Individual perceptions of distributional fairness in China.

Bishop, John A. ; Liu, Haiyong ; Qu, Zichong 等

We should attach great importance to the issue of income distribution and better handling income allocation methods. We should encourage some areas and some of the people to get rich first through honest working and lawful business, and thus to promote other areas. Besides economic development, we should decrease the income gap between regions and individuals to achieve a better distribution situation by reforming the tax system, increasing public spending, increase transfer payments and other measures.

INTRODUCTION

While China's recent transition to a market economy has reached unprecedented levels of economic growth, it has also been accompanied by an unprecedented level of growth in economic inequality. Figure 1 plots per capita income growth and the Gini coefficient of inequality for the period 1975-2009.(1) Riskin et al (2002) While incomes grew 26 fold, the Gini coefficient more than doubled, characterize this recent period of economic transition with a fitting book title, China's Retreat from Equality.

During their tenure as Party General Secretary, Hu and Premier Wen proposed a model of development called the Harmonious Society. Their aim was to reduce inequality and redirect the strategic planning away from the current 'GDP First and Welfare Second' policies. Hu and Wen recognized that certain segments of the Chinese population had been left behind and took a number of high-profile trips to the poorest areas of China with the stated goal of understanding these areas better. Hu and Wen also attempted to move China away from a policy of favoring economic growth at all costs and toward a more balanced view of growth that factors in social inequality and environmental damage.

In order to assess Chinese people's responses to rising inequality during a period of economic transition we incorporate data from two sources, the Chinese Household Income Project (CHIP) and the World Values Survey (WVS). Both of these data sources contain responses to questions related to income distribution. These questions contain subtle differences in their wording and may elicit different responses from the respondents. However, in each case we hypothesize that those who have benefitted most from the economic reforms will be less critical of the current income distribution.

[FIGURE 1 OMITTED]

Previous literature

While there is much anecdotal evidence in the press (c.f., Xinhua News Service, 2005) as well as in statements by government leaders (see above quotation) that the Chinese people believe that the current growth path is 'unfair' there has been very little systematic research on this topic. Wang and Davis (2008) and Whyte (2010) provide evidence from the 2004 China Justice Survey that more than 80 % of urban households considered income inequality 'too large'. Furthermore, in the same data they find that more than 70% of households found that the income distribution became 'more polarized'.

Han and Whyte (2009) use survey data on opinions of distributive injustice in China to assess the likelihood of widespread discontent. Similar to Knight et al. (2009) who study happiness, Han and Whyte find that subjective variables (such as feelings about the degree of corruption) are more useful than objective variables such as age and education. Han and Whyte state that 'distributive injustice attitudes in any society are influenced not simply by current objective status positions, but also and sometimes even more powerfully by subjective factors', such as subjective perceptions of one's social status, past experiences with upward and downward mobility, relative aspirations and using other reference groups to judge one's success or failure. They conclude that 'objective status is a poor guide to perceptions of current inequalities as unjust, subjective measures of relative status and mobility experiences are a much better guide'.

Han and Whyte (2008) compare popular attitudes toward distributive injustice between Beijing and Warsaw. They suggest that 'it is apparent that objective social status predictors generally have fairly weak and inconsistent associations with the four distributive injustice scales" and conclude that 'generally subjective predictors have stronger and more consistent association with these distributive justice scales than do the demographic and objective social status measures'. Grosfeld and Senik (2010) provide evidence of changing attitudes to inequality during the transition to a market economy. They argue that 'the subjective perception of inequality is one of the key elements of the attitudes toward reforms' (p. 2). Similar to the case of China they find 'increasing public sentiment that the process of income distribution is flawed and corrupt' (p. 1).

Finally, Sanfey and Teksoz (2007) use the WVS to study the effect of economic transitions on subjective well-being. They find mixed results in terms of the relationship between happiness and inequality. For non-transition countries, a higher Gini coefficient is associated with higher levels of reported happiness. For transition economies, they find the opposite result that they attribute to a 'lingering dislike of inequality that was characteristic of socialist systems' (p.726).

Econometric analysis with subjective variables

The use of subjective variables in analyzing people's attitude toward fairness can be problematic. For example, there may be cognitive factors that affect the way people answer the survey questions. Furthermore, the ordering of the questions can affect the answers: if people are asked their employment status before being asked about primary factors of inequality, it is more likely for them to answer 'unemployment' in the later question. The survey wording is also important. People often provide different answers based on the positive or negative framing of the question. Finally, the social nature of the survey procedure also appears to play a large role in shaping answers to subjective questioning. Respondents might try to avoid answering questions that might bring negative consequences to themselves such as questions about income, tax, and other social issues when the survey is administrated by a government agent.

