A Study of the Impact of Social Welfare Policies on Household Saving Rate in China By Borui Xiao Advised by Professor Lakshman Krishmurthi Submitted to the Department of Mathematical Methods in Social Sciences
Abstract The Chinese saving pattern has always been a great puzzle to many scholars who intend to understand the constantly high saving rate in the past three decades that exceeded 50 percent of gross domestic product (GDP) from 2007 to 2011 and hovers around 50 percent from 2011 onwards. Even though many theories have been suggested to understand the saving pattern in China, the Chinese saving puzzle remains unsolved. In this paper, I aim to understand the household saving pattern in China as related to the social welfare level. Using data from National Bureau of Statistics (NBS), World Development Indicators (WDI) as well as findings from existing studies, I will analyze the Chinese household saving pattern both quantitatively and qualitatively. Although there are still many complex issues to be studied, I conclude from this study that there is a correlation between household saving rate and the social welfare level, and that due to
the significant differences between the saving patterns of urban and rural households, urban household saving rate and rural household saving rate are impacted by different areas of social welfare. Introduction I. The Chinese saving puzzle The Chinese saving pattern has always been a great puzzle to many scholars. Along with the rapid economic growth of China over the past three decades, its aggregate saving rate has increased from 35 percent of gross domestic product (GDP) in 1983 to 49 percent in 2015, according to the data from WDI that defines gross saving as gross national
income less total consumption, plus net transfers 1 based on data from national income accounts. China s entry into the World Trade Organization (WTO) in 2001 had a huge impact on the country s saving rate, as can be seen from a surge of saving rate from 37 percent of GDP in 2002 to 52 percent in 2009, because of a large inflow of foreign direct investment which caused a rapid capital accumulation. Figure 1 shows a comparison of aggregate saving rate among the BRIC nations, world level, upper middle income group, and East Asia and Pacific area from 1976 to 2015 (WDI, World Bank). Of all time, China s high aggregate saving, well above the world level, stays on the top of all the other countries and country groups in the graph. The components of aggregate saving are household saving, corporate saving, and government. In this paper, only household saving is studied because if social welfare policies have impact on the saving pattern, their impacts are most likely to be seen on households. Figure 2a gathers the data of rural and urban household saving rates. These data are derived indirectly from the data of per capita disposable income and per capita 1 Definition from World Development Indicators (data.worldbank.org): gross savings are calculated as gross national income less total consumption, plus net transfers
expenditure which are directly available from National Bureau of Statistics (NBS). Per capita household saving is calculated as per capita disposable income less per capita expenditure, and saving rate is calculated as per capita saving divided by per capita disposable income. Since there are some missing data which should have been used to derive the urban saving rate from 1995 to 2001 that are not accessible on NBS, I use instead the direct data from the paper of Dennis Yang et al, Why are saving rates so high in China (2011). Figure 2b shows graphically the saving pattern of urban and rural household saving which is marked as strikingly different. Whereas urban saving rate has been increasing over the years from the year 1995, rural household saving rate, after a surge from 1995 to 1999, was on the decline until 2006, fell below urban household saving rate in 2005, and has stayed below urban household saving rate ever since. According to Dennis Yang et al. (2011), in late 1970s, household saving accounted for only around 7 percent of GDP whereas in 2007 it has increased to around 22 percent. II. Existing studies
The Chinese saving puzzle has been of great interest to many scholars. Qian (1988) conducted research that analyzed saving patterns of different sectors from year 1978 to 1984, and Kraay (2000) renewed the study of Qian to cover the year from 1978 to 1995. According to Qian and Kraay in their respective studies, household savings only accounted for 6 to 7 percent of the GDP in late 1970s (in contrast to 22 percent in 2007). They also found that both government savings and corporate savings have stayed at approximately the same level at near 30 percent during the time span of their studies. Kraay suggested that this persistently high government and corporate saving rate after the Chinese reforms in 1970s and 1980s is caused by different strategies in the central planning which focused on government investment to encourage enterprise sector. During the time span of Kraay s study, household saving did not account for a large part of gross national saving. The more recent study of Song and Yang (2010) is aimed at explaining the stylized patterns of the household savings in China with data from Chinese Urban Household Survey from the year of 1992 to 2007. They found three obvious changes in the saving
pattern of Chinese households in order to adapt to the new and fast-growing environment. First of all, younger worker cohorts have the tendency to save more over the years. Second, the aggregate pension replacement rate 2 has declined from approximately 80 percent in the early 1990s to approximately 58 percent in 2007 (Song and Yang, 2010). Third, the individual age earning profiles have flattened over the past 20 years. Another recent study of Ge, Yang, and Zhang (2010) shows the correlation between the so called one-child policy and household saving patterns. By building an overlapping generation model, they concluded that the birth control policy has correlated very positively to the household saving rate. They argued that the one child policy has contributed significantly to the increase in household saving rate from late 1990s onwards. Yang, et al (2011) have conducted a research trying to explain the high saving rate in China. In the research, they divided focus into the three components of aggregate saving: corporate saving, government saving, and household saving. They pointed out that the household saving is positively related to the income of the household. More 2 Defined as the ratio of average pension per retiree to average wages per worker in specific years
