Predictors of Support for State Social Welfare Provision in Russia and China

Abstract This essay analyses the determinants of support for state social welfare provision in Russia and China on the basis of a four-stage recursive model using two waves of the World Values Survey. It hypothesises that support is a function of economic self-interest, tapped by subjective economic satisfaction and relative income; ideology including beliefs about market fairness and inequality aversion; as well as temporal context. It finds that subjective economic satisfaction reduces support; inequality aversion is a positive influence, while beliefs about market fairness matter in different ways. Support increased over the period spanning the 2008 global financial crisis.

Welfare guarantees under the command economy were never as comprehensive in China as in the Soviet Union. In 1978, more than 80% of China's population lived in rural areas, compared to around 30% in Russia. Chinese peasants in the 1960s and 1970s were subject to extractive policies amounting to a kind of 'socialist serfdom' (Whyte 2010, p. 20), equivalent to Soviet rural policies in the phase of forced growth before World War II (Nove 1992, pp. 271-75). Their freedom of movement being restricted, by the time reforms started, the great mass of the Chinese population had no experience of the kind of developed, stable command economy which the Soviet Union established after the war and which had begun to take shape in urban China (Whyte 2010, chapter 2). Gao and Riskin report that from 1988 to 2002 the share of social benefits in China's urban incomes fell from 44% to 25%, but the share of social benefits in rural incomes remained under 1% throughout this period (Gao & Riskin 2009, pp. 25, 31).
The manner of reforming the command economy in China probably did much to legitimise pro-market welfare values, whilst in Russia transition had precisely the opposite effect (Denisova et al. 2010;Grosfeld & Senik 2010). To some extent, this contrast is likely to reflect the fact that market reforms in China took place over a longer period, showing the Chinese Communist Party's (CCP) determination to experiment at a small scale before committing to an overall plan (Nolan 1995, p. 308). Soviet economists prepared various comprehensive reform blueprints, but none were implemented until after the break-up of the USSR in 1991. Amidst a civilisational crisis, reforms then proceeded very rapidly, in order to exploit a window of political opportunity (Hough 1997). Although apologists for shock therapy would argue that it was never fully implemented in Russia (Åslund 2002), it is doubtful that the general population understood the difference between well-executed and abortive shock therapy. After 1992, Russia lost the best part of a decade of economic growth, with a prolonged period of recession lasting from the start of transformation to the end of 1996 and particularly sharp falls in 1993 (-8.7%) and 1994 (-12.7%). China, by contrast, maintained an average growth rate of 9. 8% from 1978 to 2000, only stuttering in 1989 and 1990 with growth rates around 4% because of the Tiananmen massacre and subsequent clampdown. 1 If we consider the historical legacies, therefore, it is no surprise to find that the comparative data assembled by Whyte (2010, chapter 4) suggest that Chinese in 2004 were generally more pro-market than East Europeans and Russians of the mid-1990s.
For reasons just described, social scientists, particularly those who work on China, are apt to dismiss similarities between the Russian and Chinese transformations. Writing in the introduction to a major edited volume on social inequality in China, Davis and Wang for instance, write 'in almost every parameter China's postsocialist trajectory does not parallel those of Russia and Eastern Europe in initial outcomes or pace of change' (Davis & Wang 2009, p. 3). In spite of these differences, in terms of social policy, the challenges the two countries face are quite similar, including regressive and poorly targeted social welfare programmes, tax systems which likewise put a disproportionate burden on lower and lowermiddle income households in the formal sector, co-opted trade unions, widespread corruption, large regional as well as urban-rural inequalities and social exclusion (Remington 2015, p. 8). It is this common set of problems that makes the two countries interesting cases for comparison and contrast.
Extant research on support for state social welfare provision in China and Russia focusses on economic self-interest variables and ideological determinants (Kluegel et al. 1999;Ravallion & Lokshin 2000;Whyte 2010), but relies on data collected at different time periods using different questionnaires and measures, which makes it difficult to obtain a comparative perspective. They also do not cover the period spanning the global financial crisis of [2008][2009], which affected the economies of both countries.
This study approaches the dynamics of support for state social welfare provision through statistical models constructed using survey data from two roughly contemporaneous surveys from each country using a common questionnaire. Time is treated here primarily as an 'event' variable, measured in months before and after the global financial crisis. Economic selfinterest is tapped by income and satisfaction with household economies, controlling for social structural characteristics. Ideological influences include beliefs about the way the market economy works as well as normative aversion to inequality. In order to model indirect effects, the regressions presented are recursive, with household economic satisfaction, beliefs about the market and inequality aversion assumed to mediate between social structural variables and support for state social welfare provision.
