The long-term impact of the Communist Revolution on social stratification in contemporary China.

The Chinese Communist Revolution that culminated in the 1949 founding of the People's Republic of China fundamentally transformed class relations in China. With data from a nationally representative, longitudinal survey between 2010 and 2016, this study documents the long-term impact of the Communist Revolution on the social stratification order in today's China, more than 6 decades after the revolution. True to its stated ideological missions, the revolution resulted in promoting the social status of children of the peasant, worker, and revolutionary cadre classes and disadvantaging those who were from privileged classes at the time of the revolution. Although there was a tendency toward "reversion" mitigating the revolution's effects in the third generation toward the grandparents' generation in social status, the overall impact of reversion was small. The revolution effects were most pronounced for the birth cohorts immediately following the revolution, attenuating for recently born cohorts.

The Chinese Communist Revolution was a class-based revolution with peasants as its main supporters (1). To maintain class-based interests in the newly founded People’s Republic of China (PRC) in 1949, the Chinese government soon registered every citizen as belonging to 1 of 3 broad classes according to his or her presumed role in the revolution: “good class” (“red class” revolutionary cadres, revolutionary soldiers, and revolutionary martyrs as well as industrial workers and poor and lower-middle peasants), middle class (middle- and upper-middle peasants, urban routine staff, small businessmen, intellectuals, and professionals), and “bad class” (also called the “black class,” including landlords, rich peasants, capitalists, capitalist roaders, counterrevolutionaries, rightists, and “bad elements,” such as criminals) (ref. 2, ). This classification scheme was largely property based, as it would classify essentially anyone with property at the time of the revolution as bad class. However, the presence of revolutionary classes also allowed persons from families with property to be classified as good class if they had contributed to the revolution in significant ways. Using this class scheme, the PRC government devised and implemented a series of class-based preferential social policies (1, 2).

One of the most visible and most consequential policies was the distribution of educational resources in favor of good-class children at the expense of bad-class children (2–5). Redistribution strategies included rapidly expanding education at all levels, opening special schools for government cadres only, and developing college admission policies that aimed to increase the enrollment of good-class students and restrict the number of bad-class students (). More than 60 y have passed since these policies were first implemented. Most grandchildren of the revolutionaries have now completed their educations. A natural question to ask is whether these preferential policies actually succeeded or failed to transform the social stratification order by favoring descendants of those disadvantaged classes who supported the revolution.

There are good reasons to suspect that these policies may not have worked. Bad-class citizens had relatively high levels of education and enjoyed social and economic privileges prior to the revolution. It is well known that family socioeconomic status (SES) has a significant influence on children’s educational outcomes in almost all societies (6, 7). Prior research has shown that family SES has persistent large effects over time, and this is true even in Eastern European countries that experienced a transition to socialism similar to that in China after World War II (7–9). Through both economic resources and noneconomic means, the latter of which include parenting styles, socialization, development of noncognitive skills, and social networks, parents transmit their social advantages or disadvantages to their children (10–12). While the revolution took away economic resources from bad-class citizens, they still possessed superior cultural and educational resources that could be used to advance their children’s education (8, 9). Research has also shown that family SES effects persist even when education expands, as high-SES parents can find ways to use resources in competition for higher levels of education as lower-level education becomes more widely available (13). Further, even if the influence of parents’ SES on the immediately following generation is blocked or reversed by a revolution, recent research suggests that the third generation may still be affected by grandparents’ SES (14–16).

However, the Chinese Communist Revolution was not an ordinary social event. It was a radical class struggle intended to fundamentally transform social relations between 2 opposing classes: good versus bad (1, 2). Even after properties of the formerly privileged classes were taken away by the revolution, children from these classes inherited permanent bad-class designations and suffered the same effects, being considered “bad” people to be “cleaned up” in postrevolutionary China. Thus, bad-class children were not only deprived of equal rights to education (3); they were also subject to social isolation, systematic discrimination, and sometimes physical abuse (2, 17). In this study, we examine the long-term impact of these class-based preferential policies on the social stratification order in postrevolutionary China.

Intergenerational and Multigenerational Social Mobility

Intergenerational social mobility refers to the difference in social status—educational attainment in our study—between parents and children. In social mobility research, absolute mobility can be distinguished from relative mobility (18). Absolute mobility is driven by overall societal changes, also called “structural changes.” For example, when education expands, children’s education tends to surpass parents’ education. However, even when expansion occurs, we should ask whether social origin allows certain children to benefit more from the expansion than others, raising the question of relative social mobility, or social fluidity. Past research has found the amount of relative social mobility to be persistent in most industrialized nations (7, 18). In our study, we rescaled educational attainment into percentile scores within a gender-specific birth cohort. Thus, we purged the overall changes in educational distribution in measurement. As a result, all of the analyses reported in this study are concerned with relative mobility except when we discuss structural changes.