These survey design issues result in several possible sources of measurement error. First, the mean of the measurement errors will not necessarily be zero within a survey. For instance, this could be caused by the design of the survey such as positive/negative framing, and as a result the upward and downward biases may not cancel each other in the survey sample. In addition, the measurement error may be correlated with observable individual characteristics. It is also possible that the misreporting/accuracy of the questions may vary in different demographic groups. For example, in politically more liberal regions citizens may be more likely to point to corruption as a source of distribution unfairness. For a more complete discussion of measurement error in survey data see Greene (2012, pp. 784-798).

One way to address the criticisms of subjective survey results is to use multiple data sets. In our empirical analysis we use both the CHIP and WVS data. We also estimate alternative models using different measures of income to assess the robustness of our analysis and comment on the predictive power of various control variables.

THE CHIP AND 'FAIRNESS'

In order to study the Chinese response to rising inequality during the economic transition to a market economy we use two different data sets, the CHIP and the WVS. The 2002 CHIP data contains a set of 'intention' questions. The most important of these are our dependent variables: 'do you think the income distribution is fair or not at all around the nation' and 'do you think the income distribution is fair or not at all in your city'. This part of the questionnaire also includes current income prospects, your current perceived income quartile, the main social problem in your city (eight choices, including corruption), and your self-reported happiness (our measure of current status).

The CHIP data also contain a wide range of demographic and economic variables, including income, marital status, household size, heads age and income, employment status, Communist Party affiliation and so on. The data source definitions used in this study are presented in Table 1. (2) Summary statistics are provided in Table 2.

Descriptive statistics

The descriptive statistics for the 6,374 families included in the sample are reported in Table 2. The families are divided into three different groups based on their subjective perception of fairness: fair, unfair, and extremely unfair. There are 804 families that report 'fair', 3,276 families that report 'unfair', and 2,294 families that report 'extremely unfair' (see Figure 2).

Among these three different response groups, the 'fair' group has the highest average income of [yen] 25,249, followed by the 'unfair' group ([yen] 24,748) and the 'extremely unfair' group ([yen] 21,841). Figure 3 illustrates some differences between the three response groups. Beginning with some objective measures, we find that low-income households are more prevalent in the extremely unfair group (17.89 % of the total) than the fair group (4.73 % of the total), that elder people are more likely to choose fair over extremely unfair (38% versus 35%), and that heads with a high school degree or above have a higher 'fair' response rate (47.9%) than an 'extremely unfair' response rate (39.7%). Other objective variables such as Party membership or low to middle education (not shown) do not vary widely in their responses to the income distribution fairness questions.

Turning to the subjective variables, family heads in the 'fair' group are more likely to be 'happy' (79%) than those in the 'extremely unfair' group (44%). Current income expectations are different among the three groups; 63 % of all families in the 'fair' group report rising incomes whereas only 39% of persons in the extremely unfair group report rising incomes. Households that believe that the income distribution is 'fair' are somewhat less likely to report corruption as the most pressing social issue than those in the extremely unfair group.

[FIGURE 3 OMITTED]

Regression analysis

Our hypothesis is that reform 'winners' will be more accepting of the current income distribution, whereas reform 'losers' are more likely to believe that the income distribution is extremely unfair. We hypothesize that the winners are those with high education and income; losers are the poor and illiterate. Our prediction for Party member is ambiguous: rank and file Party members may have suffered under reform while higher ranked Cadre saw the economic outcome improved. To account for this we include both Party and Cadre on our regression. Respondents with rising income prospects should also be less likely to view the income distribution as extremely unfair. Households with strong negative attitudes toward corruption should be more likely to view the income distribution as extremely unfair. Finally, we include controls for age, gender, and residing region of the country.