This study is the first to take a systematic approach to comparing Russian and Chinese support for state social welfare in the light of current theories. It shows that subjective economic satisfaction reduces support. Although relative income has the same effect in Russia, it is not strong, and in China, those with higher incomes are actually slightly more in favour of social welfare provision once indirect effects are taken into account. In both countries, ideology matters most, with inequality aversion emerging as the strongest single influence on support for social welfare provision. Beliefs about market fairness matter in different ways: attribution of success to non-merit factors emerges as a negative influence in China while the belief that getting rich is a zero-sum game is a positive influence in both countries. Finally, the 2008-2009 financial crisis or events contemporaneous with it had a positive impact on support for state social welfare in Russia and China. This impact cannot be explained by reference to economic self-interest or ideology alone.
The remainder of this essay is structured as follows: the next section applies concepts from the literature on support for state social welfare provision to develop four hypotheses. The description of data, measures and methods follows. The results section presents the recursive models and describes overall effects. The discussion gives an interpretation and acknowledges limitations of the study. The conclusion considers implications of the findings for political and policy change and makes some suggestions for further research.

Hypotheses
Based largely on the experience of Western countries, most studies of distributive justice attitudes suggest that citizens in poorer households are more likely to support government intervention to redistribute income in favour of the less well-off, typically by using progressive taxes to fund social welfare programmes (Svallfors 1997(Svallfors , 2006Cusack et al. 2006;Rehm 2009Rehm , 2011Margalit 2013). Using data collected in Russia in the mid-1990s, Kluegel et al. (1999, p. 272) showed that support for government intervention to equalise living standards was negatively related to current and prospective evaluations of subjective living standards, and the same pattern was found also in Bulgaria, the Czech Republic, East Germany and Hungary. Ravallion and Lokshin (2000, p. 100) showed that amongst those who expected to live better or expected no change in their circumstances, support for restricting the incomes of the rich (and, implicitly, raising the relative incomes of the poor) was negatively related to household expenditure and to upward consumption trajectory, and positively related to concern about providing for basic needs. In China, a regression analysis of support for government intervention to equalise living standards showed that income was insignificant, subjective changes in the standard of living were a marginal positive influence, but a measure of relative social status comparing current living standards to those of one's peers proved to be a strong negative influence (Han & Whyte 2009, pp. 201-3). The balance of evidence seems to suggest that the general association of lower income and economic dissatisfaction with support for programmes designed to equalise incomes or to reduce poverty seems to hold in both Russia and China, although the precise contours of the relationship depend on measurement strategies and what other variables are included as controls. Setting these details aside for the moment, we would expect that in both countries: H1. Households in worse economic circumstances are more likely to support state social welfare provision.
The literature on distributive justice attitudes pays considerable attention to ideological sources, such as beliefs about fairness in the way the economy works (Kluegel & Smith 1986;Fong 2001;Alesina & Angelotos 2005). Kluegel et al. (1999) as well as Whyte (2010) proceeded from the expectation, that if respondents believed the economy worked in a fair way, they would be less inclined to support government intervention to equalise living standards. Some of Kluegel et al.'s (1999) findings support this notion. For instance, they constructed a measure of 'negative attributions for wealth' based on two items, the belief that the rich are dishonest and that the economic system allows them to take unfair advantage of others. In four out of the five post-Communist countries in their study, including Russia, negative attributions of wealth had significant positive coefficients when regressed on support for government intervention to equalise incomes (Kluegel et al. 1999, p. 272).
However, not all of Kluegel et al.'s (1999) findings fit neatly into this understanding of the relationship between perceived fairness and support for government intervention. For example, they deployed a measure of 'legitimate inequality' based on four items: the belief that people have opportunities to get ahead, that they are rewarded for their efforts, that they are rewarded for intelligence and skill, and that they get what they need. This measure had no effect on support for government intervention in Russia, the Czech Republic and East Germany, was a positive influence in Bulgaria, and a negative influence only in Hungary. Differences in ideological context would lead us to expect that the relationship between perceived unfairness and support for state welfare might be different in different countries. Nevertheless, given that state social welfare provision is normally perceived as a way of ameliorating unfairness, it is justifiable to hypothesise that in both Russia and China: H2. Beliefs that the market economy works in an unfair way increase support for state social welfare provision.