Most research on mobility has been concerned only with parent–child relationships. This practice has been largely necessitated by data limitation, as researchers seldom have access to data pertaining to social status across multiple generations in the same family. This approach is justifiable if social mobility across multiple generations is Markovian: grandparents affect grandchildren entirely through affecting the middle generation. Recent research has shown, however, the inadequacy of the Markovian assumption, as it has found direct effects of grandparents’ social status on grandchildren’s social status unmediated by the middle generation, through potential causal mechanisms such as material resources, cultural and social capital, and direct grandparenting (14–16). The potential presence of non-Markovian grandparent effects is important to our study because it raises the question of whether the impact of the Chinese Communist Revolution on social stratification was short term or long term.

Let us consider the following thought experiment with 3 generations: grandparent (G1), parent (G2), and grandchild (G3). At the time of the revolution, each member of G1 was registered by the government in a specific family class. This class designation was passed along the paternal line onto G2 and G3. Let us suppose that the revolution succeeded in interrupting the prevailing stratification regime—the continuation of social status from G1 to G2—so that it suppressed the educational outcomes of G2 citizens from the bad class (previously privileged classes) while at the same time promoting educational outcomes of good-class G2 members, mostly workers and peasants along with revolutionaries. Thus, G2 bad-class children received less education than they would have in the absence of the revolution. Likewise, good-class children received more education than they would have received otherwise. Both G2 and G3 grew up in postrevolutionary China, and from past research we would expect G2’s education to affect G3’s education in a stable regime. In this case, we would expect non-Markovian effects of G1 on G3.

Among bad-class G3 children, their parents (G2) received less education than they would have otherwise. However, their grandparents (G1) had high social status before the revolution. Even though the revolution deprived G1 of material assets, bad-class families still had cultural and intellectual resources and could use them for their G3’s education. Conversely, the G2 parents of good-class G3 children received more education than they would have before the revolution. The children’s grandparents (G1) had low social status before the revolution and could not contribute much in terms of intellectual or cultural resources toward G3’s education. That is, good-class G2 parents might not be able to parlay their educational advantage to their children. Note that non-Markovian grandparent effects, if found to be present in our study, could offset the initial impact of the revolution on bad-class G2. G3 would bounce back in educational attainment, and good-class G3 would lose out because the former group had a better-educated G1 than the latter. Alternatively, this can be seen as a case of Galtonian regression, reversing toward the family line average.

Since Mare’s 2011 paper (14), a debate has emerged in social stratification research concerning the presence or absence of non-Markovian multigenerational effects. If social mobility processes are truly Markovian, long-term multigenerational mobility trends can be conceptualized as the product of parent–child relationships, with more distant relationships being mediated by immediate ones. If social mobility processes are non-Markovian, it is necessary to consider how distant generations may directly affect later generations unmediated by immediate ones.

We do not believe that there is a general resolution to this debate. Social processes are often complex and vary with social context. When a radical social event, such as the Chinese Communist Revolution, occurs, it interrupts the existing social order. In particular, it renders social inheritance between 2 immediate generations temporarily unsustainable because the normal channels for social reproduction are blocked or disrupted. As a result, children from previously high status families may fail to maintain the high status enjoyed by their parents, while others from previously underprivileged families are given a rare opportunity to get ahead. We conjecture that those occasional historical moments of large-scale social interruption are likely to bring about non-Markovian multigenerational effects, as social advantages of grandparents or even distant ancestors may have long-term effects on later generations. Hence, in this research we aim to explore whether the Chinese Communist Revolution changed the stratification order at the founding of the People’s Republic of China in 1949 so radically that the third generation after the revolution has experienced non-Markovian grandparent reversing effects. Further, if we find non-Markovian grandparent reversing effects, we ask to what extent they have weakened the long-term impact of the revolution on class differences in educational outcomes for G3.

Data and Methods

We used data from a nationally representative, longitudinal household survey of over 14,000 families, the China Family Panel Studies (CFPS), to study postrevolutionary trends in intergenerational and multigenerational educational mobility in China (). In the 2010 baseline and 2012, 2014, and 2016 follow-up surveys, we collected appropriate data for our research: retrospective and prospective data on educational outcomes for 3 generations ().