We propose three models that differ only by the measure of income used: the first model uses a subjective measure of income, the second model uses an objective measure of income, and the third model uses education indicator variables as a proxy for income. In particular, we estimate the following four ordered logit regressions. (3)

logit(outcome [less than or equal to]k) = [[alpha].sub.k] + [[beta].sub.1] (Perceived Income Quartile) + [[beta].sub.2] (Region) + [epsilon](using subjective income) (1)

logit(outcome [less than or equal to]k) = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (Actual Income Quartile) + [[??].sub.2] (Region) + u(using objective income) (2)

logit(outcome [less than or equal to] k) = (Education) + [[??].sub.2] (Region) + v (using education indicators) (3)

logit(outcome [less than or equal to] k) = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (Perceived Income Quartile) + [[??].sub.2] (Actual Income Quartile) + [[??].sub.3] (Education) + [[??].sub.4] (region) + [xi](Using subjective, objective income, and education indicators) (4)

In each of the above equations k is the value of fairness perception; in our model there are three possible responses. If k= 1, the respondent perceives 'fair', if k=2, the respondent perceives 'unfair'; if k=3, the respondent perceives 'extremely unfair'. In these parsimonious logit specifications, all slope parameters are the same for different outcomes while each level of outcome is allowed to have a different intercept term. Furthermore, we repeat each model described above with a set of controls for Party official (Cadre), Party Member (Chinese Communist Party (CCP)), Young, Male, attitudes about the government (Corruption), attitudes about one's current situation (Happy), and current income prospects (No Better). We described each of these controls (and the omitted groups) in detail in Table 1.

Table 3a shows our estimation results from the ordered logit regressions without additional controls. Recall that the regressions differ only by the measure of income, where Column 1 provides results using subjective income, Column 2 objective income, and Column 3 proxies income with education. The percentage of concordant varies between 55.1% for education and 58.7% for subjective income. Column 4 provides a combined model with both income types and education. The combined model has a slightly higher percentage of concordant (61.1%).

We find that the coefficients of the income quartiles, whether subjectively (Column 1) or objectively (Column 2) measured, are negative, significant, and declining in magnitude, relative to the omitted group (the lowest quartile). This implies that higher income households are less likely to perceive the income distribution as unfair or extremely unfair. For education, we also find that higher educated household heads are progressively less likely to perceive the income distribution as unfair or extremely unfair. When we combine subjective income, objective income, and education, we find that subjective income plays the prominent role. Only one of the objective income indicators is significant (actual top quartile) and none of the education indicators are significant.

Table 3b repeats the model presented in Table 3a with the addition of control variables other than income or education. In each case, we find the percentage of concordant rises with increases ranging from 3.7 percentage points to 8.0 percentage points. Importantly, we note that the models including subjective income (Columns 1 and 4) have the greatest predictive power. Again we find neither actual income quartiles nor the education indicators to be significant. Overall for education we find mixed results; education is statistically insignificant in models including subjective income. (4)

Examining the additional controls individually we find that all are significant in all four models except gender and Party membership (CCP). While Party membership is not significant, being a Party official (Cadre) does decrease the probability of viewing the income distribution as unfair or extremely unfair. Happy persons, and young householders, those with brighter economic prospects, and those who do not view corruption as a major social problem are more likely to have a positive view of the current income distribution.

To further investigate the additional controls in predicting a householder's view of the income distribution, we drop each of the controls separately to determine the marginal contribution to the percentage of concordant. Table 4 presents these marginal contributions. The largest contributors are Happy (2.1 percentage points), Regions (1.1), and Corruption (0.6). All combined the controls (less regions) add 5.7 percentage points to the percentage of concordant.

WORLD VALUES SURVEY

The World Values Survey (WVS) is a worldwide survey that collects information about changing social values and their impact on people's economic, social, and political life. The WVS provides representative national samples for up to 97 countries in six waves. The data for our study is from the fourth (1999-2004) and fifth (2005-2008) WVS waves (sixth wave data is not yet available). For China the actual years surveyed are 2001 and 2007. We create a pooled sample as well as examining the 2007 data separately.

This data set has been prominently studied recently by Alesina and Angeletos (2005), who investigate the role of luck and effort with regard to people's preferences for redistribution. We considera related but somewhat different question, preferences for equality. The WVS question of particular interest to us is: 'Incomes should be more equal ... or do we need larger differences in income as incentives?'

Table 5 provides the responses to this question for our complete pooled sample. The responses can range from 1 to 10 with 1 being the strongest preference for equality and 10 being the strongest preference for inequality. For convenience we define a preference for 'more equality' as responses 1, 2, or 3, and less equality as responses 8, 9, or 10. In this case, we have surprisingly 'fat tails' with 29 % of the respondents preferring greater equality and 43% preferring less equality. Clearly, respondents to the WVS express a weaker equality preference than the CHIP respondents.