Using nationwide survey data from Hungary in 2003, Toth (2008Toth ( , p. 1081) demonstrated an association between inequality aversion and an index of support for state involvement in various areas of social provision and the economy. The justification for exploring this relationship is that net of self-interest and beliefs about how the economy works, citizens may still have an ideological preference for income equality which is founded in their view of how 'the good society' should be organised. Ravallion and Lokshin (2000, p. 100) captured a similar ideological effect when they showed that in Russia in 1996 Communist Party voters were more likely to support redistribution even if their living standards had improved or undergone no change. It could be argued that inequality aversion and support for state social welfare provision are really 'the same thing', but this would be to foreclose a discussion about whether inequality aversion justifies state welfare programmes. Because state social welfare programmes are seen as a remedy to inequality, rather than the other way around, it is justifiable to hypothesise that support for state social welfare responds to inequality aversion. We can thus offer as a third hypothesis that in both Russia and China: H3. Inequality aversion increases support for state social welfare provision.
The 2008-2009 financial crisis had a relatively severe impact on the Russian economy (Drahokoupil & Myant 2010, pp. 281;Gaddy & Ickes 2010;Sapir 2010). The ruble lost 25% of its value between October 2008 and October 2009, affecting food prices amongst other imports. Reflecting weaker prices for commodities, the value of Russia's exports as a percentage of GDP fell by 47%, year on year, in the first half of 2009. Although the Russian state had sufficient reserves to cover its liabilities, many private companies did not, with the result that the state effectively bailed them out and increased control over them. Growth slowed in 2008 (5.2%) and turned negative during 2009 (-7.8%) before returning in 2010 (4.5%), 2011 (4.3%) and 2012 (3.4%).
In China the government acted quickly to pre-empt the impact of the crisis as it happened (De Haan 2010; Yee 2012). Nevertheless, the crisis did have important economic consequences for China. Specifically, this was reflected in the following (De Haan 2010, pp. 764). Tens of thousands of factories closed down and by February 2009 around 20 million workers had lost jobs, with around half of them returning to rural areas, in accordance with the traditional role of the countryside as China's 'safety valve'. To pre-empt further factory closures, the government introduced a stimulus package worth over 12% of GDP over two years, loosened credit policy, cut some taxes and issued vouchers to subsidise consumer goods purchases. As a result of the forceful policy response, the growth rate barely stuttered during this period, falling from 9.6% in 2008 to 9.2% in 2009, before surging back to 10.2% in 2010 and then slowing to 9.3% in 2011 and 7.7% in 2012.
The effects of the global economic crisis on income distribution appear to have been relatively minor in both countries. National statistical authorities reported that Gini coefficients in Russia hovered between 42. Net of changes in household economic circumstances and ideological beliefs, 'events' including both factors outside governments' control as well as their policy responses could have influenced the climate of opinion over the period leading up to the crisis and thus cued the population to react by increasing support for state social welfare provision. Putin came to power in 1999 promising continuity to the principal beneficiaries of market reform, on condition that they stayed out of politics, and also stability to the majority of the population who were still recovering from the effects of shock therapy (Rose & Munro 2002). State finances were greatly boosted by a roughly $100/barrel rise in oil prices over the course of Putin's second term, allowing the regime to accumulate foreign reserves, modernise the military and still be able to afford regular increases in pensions and other benefits. A 2007 survey showed 53% of Russians credited the Putin regime with improving social protection of the poor (Rose et al. 2011, p. 52). During this period, the regime had also partially revived Soviet ideological motifs to capitalise on popular nostalgia, and the idealisation of the past became one of the regime's 'signature tunes' (Munro 2006).
Similarly, in China the global financial crisis came against the background of a larger ideological turn. The period following Deng's 'Southern Tour' of economically advanced coastal provinces in 1992 was the apogee of market reform, culminating at the turn of the century in accelerated reforms to prepare for WTO accession. These involved an end of guaranteed employment in state-owned enterprises, and retrenchment in their work force from 76 million in 1992-1993 to 28 million by 2004 (Naughton 2008, p. 121). In the early 2000s, the new President Hu Jintao proposed the 'scientific development concept' and the notion of a 'harmonious society' as his main contributions to the development of CCP doctrine, signalling the regime's understanding of the need for an ideological palliative (Holbig 2009). 5 Although Hu Jintao may have faced resistance to his new concepts from supporters of the previous president, Jiang Zemin, eventually they were included in the 11th Five Year Plan (2006)(2007)(2008)(2009)(2010)(2011) and in 2007 made part of the CCP constitution at the 17th Party Congress. We would therefore expect that these concepts would have had some influence on Chinese public opinion during the period spanning the financial crisis. Given the traumatic nature of market reform, the emergence of compensatory ideological pressures, and the negative impact of the financial crisis on specific sectors of the economy, we would expect that in both Russia and China: H4. Events over the period spanning the financial crisis increased support for state social welfare provision.
