second, as also shown in Figure 1, the urban social benefits measured as share of household final income shrunk significantly during this period. In 1988, social benefit income on average made up 45 per cent of household final income in urban China. That share, however, kept decreasing: to 29 per cent in 1995, 27 per cent in 2002 and 20 per cent in 2007. The cutbacks in social benefits in urban China took place alongside reforms of state-owned and collective enterprises. Before the economic reforms, under the full-employment policy, nearly all urban citizens had job security and comprehensive coverage of social benefits provided for by their employers. However, the financing of these social benefits became a heavy burden for employers and a hindrance to overall economic growth. To promote economic growth and efficiency, state-owned and collective enterprises were reformed to reduce or remove the welfare burden. As a result, the financing responsibility of pensions and health care has shifted from resting solely on employers to now being shared by employers, employees and the government. Housing and food benefits have been almost eliminated.

Consistent with these policy changes, empirical evidence based on the CHIP data in Figure 3 shows that, among the urban social benefits, social insurance increased steadily while all other benefits declined as a share of household final income from 1988 to 2007. Social insurance income, mainly made up of pensions and unemployment insurance, increased from 6 per cent of household final income in 1988 to 11 per cent in 1995, 15 per cent in 2002 and 17 per cent in 2007. This trend partly reflects the strengthening of social insurance programmes through broad employer and employee contributions, and partly population ageing, which is raising the portion of pension income in total society income.

Figure 3.

Shares of various types of social benefit income in household final income in urban China

Source: Authors’ calculations using CHIP data.

On the other hand, all other benefits dropped as a share of household final income. Specifically, health care benefits initially increased from 4 per cent in 1988 to 5 per cent in 1995 and 7 per cent in 2002, reflecting the dramatic rise in health care costs and increased health consciousness among the public rather than a more generous health care system per se. From 2002 to 2007, health benefits as a share of household final income dropped to only 2 per cent, but this is an underestimation of the actual health benefits urban citizens receive. This is mainly because our estimate of health benefits in 2007 only captures in-kind health care and medical expenses paid by the work unit, but not the general benefits of participating in socialised medical insurance, including the Urban Workers Basic Medical Insurance (UWBMI) covering workers and the retired and the Urban Residents Basic Medical Insurance (URBMI) for urban residents not covered by the UWBMI, including children, students and older persons without an employment history. By the end of 2007, there were more than 180 million urban workers in the UWBMI scheme and nearly 43 million in the URBMI scheme, with a coverage rate of less than 50 per cent for both groups (Ministry of Human Resources and Social Security [MHRSS], 2008). Moving the health benefits from employer-based health care to social pooling helped to increase the health care coverage and shift the distribution of health benefits to being more equal among urban citizens.

Housing, food, other in-kind benefits and public assistance all decreased steadily from 1988 to 2007, as evident in Figure 3. Specifically, housing had been privatised, largely through subsidised purchasing over the years, yielding a sharp decrease in its share in household final income (from 19 per cent in 1988 to 10 per cent in 1995, 3 per cent in 2002 and only 0.8 per cent in 2007). Food assistance plummeted from 11 per cent of household final income to 0.7 per cent in 1995 and kept falling to only 0.3 per cent in 2007. Other in-kind benefits remained about 0.1 per cent of household final income during this period. Despite recent expansions in the Dibao programme, public assistance also decreased from 3 per cent of household final income in 1988 to 1 per cent in 1995 and 2002 and then dropped to only 0.2 per cent in 2007, suggesting that such public assistance programmes remain residual in the overall urban welfare state and on average only play a small role in contributing to total household income.

The third striking finding from Figure 1 is the consistently trivial welfare benefits in rural China. Social benefit income on average remained only 1 per cent of household final income for rural citizens from 1988 to 2002 and increased slightly to 2 per cent in 2007. In contrast to the much more comprehensive and generous social benefits in urban China, rural social benefits has always been marginal in coverage and minimal in provision. Only a very small privileged group of rural residents has had access to pensions due to their prior employment in state-owned or collective enterprises. The pre-reform, community-based rural health financing and provision system provided basic health services and wide coverage to rural residents. It was jointly funded by family contributions, a portion of the collective welfare fund and sometimes the government. The system collapsed after the rural economic reforms were launched in 1978, leaving many rural residents unable to afford the dramatically increasing health care expenses.