How do we interpret the differences between the two sets of responses? It is important to note the differences between the CHIP question and the WVS question. The CHIP question simply asks, 'Do you think the income distribution is unfair', while the WVS question points out the role of incentives. Whyte (2010) shows a similar pattern in the China Justice Survey data. When asked if the 'national income gap is too large' more than 70% (see his Table 11.la) respond affirmatively. However, the same respondents are asked 'does the income gap foster hard work' more than 80% of the respondents are neutral or affirmative in response. Our question is then: Do the same factors that lead to extremely unfair responses in the CHIP data also lead a stronger preference for equality in the WVS data?

Table 6 provides the summary statistics for the independent variables used in the logistic regression. Our logistic model using the WVS data is similar to our combined model using the CHIP data. Unlike the CHIP data, we use a binary dependent variable, the 'more equality' variable equal to one when the responses to the equality question are 1, 2, or 3 and 'more equality' equal to zero otherwise. Income in the WVS data is reported in one of the 10 income classes (not deciles) and the average respondent in our sample is near the top of the fourth income class (mean=4.883). Also included are age, gender, education, and Party membership indicator variables. The subjective independent variables are 'happiness' (scale: 1-4) and 'attitudes toward bribery' (1 = no tolerance and 0 otherwise). Attitudes toward bribery are used to proxy the 'corruption' variable used in the CHIP data.

Table 7 provides logit regression results using the WVS data. We present two sets of WVS results: a pooled model (2001 and 2007) with a year indicator, and a separate set of estimates for 2007. We provide separate estimates for 2007 as the Party variable is not available in 2001. Unlike the earlier CHIP results we find that Party membership is significant; Party members are less likely to desire greater equality. Importantly, bribery in the WVS data, like corruption in the CHIP data, is statistically significant in both models. Both education and income are significant and show the expected signs. The overall predictive power of both the pooled model (64.1% concordant) and the 2007 only model (66.8% concordant) is similar to the CHIP models in Table 3b. Interestingly, the indicator for 2007 is negative; this suggests that preferences for equality are weakening over time.

Table 7, Column 3 provides CHIP results with a model specification similar to that used with the WVS data. In the binary model, a fair or unfair response is given a value of zero, while extremely unfair is one. Higher income households, higher educated heads (the omitted group), people who tolerate corruption, and people who report being happy are all less likely to report that the income distribution is extremely unfair. Again, we find that in 2002 Party membership is not significant. Happiness, significant in the CHIP data, is not statistically significant in either of the WVS models.

CONCLUSION

In the wake of China's enormous success transitioning to a market economy, it is widely believed by policymakers that the country's income distribution has become excessively unfair. We hypothesize that reform 'winners' (educated, high income, higher ranking Party officials) will express less dissatisfaction with the current income distribution and reform 'losers' (less educated, lower income, lower ranking Party members) will express greater dissatisfaction with the current income distribution.

To test this hypothesis we use two data sets, the 2002 CHIP and the WVS, both of which ask questions regarding equality preferences. The questions asked in each survey contain subtle differences in their wording and elicit different responses from the respondents. However, we find that the same factors that lead to unfair or extremely unfair responses in the CHIP data generally lead to a stronger preference for equality in the WVS data. In both the surveys, we find that the perception of unfairness is highly correlated with perceived (subjective income) income, current prospects, and status of Party membership. Higher ranking Cadre members are more likely to view the income distribution as fair. Rank and file Party members in the 2002 CHIP data are no more likely than non-members to find the current income distribution fair; however, in 2007 WVS Survey Party members are more likely to appreciate the role of incentives when expressing preferences for equality. The role of education is mixed--higher educated persons are more likely to recognize the role of incentives, while the role of education in identifying households who view the current income distribution as unfair is sensitive to the model specification, particularly in the presence of subjective income. A strongly negative attitude toward corruption is associated with both an unfair (or extremely unfair) view of the current income distribution in the CHIP data and a strong preference for greater equality in the WVS data. One final observation can be made from the WVS data: a negative year indicator in our pooled model suggests that the average respondent's preference for equality is falling over time. This suggests that the 'lingering dislike of inequality' in a former socialist country is declining over time.

In conclusion, it appears that the average Chinese person recognizes the current income distribution as 'unfair' while he is also aware of the disincentive effects associated with greater equality. Thus, while greater equality may well be desired it is not desired at 'any cost'.

APPENDIX

Ordered logit specifications used in this paper In this paper, we adopt a parsimonious logit specification and all the slope parameters are the same for different outcomes while each level of outcome is allowed to have a different intercept term. Suppose the unconditional probability of each outcome is denoted as [p.sub.k] = Prob(outcome = k), k = 1, 2, 3, then the logit specification is given by the following:

logit ([p.sub.1]) = log [p.sub.1]/[1-[p.sub.1]] = [[alpha].sub.1] + [beta]'X logit ([p.sub.1] + [p.sub.2]) = log [[p.sub.1] + [p.sub.2]] = [[[alpha].sub.2] + [beta]'X [p.sub.1] + [p.sub.2] + [p.sub.3] = 1

This specification assumes proportional odds, as the odds ratio of the outcome Y [less than or equal to] k is independent of the category k. The odds ratio is assumed to be constant for all categories.