In the post-Communist world, differences of age, gender, education and sometimes place of residence have been shown to affect distributive justice attitudes. For example, Kluegel et al. (1999, p. 263) found that in Russia in 1996 support for government action to equalise incomes was stronger amongst older and less educated citizens. 6 Ravallion and Lokshin (2000, p. 101) confirmed these results and showed in addition that women and those living in rural areas were more supportive of redistribution, even if they expected their living standards to improve or expect no change. In China, by contrast, age has been found to make no difference to support for government intervention to equalise incomes, education is a positive influence, whilst female gender and rural residence are negative influences, controlling for differences in economic circumstances (Han & Whyte 2009, p. 203). Because social structure affects support for social welfare provision and related attitudes differently in Russia and China, we need to control for social structural position in terms of age, gender, education and settlement size when attempting to measure the effects of economic self-interest, ideology and temporal context. The measure of support for state social welfare provision used in a WVS is the following question:

Data and measurement
How would you place your views on this scale? 1 means you agree completely with the statement on the left; 10 means you agree completely with the statement on the right; and if your views fall somewhere in between, you can choose any number in between. People should take more responsibility to provide for themselves OR the government should take more responsibility to ensure that everyone is provided for.
Distributions for China and Russia in the two most recent waves are shown in Figure 1. A notable feature of the Russian distribution is that it is skewed towards state social welfare provision, with more than two fifths selecting the extreme end of support in each year. Chinese respondents are much more evenly distributed along the continuum. 6 See also Stephenson and Khakhulina (2000, p. 89 5.7 (s.d. 2.9). Then between 1999 and 2006 there was another jump in support for state social welfare provision to 7.6 (s.d. 2.8), followed by a further small rise to 7.9 (s.d. 2.6) in 2011 (p<0.05). The overall trend has been up in Russia, albeit with a dip below the trend line in 1999.
A survey conducted by Whyte in China in 2009 as a follow-up to his 2004 survey appears to confirm rising support for government welfare responsibility, as measured on a six-item scale of agreement that the government should be more responsible for health care, primary and secondary education, university education, employment, housing and care for the elderly (Whyte 2010;Whyte & Im 2014). The mean rose from 3.0 to 3.5 (p<0.001) (Whyte & Im 2014, pp. 66-7). Whyte's surveys also show rising support for government intervention to equalise incomes as measured by agreement or disagreement on a five-point scale with the proposition that 'distributing wealth and income equally among people is the most just method'. Mean agreement with this proposition increased between 2004 and 2009 from 2.8 to 3.0 (p<0.001). Similarly, the surveys showed rising trends in support for government levelling, a threeitem mean scale tapping agreement with the propositions that government should provide a minimum living standard for all, guarantee jobs for those who need them, and reduce the gaps between the rich and the poor. The mean for this measure rose from 3.9 to 4.0 (p<0.05). Whyte's surveys show that alongside the growth in pro-state welfare attitudes, the numbers of respondents who disagreed with the idea that it is fair that the rich should be able to get better schooling for their children, better housing and better medical care also increased from 2.5 to 2.7 (p<0.001) (Whyte & Im 2014, p. 66). In sum, surveys conducted around the same time would seem to validate the observation based on the WVS data that distributive justice attitudes were moving in favour of state social welfare provision in China over this period.
However, confirmatory evidence for a rising trend in support for state social welfare in Russia is lacking. New Russia Barometer (NRB) 8 asked respondents to choose between two statements: 'Individuals should take responsibility for themselves and their livelihood OR the state should be responsible for everyone's material security'. In the fourteenth NRB survey of 3-23 January 2005 (N=2,107), 62% agreed with the second statement; in the eighteenth survey of 18-24 June 2009 (N=1,601) only 54% agreed with it. Similarly, in nine polls conducted between 1999 and 2014, the Levada Centre asked respondents to choose from a list of four statements, as follows: (1) 'The state should provide an adequate level of welfare to all citizens'; (2) 'the state should provide help to those who have found themselves in difficult circumstances, for example, after losing their job etc.'; (3) 'the state should provide help only to those who are not able to provide for themselves; pensioners, the disabled, orphans'; (4) 'people should look after themselves and provide an adequate life for themselves without any help from the state'. (Levada Centre 2015, p. 24) If we calculate 'net support for state welfare provision' by adding the percentages agreeing with the first two statements and subtracting the percentages agreeing with the second two statements, excluding don't knows, and plot them over time measured in months, the results show substantial short-term fluctuations: 58% net support in September 1999;63% in March 2000;57% in December 2004;57% in December 2005;63% in December 2006;61% in February 2008;65% in December 2010;56% in October 2011;and 55% in September 2014. There is an overall decline in support for state welfare, but there are three peaks above the trend line, in March 2000, December 2006and December 2010 In order to test the construct validity of the WVS measure of support for state social welfare provision it is necessary to relate it to a system of independent variables whose relationships to support for state social welfare provision are relatively well-documented. The panel labelled 'Controls' in Table 1 presents a selection of such variables. These include the basic demographic attributes of age and gender, as well as measures of education and town size. Age is capped at 75 years. Education is measured as the age at which the respondent completed formal education and, to avoid outlier effects, has a floor of ten years and a ceiling of 25 years. Town size is measured on an ordinal scale from one for settlements with a population of 2,000 or under to eight for populations of 500,000 or more. 9 Residence registration status (hukou) and residence in mono-towns are omitted even though they are likely to influence distributive justice attitudes in China and Russia respectively because they are not available on the WVS data file. Since they are country-specific variables, their inclusion could in any case confound comparison of generic effects between the two countries.