We constructed 3-generation linked data in the CFPS, with adult respondents in 2010 who were born between 1940 and 1969 serving as the middle generation (G2), their parents serving as the first generation (G1), and their children serving as the third generation (G3). We measured the social status of each generation by educational attainment. China has experienced rapid improvement in educational attainment since 1949 (19), and measurement of education in absolute attained levels would not be comparable by generation, birth cohort, and gender. While young members of all class backgrounds have experienced large improvements in absolute education (), our research interest centers on the variation in the intergenerational improvement in relative education by class background, i.e., whether descendants of the bad classes have benefitted less from China’s education expansion in comparison to those of the good classes. Thus, we compared each person’s education to that of his or her peers by birth year and gender using Chinese census data and obtained education percentile scores, ranging between 0 (lowest) and 100 (highest) (ref. 15, ). When information for both parents (G2) was available, we took the average of their percentile scores. Similarly, we took the average for G1 when information from multiple grandparents was available.

To study how the revolution affected persons born in different periods, we further broke down G2 into 3 birth cohorts: 1940 to 1947, 1948 to 1962, and 1963 to 1969 (). The first cohort was born before the revolution and was affected directly by it in childhood. Born a year before the revolution or immediately after it, the second cohort suffered the effects of the Cultural Revolution (1966 to 1976), which closed down colleges and universities and sent about 17 million urban youths to the countryside for rustication (1, 4). The third cohort was born after the Great Famine of 1959 to 1962 and had the opportunity to attend college by taking college entrance examinations that were reintroduced in 1977 (1–5, 19).

Family class origin was obtained from G2 respondents in the CFPS, using a classification system originally designed for redistributing land and other properties during the Land Reform in the 1950s. Although we collected detailed class categories, we grouped them into 6 large categories with sufficient cases in our sample: red class members (revolutionary cadres, martyrs, and soldiers), workers, poor peasants, middle peasants, upper-middle-class members, and bad-class members (landlords, rich peasants, counterrevolutionaries, rightists, bad elements, capitalists, and capitalist roaders) (). Family class affiliation played a central role in the lives of all ordinary Chinese people for nearly 3 decades between the 1950s and the 1970s and dominated Chinese society during the Cultural Revolution, when class struggle became the center of all social life (17).

Historical records clearly show that the Chinese government implemented a series of class-based education policies to promote the interests of the good classes. For example, it established a large number of adult education programs dedicated to cadres from the worker and peasant classes (). In regular schools, the government steadily increased the proportions of enrolled students from peasant and worker backgrounds at the secondary, tertiary, and postgraduate levels (). In 1951, students from peasant and worker backgrounds constituted 53 and 19%, respectively, of those enrolled at the secondary and tertiary levels. They moved up to 81 and 65%, respectively, by 1965. Recall that these increases in proportions were also accompanied by a large expansion of education at the secondary level immediately after the revolution (19). Thus, the education of children from peasant and worker backgrounds improved dramatically, both in the expansion of secondary education and in their share of the enrollment. At the postgraduate level, the proportion of students from peasant and worker backgrounds increased from 15% in 1953 to 38% in 1965. For college education, the government implemented, between 1958 and 1966, a series of political screening policies to explicitly discriminate against bad-class children (). As a result of these class-based college admission policies, the proportion of students from the worker and peasant classes dramatically increased, from 55% in 1958 to 70% in 1964, while the proportion of those from the “exploiting” classes declined from 17% in 1958 to a mere 5% in 1964 (). As a result, the proportion of college students who were considered politically reliable enough to study classified majors increased steadily, from 30% in 1958 to 78% in 1964.

We conducted 3 sets of statistical analyses of the multigenerational linked data in the CFPS. First, we examined to what extent G2 from different classes experienced downward or upward educational mobility from G1. If the revolution successfully weakened the dependence of G2’s educational attainment on their social origins (the influence of G1), we would observe upward G1–G2 educational mobility in worker and peasant families but downward G1–G2 educational mobility among bad-class and upper-middle-class families. A difference in education percentile score between G2 and G1 within a family is a straightforward measure of the amount or distance of educational deviation of G2 from G1. As education percentile scores were already standardized by birth year cohort and gender, they can be used as relative educational attainment for cross-generational comparisons. G2’s educational deviation from G1 can be either positive (for upward mobility) or negative (for downward mobility). The absolute deviation represents the amount of mobility. To describe both the direction and the amount of mobility by class origin on average, we constructed 2 indexes. For the first, we considered any mobility as a G2 − G1 difference by at least 5%. For the second, we considered long-range mobility as a G2 − G1 difference by 50% or greater. Obviously, long-range mobility represents a huge deviation in social status between G1 and G2. We intended it to capture the impact of the interruption of stratification order caused by the revolution. In both cases, we took the difference between the percentage of upward mobility and the percentage of downward mobility and called it an “index of advantage.”