REFERENCES

Alesina, A and Angeletos, G-M. 2005: Fairness and redistribution. The American Economic Review 95(4): 960-980.

Chinese Household Income Project (CHIP). 2002: ICPSR21741-vl. Ann Arbor, MI: Inter-university Consortium for Political and Social Research, 2009-08-14. doi:10.3886/ICPSR21741.vl.

Greene, W. 2012: Econometric analysis, 7th Edition. Prentice Hall: Upper Saddle River, NJ.

Grosfeld, 1 and Senik, C. 2010: The emerging aversion to inequality: Evidence from subjective data. Economics of Transition 18 (1): 1-26.

Han, C and Whyte, MK. 2008: Popular attitudes toward distributive justice: Beijing and Warsaw compared. Journal of Chinese Political Science 13 (1): 29- 51.

Han, C and Whyte, MK. 2009: The social contours of distributive injustice feelings in contemporary China. In: Davis, DS and Wang, F (eds). Creating Wealth and Poverty in Post-Socialist China. Stanford University Press: Palo Alto, CA. pp. 193-212.

Hu, J. 2005: People's Daily (Renmin Ribao), 27 June.

Knight, J, Song, L and Gunaatilaka, R. 2009: Subjective well-being and its determinants in rural China. China Economic Review 20(4): 635-649.

Lee, HY. 2000: Xiagang, the Chinese style of laying off workers. Asian Survey 40(6): 914-937.

National Bureau of Statistics of China. 2000: Chinese Urban Household Survey, Beijing, China.

Riskin, C, Zhao, R and Li, S. 2002: China's retreat from equality: Income distribution and economic transition. New York: Studies of the East Asian Institute, Columbia University.

Sanfey, P and Teksoz, U. 2007: Does transition make you happy? Economics of Transition 15(4): 707-731.

The World Bank. 2012: China|Data|Economic Indicators. The World Bank, http://data.worldbank.org/country/china, accessed 23 February.

The Xinhua News Agency. 2005: Widening income gap, the most serious social problem in China. People's Daily Online. 9 July, http://english.peopledaily.com.cn/200507/09/eng20050709_195106.html, accessed 23 Feburary 2012.

Wang, F and Davis, DS. 2008: Creating wealth and poverty in postsocialist china. Stanford University Press: Paio Alto, CA.

Whyte, M. 2010: Fair versus unfair: How do Chinese citizens view current inequalities? In: Oi, J,

Rozelle, S and Zhou, X (eds). Growing Pains: Tensions and Opportunity in China's Transformation. Walter H. Shorenstein Asia-Pacific Research Center: Stanford, CA.

World Values Study Group. 2008: World values survey, 1995-2008. ICPSR: Ann Arbor, MI.

(1) See Lee (2000) for an explanation for the dip in the Gini between 1995 and 2000.

(2) We consider the urban households only as they provide a more reliable measure of household income. We examine only the 'inequality in the nation' responses as the city responses are quite similar.

(3) See Appendix for a description of ordered logit.

(4) In Table 7 we compare CHIP results with WVS data, which lacks a subjective income variable. In this case education is statistically significant.

JOHN A BISHOP [1], HAIYONG LIU [1] & ZICHONG QU [2]

[1] East Carolina University, Brewster A439, 10th St, Greenville, NC 27858, USA. E-mail: liuh@ecu.edu

[2] Georgia State University, Andrew Young School of Policy Studies, Atlanta, GA, 30302, USA.