The variables in the middle panel of the table are in two groups. One is entitled 'economic self-interest' and includes self-assessed household income measured on a ten-point ordinal scale plus household economic satisfaction measured on a ten-point scale. The second is entitled 'ideological influences'. It includes two beliefs about the operation of the market economy, as described below, plus inequality aversion.
Each of the questions tapping beliefs about the market economy is introduced as per the question on support for state welfare: 'How would you place your views on this scale? 1 means you agree completely with the statement on the left; 10 means you agree completely with the statement on the right; and if your views fall somewhere in between, you can choose any number in between'. The first question taps the view that getting rich is a zero-sum game, requiring respondents to choose between 'Wealth can grow so that there's enough for everyone OR people can only get rich at the expense of others'. Chinese tend to disagree that getting rich is a zero-sum game. In the first WVS survey in China in 1990, the mean was 3.4 (s.d. We can conclude from this that first, Russians generally perceived the market as more of a zero-sum game than Chinese. Second, the trend in both countries has been to perceive increasingly zero-sum competition, albeit with some 9 For 2012, the China data file only codes settlement size as 100,000-500,000 people or more than 500,000 people, measured as points 7 and 8 on the scale. This corresponds to the population of the primary sampling unit (Personal communication with Professor Jie Yan, Research Centre for Contemporary China, Peking University, 20 December 2014). Given that the sampling method was identical in the 2007 survey, for which a more detailed breakdown of town size was provided, town size for China in 2012 has been centred on the 2007 mean. 10 The question was not asked in wave four, 2001. fluctuations. This is a gentler trend in China with the rise occurring between the mid-2000s and 2012, whilst the biggest rise in Russia occurs between the mid-1990s and mid-2000s. The second index of beliefs about the market economy is the one used by Alesina and Angelotos (2005) to tap the view that success depends on non-merit factors, requiring respondents to choose between 'In the long run, hard work usually brings a better life OR hard work doesn't bring success; it's more a matter of luck and connections'. In the first WVS survey in 1990, the Chinese mean was 5.0 (s.d. 3.1), in 1995 it was 3.5 (s.d. 2.8), 3.6 (s.d. 2.8) in 2007 and 3.7 (s.d. 2.4) in 2012. The Russian means were 4.1 (s.d. 2.9) in 1990, 4.7 (s.d. 3.1) in 1996, 5.8 (s.d. 2.9) in 2006 and 4.9 (s.d. 2.8) in 2011. We can conclude that first, Chinese are generally less likely than Russians to attribute success to non-merit-based factors. Second, the trends are in different directions: Chinese attribution of success to nonmerit factors fell after the turn towards the market in the early 1990s; Russians in the same period have been increasingly likely to attribute success to non-merit factors, although there are fluctuations about the trend line.
The WVS measure of inequality aversion has the same format as the questions above, with respondents asked to choose on a ten-point scale between two statements: 'Incomes should be made more equal OR we need larger income differences as incentives'. The coding is reversed so that higher values mean more inequality aversion. and they rise to 7.6 (s.d. 2.6) (p<0.001) for 2011. Both populations appear from these data to have become more averse to inequality. As with support for state social welfare provision, the trend line is up in both countries, but in Russia there are substantial fluctuations above and below the line. The 2011 survey in particular shows a substantial jump, which remains to be explained.

Methodology
In order to ensure that the modelling procedures are appropriate for a reasonable set of causal assumptions about the relationships in the data, this essay uses Mplus software to develop a two-level recursive model of support for state social welfare provision (Muthén & Muthén 2012). The analysis is conducted separately for each country but with identical coding so that coefficients can be compared between countries. The two surveys from each country are pooled in order to allow measurement of the impact of time. Parameter estimates take account of the fact that responses from a single survey are clustered by time of survey. To facilitate interpretation, all variables except gender are treated as continuous and centred on their grand mean. Female gender is treated as a dummy variable. Missing data are handled through the use of multiple imputation procedures in SPSS to create five complete datasets using fully conditional specification (Van Buuren et al. 2006). Parameters reported are thus pooled estimates from five imputations.