specifically, they showed the demographic saving patterns: the saving rates of the lowest quantile of the income group have remained between 5 percent to 10 percent in most years, whereas that of the highest income quantile has increased steadily and rapidly over the years, jumping from 10 percent in 1988 to 34 percent in 2007. They compared their study to that of Dynan el al (2004) and found that the pattern of higher saving rate among the highest income group is consistent with that of the household saving patterns in developed countries. They then concluded that for China, the persistent increase in saving rate might be related to the growing income inequality over the years after the economic reforms. III. Research method Having reviewed many existing studies, I have found that almost all of them focused on explaining the saving rate in a broad spectrum of factors. In this paper, I intend to understand how household saving is influenced by social welfare policies. My research method includes two directions: qualitative analysis and quantitative analysis. Through
the qualitative method, I conduct a series of literature review of existing studies as well as of recent major social welfare policy changes in China. Through the quantitative method, I try to quantify social welfare level using different economic indicators and the run regression of household saving rate on these indicators to analyze the correlations.
Qualitative findings The study of Chamon and Prasad (2010) has shown that the age-saving profile is becoming U-shaped as saving rate has dramatically increased since 1990s, suggesting saving being the highest among the youngest and oldest households. They postulated since this increase coincided with the dramatic economic change of China that has transformed from a closed economy to a market-oriented economy and from agriculturally-focused to industry-driven, the U-shaped profile can be explained by growing income uncertainty and inequality as well as pension reforms. Using data from the China Health and Nutrition Survey (CHNS), Chamon and Prasad found that prior to the pension reform, the pensions received by urban workers from their employers (which were mostly SOEs) had a replacement ratio of around 75 to 80 percent relative to the average wage. Under the reform which benefited workers who retire after the year 1997, workers receive a social pension about 20 percent of the average local wages. Therefore, they claimed that the replacement ratio for the transition pension should be about 60
percent of the average wage (Chamon and Prasad, 2010). By building a calibration model, Chamon and Prasad managed to show that 6-8 percentage points of increase in the household saving rate can be explained by a decrease of the replacement ratio from 75 percent to 60 percent. The model shows that both increasing income uncertainty and pension reforms can be used to explain the U-shape age-profile patttern in urban household savings in China. The study of Yang et al (2011) also shows interest in the U-shape age-profile in household saving pattern in China. Yang pointed out that the households with older people have experienced faster growth in saving that the overall population since the late 1990s. Like Chamon and Prasad in their study, Yang also found out that this pattern can be explained by the decline in pension incomes for retired workers so that those families with older people tend to save more in order to maintain their daily consumption. The fact that families with children tend to save less can be explained by the rapid increase in costs associated with raising children over the years. He further explained that even though the income for households with children has increased over the years, the
increase in their expenditures on raising their children, such as education costs, nourishment, travel, etc, has offset the increase in income, thus dragging down their savings in respect to those households without children. While Chamon and Prasad as well as Yang et al are interested in studying the demographics of households to analyze household saving patterns, Nabar (2011) has pointed out that the household saving rate is also correlated with interest rates. Using the panel data of saving rates from China s provinces over the years from 1996 to 2009, Nabar found that the urban household savings is negatively correlated to real interest rates over this period, and the correlation is particularly strong for the years after 2003, suggesting that the Chinese household saving pattern is driven by a precautionary mindset : Chinese households save with a target level of saving in mind. When the return to saving declines (increases), it becomes more difficult (easier) to meet a target and households increase (lower) their saving out of current disposable income to compensate (Nabar, 2011). He suggested that because of the negative correlation between saving
rate and real interest rate, the government can increase real deposit rates in order to help lower household saving and therefore, increase domestic consumption. Like Nabar (2011), Carroll and Weil (1994) also pointed out habit formation as an alternative explanation of the rising household saving despite the rapid income growth. They suggested that the so-called consumption inertia is more of a historical and cultural social behavior. Their argument that the low consumption growth of Chinese household has lagged behind the income growth during the years of economic and social reform leading to a growth in saving rate is evidenced by empirical findings that variations in household savings by provinces over time and space are influenced by the lagged saving rates, a result that is in line with the theory of inertia or persistence (Horioka and Wan, 2007). However, the habit theory is not very convincing according to the study of Modigliani and Cao (2004). They showed that for the two decades from mid 1950s to mid 1970s, the household saving rates have stayed very low (below 5 percent) and the dramatic increase occurred during the years of reform. There is no data or any other empirical evidence that can how the assumed positive correlation between current