Table 1: Data source definitions CHIP (2002)

Variables                     Definitions

Fairness                      = 1 if fairness to respondent's
                              perception is 'fair'; = 2 if fairness to
                              respondent's perception is 'unfair';
                              = 3 if fairness to the responder's
                              perception is 'extremely unfair

Subjective perceptions

  Perceived top quartile      = 1 if respondent perceives his/her
                              income as the top 25% of the
                              population; = 0 otherwise

  Perceived second quartile   = 1 if respondent perceives his/her
                              income as the top 25%-50% of
                              the population; = 0 otherwise

  Perceived third quartile    = 1 if respondent perceives his/her
                              income as the 50%-75% of the
                              population; = 0 otherwise

  Perceived fourth quartile   = 1 if respondent perceives his/her
                              income as the bottom 25% of the
                              population; = 0 otherwise

  Corruption                  = 1 if the respondent think corruption
                              is the main social problem in
                              his/her city; = 0 otherwise

  Happy                       = 1 if the respondent feels happy;
                              = 0 otherwise

                              = 1 if the respondent feels unhappy;
                              = 0 otherwise (omitted group)

                              = 1 if the respondent feels extremely
                              unhappy; = 0 otherwise (omitted group)

  No Better                   = 1 if the respondent feels the living
                              standard had increased compared
                              with 1995; = 0 otherwise

  Income great                = 1 if the respondent expected to have a
                              great increase in income in the
                              future; = 0 otherwise (omitted group)

  Income little               = 1 if the respondent expected to have
                              little increase in income in the
                              future; = 0 otherwise (omitted group)

Objective perceptions

  Actual top quartile         = 1 if the respondent's income is among
                              the top 25% of the population;
                              = 0 otherwise

  Actual second quartile      = 1 if the respondent's income is among
                              the second 25% of the population;
                              = 0 otherwise

  Actual third quartile       = 1 if the respondent's income is among
                              the third 25% of the population;
                              = 0 otherwise

  Actual fourth quartile      = 1 if the respondent's income is among
                              the fourth 25% of the population;
                              = 0 otherwise

  CCP                         = 1 if the respondent is member of CCP;
                              = 0 otherwise

  Cadre                       = 1 if the respondent is cadre member of
                              CCP; = 0 otherwise

  Young                       = 1 if the respondent's age is under 25;
                              = 0 otherwise

  Mate                        = 1 if the respondent is male; = 0 if the
                              responder is female

Educational dummy variables

  ES                          = 1 is the respondent has a highest
                              degree of elementary school;
                              = 0 otherwise

  MS                          = 1 is the respondent has a highest
                              degree of middle school; = 0 otherwise

  HS                          = 1 is the respondent has a highest
                              degree of high school; = 0 otherwise

  SC                          = 1 is the respondent has finished some
                              college; = 0 otherwise

  HC                          = 1 is the respondent has a college
                              degree or higher; = 0 otherwise

Regional dummy variables

  LN                          = 1 if the respondent lives in Liaoning
                              Province; = 0 otherwise

  BJ                          = 1 if the respondent lives in Beijing;
                              = 0 otherwise (omitted group)

  HN                          = 1 if the respondent lives in Hunan
                              Province; = 0 otherwise

  is                          = 1 if the respondent lives in Jiangsu
                              Province; = 0 otherwise

  AH                          = 1 if the respondent lives in Anhui
                              Province; = 0 otherwise

  HB                          = 1 if the respondent lives in Hubei
                              Province; = 0 otherwise

  GD                          = 1 if the respondent lives in Guangdong
                              Province; = 0 otherwise

  SX                          = 1 if the respondent lives in Shanxi
                              Province; = 0 otherwise

  GS                          = 1 if the respondent lives in Gansu
                              Province; = 0 otherwise

  YN                          = 1 if the respondent lives in Yunnan
                              Province; = 0 otherwise

Table 2: Summary statistics, in different fairness groups CHIP (2002)

                                  Fairness perceptions

                               1 (Fair)          2 (Unfair)
                                N=804              N=3276

                                    Standard            Standard
Variables                   Mean    deviation   Mean    deviation

Perceived top quartile      0.015     0.121     0.008     0.089
Perceived second quartile   0.478     0.500     0.561     0.496
Perceived third quartile    0.459     0.499     0.353     0.478
Perceived fourth quartile   0.047     0.212     0.075     0.264
Corruption                  0.566     0.496     0.588     0.492
Happy                       0.791     0.407     0.587     0.492
Unhappy                     0.039     0.193     0.075     0.264
Ex unhappy                  0.009     0.093     0.011     0.106
No Better                   0.143     0.350     0.187     0.390
Income great                0.034     0.180     0.021     0.144
Income little               0.607     0.489     0.487     0.500
Young                       0.281     0.450     0.259     0.438
Male                        0.694     0.461     0.674     0.469
Illiterate                  0.053     0.225     0.056     0.230
ES                          0.128     0.334     0.140     0.347
MS                          0.340     0.474     0.346     0.476
HS                          0.167     0.374     0.178     0.383
SC                          0.225     0.418     0.205     0.403
HC                          0.086     0.280     0.076     0.265
Household income            25249     15702     24748     15713
Actual top quartile         0.280     0.449     0.279     0.448
Actual second quartile      0.264     0.441     0.256     0.436
Actual third quartile       0.245     0.430     0.239     0.427
Actual fourth quartile      0.211     0.409     0.226     0.419
Communist Party             0.398     0.490     0.392     0.488
Cadre                       0.366     0.482     0.344     0.475