The control variables plus income are assumed to directly influence satisfaction with the household economy. The control variables as well as satisfaction with the household economy are assumed to influence beliefs about the fairness of the market, inequality aversion and support for state social welfare provision. Time is assumed to affect the intercepts of both inequality aversion and support for state social welfare provision. Time is measured in months elapsed since September 2008, with negative values indicating months before and positive values months after. It should be noted that, since the data are cross-sectional, there is no way to establish the direction of causality in the models. Hence, the strict interpretation of the regression coefficients is that they indicate the net strength of associations between variables given the set of causal assumptions outlined above.
Using cross-sectional data, it is impossible to prove that one attitude is causally or temporally prior to another. Attitudinal explanations typically proceed from some justifiable assumptions about how one attitude might influence another and then search for corroboration by testing the strength of the assumed relationship. Nevertheless it is important to acknowledge that the proof of this temporal or causal ordering awaits confirmation using more sophisticated research designs such as panel studies or experiments.
Although all elements of the models are computed simultaneously, for expositional purposes it is helpful to divide the analysis into two stages. The first stage analyses the mediating variables by regressing them on their assumed determinants. The second stage of the analysis systematically tests the hypotheses about influences on support for state social welfare provision by introducing the relevant independent variables in a stepwise fashion. Time is introduced first to establish the direction of the gross change in the climate of opinion on social welfare provision. Next, the control variables plus income are introduced as the first test of the first hypothesis, that economic self-interest drives support for welfare provision. Since people may be satisfied with different relative income levels, satisfaction with the household economy is introduced at the next stage, to see whether its introduction alters the effect of income. Simultaneously, beliefs about the fairness of the market are introduced to test the second hypothesis. Finally, inequality aversion is introduced as a test of the third hypothesis. Time is maintained in all the models to test the fourth hypothesis: that support for state social welfare provision was subject to a net change in the climate of opinion.

Results
For satisfaction with the household economy, income is the strongest influence in both Russia and China (b= 0.28*** and 0.23*** respectively, Table 2, top panel). 11 Education has different effects. It is a marginally negative influence on satisfaction in Russia (b= -0.02*), but it is a positive influence in China (b= 0.06***). Town size is a positive influence in Russia (b= 0.02***), but not significant in China.
Satisfaction with household economies reduces the belief that getting rich is a zero-sum game in both Russia and China (b= -0.13*** and -0.17*** respectively). In China, education has the same effect (b= -0.06***), but income has the opposite effect (b= 0.08***). In Russia, the belief that getting rich is a zero-sum game is more common amongst men than women (b= -0.10*).
Satisfaction with the household economy reduces the belief that success depends on nonmerit factors in Russia (b= -0.13***). There are no significant effects on this variable in China.
Satisfaction with the household economy reduces inequality aversion in both societies, though more so in China (b= -0.14***) than Russia (b= -0.04*). Women are more inequality averse in both Russia and China (b= 0.30** and 0.37** respectively), and, in Russia, age 11 Here and in the table estimates significant at 0.001 level are marked ***, at 0.01 level **, and at 0.05 level *. has the same effect (b= 0.02***). In both Russia and China, the belief that getting rich is a zero-sum game increases inequality aversion (b= 0.10** and 0.15*** respectively). In China attribution of success to non-merit factors has the opposite effect (b= -0.15***). Net of all individual influences, time exerts significant and positive effects on inequality aversion in both Russia and China (b= 0.04*** and 0.02*** respectively).
A series of models of support for state social welfare provision offer systematic tests of the four hypotheses (see Table 3). Consistent with hypothesis 1, economic self-interest matters,    but it is satisfaction with the household economy rather than income which matters more. In Russia, those who are satisfied are consistently less supportive of social welfare provision (b= -0.07*** in model III and b= -0.06*** in model IV). In China, satisfaction with the household economy loses significance when inequality aversion is included (b= -0.05** in model III, b= 0.01 in model IV), implying that in China economic self-interest matters through its effect on inequality aversion. Contrary to hypothesis 2, beliefs about the fairness of the market have no direct effect on support for state social welfare provision in Russia (models III and IV). In China, both beliefs about market fairness matter. The belief that getting rich is a zero-sum game increases support for state social welfare provision, as expected (b= 0.07**, model IV). But attribution of success to non-merit factors reduces support for state social welfare provision (b= -0.12***), just as it did for inequality aversion. In line with hypothesis 3, inequality aversion is a highly significant, positive influence in both China and Russia (b= 0.30*** and 0.29*** respectively, model IV). In line with hypothesis 4, time measured in months is a consistently positive and significant influence which is robust to the introduction of all other variables in both China and Russia (b= 0.01*** and 0.004* respectively, model IV).