Lastly, Figure 1 shows that the social benefits for rural-to-urban migrants expanded substantially between 2002 and 2007. As a percentage of household final income, social benefits for migrants increased significantly from 3 per cent in 2002 to 12 per cent in 2007. Figure 4 further illustrates the structure of the social benefits for migrants in 2007. Nearly all social benefit income for migrants was from in-kind benefits, including food (6 per cent) and housing (5 per cent). These two kinds of benefits refer to catering and accommodation provided by their work units, or allowances for meals or accommodation deducted from their salary. Therefore, this may represent part of their salaries in the form of in-kind payment rather than more generous social benefits per se. Indeed, if these in-kind payments of food and housing are not counted, the share of social benefits in migrants’ household final income is reduced to a bare 1 per cent, very close to that for their rural peers. This reveals the still seriously lacking conditions of social benefits for migrants. Health benefits for migrants were minimal, mainly because migrants did not qualify for urban health insurance due to their lack of local city Hukou. Cash transfers received by migrants were also minimal, indicating that the current pension system and unemployment insurance covered only very few migrants. Migrants were also ineligible for Dibao benefits due to their lack of local city Hukou.

Figure 4.

Shares of various types of social benefit income in household final income for rural–urban migrants in China

Source: Authors’ calculations using CHIP data.

How do these trends based on the micro measure using the CHIP data compare to those using a macro measure? Below we use estimates from macro-level government welfare expenditure data to supplement the results presented above. Before we do so, it is important to note a few important defects of the administrative data. First, the available administrative data mostly only include government expenditures but exclude employer-provided benefits, an important component that we are able to capture using the CHIP data. Second, the administrative data do not differentiate government welfare expenditures by urban and rural areas or for migrants. Therefore, we are only able to obtain a crude estimate of the size and structure of the Chinese welfare state as a whole, but not by subgroups. Third, as mentioned earlier, the administrative data do not help with the analysis of the redistributive effects of social benefits.

Figure 5 shows the trends in social welfare expenditures as a percentage of GDP in the four years covered by CHIP and in 2011, the latest year for which data are available. Only government expenditures on cash transfers (including pensions and social assistance), health and education are included. Housing and food benefits are not captured due to data unavailability. Overall, total government expenditure on cash transfers, health and education as a percentage of GDP has been increasing since 1995, reaching 5.7 per cent by 2007 and 7.4 per cent by 2011. Among the three categories, cash transfers had significant increases from only 0.3 per cent in 1988 to 2.3 per cent in 2011, mirroring the similar trend revealed by CHIP data in Figure 3 (as cash transfers were mostly available only in urban China). Health expenditures also increased, reaching 1.6 per cent in 2011, echoing the similar trend shown in Figure 3 (as noted earlier, CHIP data in 2007 did not fully capture the health benefits). Education, a benefit that we are unable to analyse using CHIP data, was the largest of the three expenditures and saw a steady increase over the years, reaching 3.5 per cent of GDP in 2011. Unfortunately, the two main in-kind benefits – housing and food – that dominated the declining trend of urban social benefits from 1988 to 2007 as shown using CHIP data are unavailable for analysis using administrative data. If the data were available, we suspect that the results would be consistent with the policy changes during this period and very similar to those based on CHIP data.

Figure 5.

Shares of government social welfare expenditure in GDP

Source: Authors’ calculations using official statistics from National Bureau of Statistics [NBS] (various years).

Impact on income inequality

How big is the impact of the Chinese welfare state on income inequality? We investigate whether and to what extent the income redistribution through social benefit transfers has raised or reduced the overall income inequality by comparing the pre- and post-transfer Gini coefficients. The greater the reduction in the Gini coefficients due to the social benefit transfers, the greater their redistributive effect.

Table 2 presents the results on the impact of social benefits on income inequality as measured by the Gini coefficient.Footnote ¹ Overall, urban social benefit transfers helped reduce income inequality through the period 1988 to 2007. However, their impact kept declining over time. Rural social benefits had only negligible effects, turning from slightly reducing income inequality in 1988 and 1995 to slightly increasing it in 2002 and 2007. Social benefits for migrants raised income inequality in both 2002 and 2007, though to a small extent.