                           Fairness perceptions

                           3 (Extremely unfair)

                                 N=2294

                                      Standard
Variables                    Mean    deviation

Perceived top quartile      0.006       0.075
Perceived second quartile   0.575       0.495
Perceived third quartile    0.239       0.426
Perceived fourth quartile   0.178       0.383
Corruption                  0.660       0.474
Happy                       0.445       0.497
Unhappy                     0.157       0.364
Ex unhappy                  0.041       0.198
No Better                   0.277       0.448
Income great                0.021       0.145
Income little               0.374       0.484
Young                       0.219       0.414
Male                        0.669       0.471
Illiterate                  0.051       0.220
ES                          0.167       0.373
MS                          0.386       0.487
HS                          0.165       0.371
SC                          0.172       0.377
HC                          0.060       0.237
Household income            21841       14898
Actual top quartile         0.199       0.399
Actual second quartile      0.237       0.425
Actual third quartile       0.267       0.443
Actual fourth quartile      0.297       0.457
Communist Party             0.363       0.481
Cadre                       0.275       0.447

Table 3a: Perceptions of fairness and individual characteristics
CHIP (2002), without controls

                        Subjective     Objective
                          income        income

Perceived               -1.6017           --
top quartile            (0.2832) ***

Perceived second        -0.8601           --
quartile                (0.0823) ***

Perceived third         -1.3662           --
quartile                (0.0877) ***

Actual top quartile        --          -0.6319
                                       (0.0715) ***

Actual second              --          -0.4108
quartile                               (0.0690) ***

Actual third               --          -0.2164
quartile                               (0.0685) ***

MS                         --             --

HS                         --             --

SC                         --             --

HC                         --             --

Percentage of             56.6           58.7
concordant

                       Education        Combined
                       variables       variables

Perceived                  --          -1.4533
top quartile                           (0.2866) ***

Perceived second           --          -0.8178
quartile                               (0.0849) ***

Perceived third            --          -1.2628
quartile                               (0.0942) ***

Actual top quartile        --          -0.2119
                                       (0.0800) ***

Actual second              --          -0.0911
quartile                               (0.0742)

Actual third               --           0.00343
quartile                               (0.0712)

MS                      -0.0299         0.0705
                        (0.0677)       (0.0690)

HS                      -0.1860        -0.0252
                        (0.0798) **      (0.0821)

SC                      -0.3211        -0.0783
                        (0.0768) ***   (0.0804)

HC                      -0.4075        -0.0428
                        (0.1054) ***   (0.1106)

Percentage of             55.1           61.1
concordant

Notes: Robust standard errors in parentheses. * significant at 10%;
** significant at 5 %; ***significant at 1%; all regressions include
regional indicator variables. Positive sign implies greater
dissatisfaction with the income distribution.

Table 3b: Perceptions of fairness and individual characteristics
CHIP (2002), with controls

                              Subjective      Objective
                               income,         income,
                              education       education

Perceived top quartile        -1.0482            --
                              (0.2892) ***
Perceived second quartile     -0.6419            --
                              (0.0866) ***
Perceived third quartile      -0.9467            --
                              (0.0997) ***
Actual top quartile               --          -0.3431
                                              (0.0770) ***
Actual second quartile            --          -0.2086
                                              (0.0723) ***
Actual third quartile             --          -0.1126
                                              (0.0702)
MS                                --             --

HS                                --             --

SC                                --             --

HC                                --             --

Cadre                         -0.1443         -0.1544
                              (0.0561) **     (0.0568) ***
CCP                            0.0559          0.0452
                              (0.0533)        (0.0534)
Corruption                     0.3078          0.2962
                              (0.0507) ***    (0.0505) ***
Happy                         -0.6423         -0.7310
                              (0.0530) ***    (0.0517) ***
No Better                      0.1552          0.3488
                              (0.0664) **     (0.0618) ***
Young                         -0.1645         -0.1825
                              (0.0586) ***    (0.0586) ***
Male                          -0.0326         -0.0270
                              (0.0541)        (0.0541)
Percentage of concordant       64.6            63.7

                             Education
                              without        Income and
                               income        Education