In terms of control variables, in Russia, age and female gender are positive predictors of support for state social welfare provision (b= 0.01** and 0.15** respectively, model IV), whilst town size is a negative predictor (b= -0.04***), and education makes no difference. In China, by contrast, only town size matters, and it only emerges as a positive influence (0.31*) once inequality aversion is introduced as a control.
To see the overall patterns of determination of support for social welfare provision, we need to take into account indirect as well as direct effects. Figures 2 and 3 model the total direct and indirect effects of all the significant independent variables, computed by multiplying their coefficients by their mean values in each year. Figure 2 shows that, on the ten-point scale, in Russia age, gender, education and town size all have equal or nearly equal effects in the two years, ranging from -0.22 for town size to +0.70 for age. In line with hypothesis 1, taking into account indirect effects, both income and satisfaction with the household economy matter, reducing support for state social welfare provision by a total of -0.40 (-0.10-0.30) in 2006 and -0.42 (-0.07-0.35) in 2011. The belief that getting rich is a zero-sum game increases support for state social welfare provision, as expected under hypothesis 2, by 0.18 in 2006 and 0.16 in 2011. As per hypothesis 3, inequality aversion is the largest single influence, increasing support for social welfare provision by 1.33 in 2006 and 2.20 in 2011, reflecting the increase in its value. Consistent with the fourth hypothesis, time matters, demonstrating a net change in the climate of opinion worth more than one point, from -0.48 in 2006 to +0.56 in 2011. In China, the picture is somewhat different (see Figure 3). Again, gender, education and town size have the same or similar impacts in each year. Town size has the largest effect, 1.61 in each year, whilst other social structural effects are minor. Through its impact on getting rich is a zero-sum game, income emerges as a small net positive influence on support for state social welfare provision, worth 0.02 in each year, but the effect is vastly outweighed by the negative effect of satisfaction with the household economy, worth -0. Inequality aversion is a positive influence in both years, as per hypothesis 3, with an effect of 1.50 in 2007 and 1.98 in 2012, reflecting the increase in its value. Consistent with the fourth hypothesis, time matters, again demonstrating a net change in the climate of opinion worth more than one point on the ten-point scale, from -0.29 to 0.83.

Discussion
The economic self-interest arguments receive strong support from the analysis. The fact that the net effect of income has an unexpected sign in China does little to change this conclusion, since it is so small (0.02 on the ten-point scale), and the negative effect of household economic satisfaction is so large (-0.44 to -0.50). The positive effect of town size in China is consistent with Whyte's (2010, chapter 1) legacy-based argument that urban Chinese historically expect more from the state in terms of social welfare provision. The Russian results are likewise consistent with theory, since both income and satisfaction with the household economy are negative influences. The negative effect of town size in Russia is consistent with Ravallion and Lokshin's (2000) results and with the notion that legacy attitudes are stronger in the countryside and smaller towns. Source: Computed from regression model IV in Table 3, panel A, using means of independent variables reported in Table 1.
The fact that the belief that getting rich is a zero-sum game is associated with support for state social welfare provision in both societies is consistent with the proposition that those who think the market is unfair are more likely to support social welfare as a corrective. This finding echoes the classic Meltzer and Richard (1981) model, which assumes that the government is a fair umpire and accountable to tax payers, and therefore perceptions of unfairness should lead to demands to use public money for social welfare.
Attribution of success to non-merit factors has a negative sign in both China and Russia, although it is statistically significant only in the former. This means that Chinese who take the cynical view that success depends on luck and connections tend to believe individuals should provide for themselves and their families. It also means that those who believe that in the long-run hard work brings success also tend to believe the state should do more to provide welfare. We can interpret this as the result of China's ideology of the 'socialist market economy': on the one hand, the state should be benevolent in providing social welfare, and on the other hand, if citizens work hard, they will be rewarded. As Ching Kwan Lee argues on the basis of extended interviews with working and middle class respondents in Beijing 'popular consciousness does not mechanically substitute socialist values of social equality Source: Computed from regression model IV in Table 3, panel B, using means of independent variables reported in Table 1. and need-based allocation with the market ideology of competition and unequal distribution. Rather, there has been a process of layering, mixing and multiplying moral standards and justice claims' (Lee 2009, p. 228). Why is the negative correlation of the support for state welfare and attribution of success to non-merit factors stronger in China than in Russia? It seems possible that Russians, having been through a more radical turn towards pro-market ideology in the 1990s, accompanied by the up-ending of expectations of virtue-based reward, are cross-pressured by the state re-adopting some ideological symbols from the Soviet era whilst failing to resume socialist welfare obligations. The strong positive effects of inequality aversion on support for state social welfare provision reinforce the message that Russians and Chinese agree at least normatively with the idea of social welfare as a way of ameliorating inequality. Although this may seem like an obvious point, it does not feature in two of the leading extant studies on distributive justice attitudes in Russia and China (Kluegel et al. 1999;Whyte 2010), even though the political connection between Communist ideology and state social welfare provision was noted in another widely cited article on Russia (Ravallion & Lokshin 2000).