consumption growth and that in the past. Zhou (2007) went even further by completely rejecting the thrifty Chinese factor as a significant explanation of household saving puzzle. Using the data from CHNS from year year of 1988 to 2003 and after controlling for other factors that may affect saving rate, Zhou finds that younger cohorts tend to save more than older cohorts who are believed to carry usually more cultural habit or tradition than the younger generation. Welfare reforms since the late 1990s have included health care, pension reforms, unemployment insurance, and maternity benefits. In 2006, agricultural tax has been abolished so that the disposable income for rural households who depend largely on agriculture, has increased. As Yang (2011) has pointed out, household income is positively correlated with saving rate. Thus the increase in disposable income leads to an increase in saving rate after 2006 which can be seen from Figure 2b. Furthermore, in terms of the social relief in urban areas, a policy of providing basic living allowances to subsidize the living costs of those who live under the poverty line (which is defined locally) was carried out in 1999, and in 2004, was extended to rural areas with some modification
to cater for the specific situations for rural households. In 2003, cooperative medical care was introduced to rural areas, under the scheme of which medical bills that are usually quite expensive are covered by a cooperative fund that covers those who have signed up for the cooperative medical care program. Each individual in a rural household as well as the local and central governments give some money (different for individuals and for the government) to the fund. Quantitative findings I. Quantifying social welfare level There are three main areas of social welfare to be measured: poverty reduction, education, and medical aid. In terms of poverty reduction, I use indicators of GINI index that indicates the income inequality in a country, the percentage of urban as well as rural population receiving minimum living allowance which are calculated indirectly from direct data of National Bureau of Statistics, and unemployment rate from World Bank. In terms of education level, I use the data from WDI that include government education
expenditure and gross secondary enrollment ratio. To measure medical aid, I use the indicator of the percentage of urban and rural population receiving medical aid which are calculated indirectly from data from NBS, as well as government healthcare expenditure which is extracted directly from NBS. Because of the striking difference between rural household saving pattern and urban household saving pattern, I build to separate regression models to analyze the impact of social welfare level on urban households and rural households respectively. II. Urban household saving 1. Urban household saving as explained by labor market, education level and poverty By running a multi-variable regression of urban household saving rate on unemployment rate, government education expense and the percentage of urban population that receives minimum living allowance. The regression results can be found in figure 3. The R-squared of 97.59% and adjusted R-squared of 97.04% suggested the
relative success of the model. The p-values of the coefficients suggest their high significance. The interpretation of the result is rather intuitive. Unemployment rate has a negative correlation with urban household saving rate because as more people in the society are unemployed, the overall income level declines and as suggested by many existing studies, income decline leads to a decline in saving rate, therefore explaining the negative sign of the coefficient of unemployment on urban household saving rate. The positive correlation between government education expense and urban household saving rate can be intuitively explained as well. As government spends more on education, households with children can spend less (because of education subsidies, government scholarships, etc) on education and thus save more. The positive correlation between the percentage of urban population that receives minimum living allowance and urban household saving rate is, on the other hand, rather ambiguous. While the intuition goes that as more people are receiving the minimum living allowance, it is a sign of rising poverty level, and thus will lead to less saving since income level is positively correlated
with saving rate. However, we fail to take into account the fact that as more people receive the minimum living allowance, the disposable income of these people increases and thus have more to save. The cultural habit can partially explain this phenomenon: the poor population who live on so little money tend to save the extra money they get and only spend the absolute essentials such as inexpensive food and drink, and save most of the extra money in fear of further poverty. In conclusion, this regression model shows that the key influences on the saving pattern of urban households are the labor market, education costs, and poverty level. 2. Urban household saving as explained by the inequality of income distribution According to Yang et al (2011), the saving rate of the highest income quantile has increased rapidly and dramatically from around 10 percent in 1988 to around 35 percent in 2007 whereas the saving rate of the lowest income quantile has stayed roughly the same or varied only in a small range during the same time span. Therefore, I am also interested in running another regression of the urban household saving rate on GINI index which is an indicator of income inequality. The regression results are shown in figure 4.