TABLE 2

Impact of social benefit transfers on income inequality as measured by the Gini coefficient

Source: Authors’ calculations using CHIP data.

Specifically, in 1988, urban social benefits reduced the overall income inequality as measured by the Gini coefficient from 0.268 to 0.224, a reduction of 16 per cent. However, urban social benefits reduced the overall income inequality by 12 per cent in 1995, 7 per cent in 2002 and only 0.7 per cent in 2007, suggesting that the equalising role of urban social benefits kept declining during this period and reached a minimal level by 2007. Rural social benefits were only able to have little impact on the overall income inequality given their small size. Rural social benefits reduced income inequality by 0.2 per cent in 1988 and 1995, but raised income inequality by 0.3 per cent in 2002 and 0.2 per cent in 2007. For migrants, social benefits actually helped raise the income inequality levels in both 2002 and 2007, and the increase was greater in 2007 (2.4 per cent) than in 2002 (1.6 per cent), indicating that social benefits for migrants were distributed more unevenly than pre-transfer income itself. Indeed, the majority of social benefits for migrants were food and housing provided by their work units, especially in 2007. These benefits were more likely to go to those with better jobs than their less advantaged peers, therefore further skewing the income distribution among the migrants.

Progressivity (or regressivity)

Across the developed and developing worlds, social benefits are one of a nation's primary mechanisms for redistribution of resources and achieving social justice. What is the redistributive nature of the Chinese welfare state? Is it regressive (i.e., favoring the rich) or progressive (i.e., favoring the poor)? We compare the economic distances at the bottom and the top of the income distribution between pre- and post-transfer incomes to address this question. We estimate the economic distances between households at three different points (10th, 50th and 90th percentiles) of the income distribution, representing the poor, the median and the rich households in the society respectively. Specifically, the economic distance between the bottom- and middle-income households will be captured by the P10/P50 ratio, with the P90/P50 ratio reflecting the economic distance between the top- and middle-income households. By comparing these two ratios, the income inequality gap between the bottom and the top of the distribution is revealed vividly (Gao, Reference Gao2010; Garfinkel et al., Reference Garfinkel, Rainwater and Smeeding2010). A progressive social benefit system implies that resources are reallocated from the rich to the poor, narrowing the economic distance between the rich and the poor. If the economic distance is enlarged at both ends, the social benefit system is clearly regressive. If social benefits reduce the economic distance at the bottom (indicating progressivity) but increase it at the top (indicating regressivity), the redistributive nature is ambivalent and the extent of progressivity or regressivity is determined by which of the two effects dominates.

Figures 6–8 show the changes in the redistributive effects of social benefits for urban, rural and migrant populations respectively from 1988 to 2007. The length of the bars represents the economic distance between low- and high-income households, with the value of the P10/P50 ratio at the lower end and the value of the P90/P50 ratio at the higher end. For urban residents, social benefits shortened the economic distance between the poor and the rich and were progressive in all years. Furthermore, the reduction of the economic distance became greater over time, suggesting that urban social benefits have become even more progressive in recent years.

Figure 6.

Impact of social benefits on economic distance between low- and high-income households in urban China (length of bars represents economic distance between low- and high-income households)

Source: Authors’ calculations using CHIP data.

Figure 7.

Impact of social benefits on economic distance between low- and high-income households in rural China (length of bars represents economic distance between low- and high-income households)

Source: Authors’ calculations using CHIP data.

Figure 8.

Impact of social benefits on economic distance between low- and high-income households for rural-to-urban migrants in China (length of bars represents economic distance between low- and high-income households)

Source: Authors’ calculations using CHIP data.

Rural social benefits were redistributed regressively from 1988 to 2002, as indicated by the elongated economic distance bars from pre- to post-transfer income, but they became slightly progressive in 2007 because the relative position of the poor was lifted slightly by the social benefits in that year. Social benefits for migrants followed the same trend as that for rural citizens. They were clearly distributed regressively in 2002, with the poor staying intact in their relative position in the overall income distribution and the rich gaining through social benefits. In 2007, social benefits were redistributed partly regressively and partly progressively: they lowered the relative income of the poor but also reduced the relative income of the rich. Because the decrease at the top was larger than that at the bottom, social benefits for migrants appeared to be more progressive than regressive in 2007.