The effects of time on support for state social welfare provision in the models discussed above show that the upward trend in support in both societies can be explained neither by variations in satisfaction with household economies, nor by market beliefs, nor by fluctuations in inequality aversion, nor by variations in sample characteristics between surveys. The fact that time exerts a net positive influence suggests that there has been a change in the climate of opinion in both societies over the period spanning the world financial crisis.
The WVS data suggest that support for state social welfare provision was on an upward trend in both countries before the financial crisis. Russian support for state social welfare provision may have shifted in response to the regime's more or less explicit embrace of elements of Soviet ideology at the beginning of Putin's second term. Chinese opinion may have shifted in response to Hu Jintao's harmonious society agenda. Both sets of changes could have been accelerated by the political repercussions of the 2008-2009 financial crisis. However, to attribute attitudinal change to the financial crisis itself would probably be a step too far, as it would fail to take into account the ways in which the global economy is mediated and perceived through national institutions, and of the pre-existing ideological dynamics in each society.
This study has a number of limitations. First, the findings are necessarily tentative. For the sake of achieving comparability between the two cases of interest, it has been necessary to restrict the analysis to the WVS dataset. WVS has the principal advantage that working with it allows one to easily test findings in diverse contexts. Nevertheless, WVS usually fails to include the full range of measures found in more specialised and single-country surveys. Second, because this study is based on cross-sectional data, it cannot demonstrate that the relationships modelled are causal in nature. To do that would require panel or experimental data, which are not available on a comparable basis. All that is possible here is to recall the usual caveats that association does not prove causation and that all statistical models are only as tenable as their underlying assumptions.

Conclusion
A substantial portion of the literature on economic attitudes in post-socialist states is devoted to demonstrating that market justice norms are strong, despite the existence of some legacy attitudes, thus discrediting the idea that anti-market norms could lead to political instability or even the overthrow of regimes (Stephenson & Khakhulina 2000;Whyte 2010;Whyte & Im 2014). There is some evidence that the perceived fairness of market institutions influences political trust in post-socialist countries (Kluegel & Mason 2004), and it is possible that extreme unfairness might, if left unaddressed, undermine support for the regime. Certainly, this appeared to be the logic behind Hu Jintao's harmonious society initiative. However, if we look for historical transitions to democracy caused by grievances over economic inequality, Houle's (2009) analysis shows that we are likely to come away empty-handed: inequality appears to reduce the chances of a democratic regime consolidating, but doesn't affect the chances of an authoritarian regime collapsing.
Distributive justice attitudes can matter politically in ways that fall short of undermining the regime. The Putin regime backtracked on the monetisation of benefits in 2005, largely in response to public protest (Wengle & Rasell 2008). Following the 2008-2009 global financial crisis, China increased its investments in education and training for migrant workers and in affordable housing (De Haan 2010, pp. 765), and major new investments were announced in health care with the aim of achieving universal coverage (Yip et al. 2012). These may be examples of authoritarian regimes responding to public opinion about distributive justice issues, although elite-level studies would be required to demonstrate the causal link.
The present study has shown that whilst social structural variables, income and household economic satisfaction certainly matter in the determination of support for state social welfare provision, they are far from being the dominant influences. Ideology matters more, and therefore, it is important to understand the discourse within which discussions of state social welfare provision take place, and to situate that discourse within its temporal and political context.
More granular single-country data could be used to extend the present analysis into the past by looking at particular historical conjunctures in Russia or China. Multi-country datasets such as WVS could usefully extend it across space by looking at distributive justice attitudes in other large and diverse countries where the state plays an important role in the economy, such as Brazil or India.
NEIL MUNRO, School of Social and Political Sciences, University of Glasgow, Bute Gardens, Glasgow G12 8RT, UK. Email: neil.munro@glasgow.ac.uk. corresponding to electoral districts were selected, and at each sampling point interviewers proceeded by a random route to complete around 15 interviews per district. Within each household, respondents were selected to fill a quota by age and gender. The Levada Centre survey of 2011 similarly stratified the country into the same eight macro-regions. Within each macro-region settlements were stratified into five types and a total of 131 primary sampling units corresponding to urban settlements or rural districts were randomly selected with probability proportionate to population. Within each PSU, two electoral districts were selected at random and dwellings identified by following a random route. Within each dwelling respondents were identified by the next birthday method.