The R-squared of 76.99% and adjusted R-squared of 75.46% suggest the relative success of the explanatory power of model. The coefficient of GINI index, which is highly significant as suggested by the p-value, shows that 1 point of increase in GINI index increases the urban household saving rate by roughly 0.93 percentage point. Therefore, the impact of income inequality on household saving can be corroborated by the model. The implication of this finding is that for China where the growing inequality in income distribution brought about by the years of economic transition, income transfers from the rich to the poor may help to enhance the propensity of consumption. III. Rural household saving rate By running a simple regression of rural household saving rate on the percentage of rural population that receives medical aid. The regression results can be found in figure 5. The R-squared of 79.33% and adjusted R-squared of 75.88% suggested the relative success of the model. The p-values of the coefficient shows that the factor is highly significant.
The positive sign of the coefficient of the percentage of rural population that receives medical aid is rather intuitive. As more individuals receive medical aid in rural area, households can spend less on medical bills and hospitalization, and thus have more to save. Since the regression model is percentage point to percentage point, the coefficient factor of 0.67 implies a rather high impact of medical aid on rural saving rate whereas it does not play a significant role in urban household saving. I also ran a regression of the rural household saving rate on unemployment rate and education, and it is very counterintuitive to find that neither the labor market nor the education level has significant impact on the saving rate in rural households. A tentative explanation is that mot rural population lives on agricultural activities, so that their daily lives are not significantly influenced by the labor market. The recent increase in the population of migrant workers that come from the countryside to big cities to find low-paid jobs may increase the impact of labor market on saving rate in the near future. Besides, since the education cost in rural area is already very low (almost free), government expenditure on education will only have very little marginal effect on their disposable income. However,
since the lack of data available in that only 8 data points are available from the NBS, the explanatory power the model is inconclusive.
Conclusion This paper aims at understanding the impact of social welfare policies on household saving rate in China. The reasons behind the Chinese saving puzzle are complicated. By conducting both literature review and quantitative analysis, I find a strong difference between the saving patterns of urban households and rural households. Social welfare programs regarding employment, education and income distribution have a much more significant impact on urban household savings, while those regarding medical aid have more impact on rural household savings. Yang et al (2011) also pointed out that household saving in China will be reduced by the inevitable slowdown in the labor earnings growth as well as the gradual steepening of the age-earning profile. Song and Yang has further suggested that, despite that fact the existing pension contributions under the three-pillared system have not yet met the
targeted levels, the pension system will become better to meet the levels given the government s intent to build a welfare state. Therefore, a much more comprehensive pension system will have two important consequences. First of all, since more people are insured with higher level of pension replacement ratio and are therefore exposed to less retirement risk, they tend to have less incentive to save a lot during their working time. Furthermore, according to Yang, a more robust implementation of the three-pillared pension system that is expected to raise the pension contribution from employers from the 2007 level of 5 percent to the target of 17 percent wage taxes have the potential to reduce both corporate saving and household saving.
Reference Carroll, Christopher D. and David N. Weil. 1994. Saving and growth: A Reinterpretation, Carnegie-Rochester Conference Series on Public Policy 40: 133-192. Chamon, Marcos D. and Prasad, Eswar S. 2010. Why Are Saving Rates of Urban Households in China Rising? American Economic Journal: Macroeconomics, 2(1): 93-130.