Perceived top quartile           --           -0.9888
                                              (0.2917) ***
Perceived second quartile        --           -0.6298
                                              (0.0887) ***
Perceived third quartile         --           -0.9098
                                              (0.1039) ***
Actual top quartile              --           -0.1728
                                              (0.082) **
Actual second quartile           --           -0.0662
                                              (0.0754)
Actual third quartile            --           -0.00071
                                              (0.0721)
MS                             0.0390          0.0996
                              (0.0692)        (0.0700)
HS                            -0.0466          0.0324
                              (0.0829)        (0.0843)
SC                            -0.0592          0.0499
                              (0.0869)        (0.0890)
HC                            -0.0229          0.1348
                              (0.1156)        (0.1187)
Cadre                         -0.1821         -0.1327
                              (0.0613) ***    (0.0619) **
CCP                            0.0255          0.0667
                              (0.0538)        (0.0542)
Corruption                     0.2894          0.3087
                              (0.0504) ***    (0.0507) ***
Happy                         -0.7651         -0.6336
                              (0.0513) ***    (0.0532) ***
No Better                      0.3867          0.1470
                              (0.0612) ***    (0.0665) **
Young                         -0.1531         -0.1815
                              (0.0594) ***    (0.0600) ***
Male                          -0.0102         -0.0421
                              (0.0539)        (0.0544)
Percentage of concordant       63.4            64.8

Notes: Robust standard errors in parentheses.
* significant at 10%; ** significant at 5%; *** significant at  1%;
all regressions include regional indicator variables. Positive sign
implies greater dissatisfaction.

Table 4: Marginal contribution to percentage of concordant with both
income and education variables CHIP (2002)

               Marginal contribution
Variables    to percentage of concordant

CCP/Cadre                0.1
Corruption               0.6
Happy                    2.1
No Better                0.1
Young                    0.1
Male                     0.0
Regions                  1.1

Table 5: WVS'Equality' response (pooled data)

          Desires more equality

  1       2       3       4       5

11.9%   9.4%    7.4%    3.8%    8.2%

          Desires less equality

  6       7       8       9      10

7.0%    9.4%    15.9%   10.7%   16.3%

Notes: Question: Incomes should be more equal (1) ... Or
do we need larger differences in income as  incentives (10)?.

Data Source: Author calculations

Table 6: WVS summary statistics (pooled data)

Variables                   Mean    Standard error

More equality               0.287       0.000
Scale of incomes            4.833       0.046
Young                       0.452       0.011
Male                        0.505       0.011
Illiterate                  0.119       0.007
ES                          0.146       0.008
MS                          0.147       0.007
Political affiliation (a)   0.091       0.006
Bribery                     1.577       0.032
Happiness                   2.063       0.015
Year 2007                   0.598       0.101

(a) Available for 2007 only.

Table 7: WVS and CHIP logit results
(1=more equality/extremely unfair; 0=otherwise)

                               Pooled       2007
Variables                     estimate    estimate       CHIP

Intercept                     -0.788 *     -1.173       -0.021
Income                        -0.045 *     -0.071 *     -0.011
                              (0.027)      (0.038)      (0.002)
Young                         -0.009        0.135       -0.249 ***
                              (0.134)      (0.137)      (0.066)
Male                          -0.222 **    -0.189       -0.024
                              (0.101)      (0.132)      (0.060)
Illiterate                     0.643 ***    0.666 ***   -0.185
                              (0.216)      (0.201)      (0.128)
ES                             0.797 ***    0.779 ***    0.152 *
                              (0.198)      (0.192)      (0.84)
MS                             0.568 ***    0.540 ***    0.131 **
                              (0.189)      (0.192)      (0.063)
Happy                         -0.236 *     -0.126       -0.667 ***
                              (0.123)      (0.163)      (0.056)
Bribery/Corruption             0.434 **     0.595 ***    0.289***
                              (0.208)      (0.2453)     (0.050)
Political affiliation/Party      --        -0.569 *     -0.013
                                           (0.224)      (0.059)
Year 2007                     -0.350 *        --           --
                              (0.184)
Percentage of concordant       64.4          66.7         63.8
Sample size                    2156          1288         6374

Notes: * significant at 10%; ** significant at 5%; *** significant
at 1%; standard errors are in parentheses.

Figure 2: Percentage of respondents with fairness perceptions

Fair (n=804)                   12.61%
Unfair (n=3276)                51.40%
Extremely Unfair (n=2294)      35.99%

Source: Author calculations from CHIP(2002)

Note: Table made from bar graph.