Ge, Suqin, Yang, Dennis Tao and Zhang, Junsen. 2010. One-Child Policy and the Chinese Household Saving Puzzle: A Cohort Analysis, Working Paper, The Chinese University of Hong Kong. Horioka, Charles Yuji and Junmin Wan. The Determinants of Household Saving in China: A Dynamic Panel Analysis of Provincial Data, Journal of Money, Credit and Banking 39(8): 2077-2096. Kraay, Aart. 2000. Household Saving in China, World Bank Economic Review 14(3): 545-70. Modigliani, Franco. 1970. The Life Cycle Hypothesis of Saving and Intercountry Differences in the Saving Ratio, in Induction, Growth and Trade, eds. W. A. Eltis, M. FG. Scott, and J. N. Wolfe. Oxford: Clarendon Press, 197-225 Modigliani, Franco and Shi Larry Cao. 2004. "The Chinese Saving Puzzle and the Life- Cycle Hypothesis," Journal of Economic Literature 42(1): 145-170. Nabar, Malhar. 2011. Targets, Interest Rates, and Household Saving in Urban China, Working Paper, International Monetary Fund Qian, Yingyi. 1988. Urban and Rural Household Saving in China, International Monetary Fund Staff Papers 35(4): 592-627. Song, Zheng Michael and Yang, Dennis Tao. 2010. Life Cycle Earnings and Saving in a Fast-Growing Economy. Working Paper, Chinese University of Hong Kong. Yang, Denni Tao and Zhang, Junsen and Zhou, Shaojie. 2011. Why Are Saving Rates So High in China?, IZA DP No. 5465 Zhou, Shaojie. 2007. Essays on Household Consumption and Household Saving Behavior of Chinese Urban Residents, Ph.D. Dissertation, The Chinese University of Hong Kong. Appendix 1: Tables and graphs
Source: World Development Indicators Figure 1: The aggregate saving rate (as % of GDP) of China, Brazil, Russia, India, East Asia & Pacific (all income levels, aggregate), World, and Upper middle income (aggregate) (legend from left to right)
Figure 2a: Household saving Figure 2b: Household saving
. reg Urban_SR Unemployment_Rate Gov_Edu_Expense Urban_MLA Source SS df MS Number of obs = 17 F(3, 13) = 175.81 Model.042152351 3.014050784 Prob > F = 0.0000 Residual.00103898 13.000079922 R-squared = 0.9759 Adj R-squared = 0.9704 Total.043191331 16.002699458 Root MSE =.00894 Urban_SR Coef. Std. Err. t P> t [95% Conf. Interval] Unemployment_Rate -.0335574.0135173-2.48 0.027 -.0627598 -.004355 Gov_Edu_Expense 5.82e-10 3.85e-11 15.12 0.000 4.99e-10 6.65e-10 Urban_MLA 1.084843.2273 4.77 0.000.5937915 1.575895 _cons.3124027.0633546 4.93 0.000.1755333.449272 Figure 3: Regression result 1. reg Urban_SR GINI Source SS df MS Number of obs = 17 F(1, 15) = 50.20 Model.033254184 1.033254184 Prob > F = 0.0000 Residual.009937147 15.000662476 R-squared = 0.7699 Adj R-squared = 0.7546 Total.043191331 16.002699458 Root MSE =.02574 Urban_SR Coef. Std. Err. t P> t [95% Conf. Interval] GINI.9334231.131747 7.08 0.000.6526111 1.214235 _cons -.1775987.0589264-3.01 0.009 -.3031974 -.0519999 Figure 4: Regression result 2
. reg Rural_SR Med_Rural Source SS df MS Number of obs = 8 F(1, 6) = 23.02 Model.00197417 1.00197417 Prob > F = 0.0030 Residual.000514535 6.000085756 R-squared = 0.7933 Adj R-squared = 0.7588 Total.002488706 7.000355529 Root MSE =.00926 Rural_SR Coef. Std. Err. t P> t [95% Conf. Interval] Med_Rural.6790451.1415267 4.80 0.003.3327417 1.025348 _cons.2011876.0075064 26.80 0.000.1828202.219555 Figure 5: Regression result 3 Appendix 2: Data
Table 1: Health expenditure data Table 2: Social relief data Table 3: Compiled data for regression (1) Data source: WDI
Table 4: Compiled data for regression (2) Data source: United Nations, WDI, NBS