In the second step, we examined G3’s education by regressing G3’s education scores on G2’s education scores and G2’s educational deviation from G1, controlling for G3’s gender and sibling size (). The intergenerational effect of G2’s education on G3’s education was measured by the regression coefficient of G2’s education scores. G2’s educational deviation from G1 represents G1’s non-Markovian grandparent effect on G3, which we expected to be negative as G3 tends to “revert” to the social position of G1 net of the mediation through G2.

We know that the grandparent effect, if present, would improve the education of descendants of the bad classes and suppress the education of descendants of the good classes. If we find this effect in our study, we will still want to assess its importance for the long-term impact of the revolution. Thus, we conducted a counterfactual exercise in which we removed the non-Markovian grandparent effect from our model and then simulated the education outcomes by class origin.

Finally, since G2 was born in different historical periods when the revolution broke out, we were interested in testing whether those who were born later were less affected by the revolution than those who were born earlier. Of course, we were not able to distinguish confounding effects of other historical events, such as the Great Famine (1959 to 1962) and the Cultural Revolution (1966 to 1976). The best that we could do was to examine cohort variation and interpret the variation with caution.

Educational Attainment by Family Class and Generation

In Fig. 1, we present the trend in educational attainment from G1 to G3 by family class. Because mobility rates are confounded by parental fertility (20, 21), we adopted the standard approach of focusing on one generation at a time, treating parent or child status as individuals’ attributes (20). Fig. 1 shows a clear upward pattern of educational mobility among descendants of poor peasants, workers, and red class members. Although G1 of poor peasants started at very low educational attainment, at the 42nd percentile, it increased steadily from G1 to G3 by about 5%. Education of the worker class and the red class enjoyed a greater increase, by 12% from G1 to G3. As a result, education of G3 from the poor-peasant class fell just short of the median (at 46%), education of G3 from the worker class surpassed the median by 12% (at 62%), and red class G3’s education achieved the 78th percentile.

Fig. 1.

(A–F) The mean education percentile score of G1, G2, and G3 by family class origin. The 95% confidence intervals are provided. The education percentile score of G1 is an average of education percentile scores from both paternal and maternal parents of G2. Men and women were pooled. Statistics were appropriately weighted by number of siblings and sampling weights.

Relative mobility is a zero-sum game. For the good classes to gain status, other classes had to lose out. Indeed, our results show that the bad class and the upper-middle class lost out over generations. The bad class enjoyed an education advantage in G1 (at 52%), but they fell below the median in G2 (at 46%) and regained slightly to almost reach parity with the median in G3 (at 50%). The upper-middle class had a huge initial advantage with their percentile at 67% in G1, but their advantage gradually lost ground, with their education percentile declining significantly to 58% in G2 and then to a level slightly above the median at 54% in G3.

Educational Mobility from G1 to G2

How did the trends by generation depicted in Fig. 1 come about intergenerationally? We now take a close look at how the revolution affected G1-to-G2 social mobility by class. We analyzed detailed mobility tables by class () and summarize the main results in Fig. 2, which displays the index of advantage by family class for overall mobility and long-range mobility.

Fig. 2.

The index of advantage for G2’s educational mobility from G1 by mobility distance and family class origin. Mobility is defined as a 5% or larger educational difference between G2 and G1. Long-range mobility is defined as a 50% or larger difference between G2 and G1. The index of advantage is the net mobility, the difference between upward mobility and downward mobility.

Overall, a slightly larger portion of G2 from the poor-peasant class experienced relative downward mobility than upward mobility (at a net disadvantage of 8%). However, this overall pattern was driven by the downward mobility of women, not that of men. G2 men from the peasant class actually experienced 4% net upward mobility (). Since sons were traditionally favored over daughters in rural China, men from the peasant class were likely to seize new educational opportunities created by the revolution, while rural women’s education was impeded all of the way up until the Cultural Revolution (22). The data also reveal that a noticeable proportion (around 3% overall) of G2 from the poor-peasant class experienced long-range upward mobility for both men and women (), likely attributable to the government’s class-based preferential policies. Those in G2 of the worker class and red class were the main beneficiaries of the revolution. The worker class experienced more upward than downward mobility for both overall mobility (by 20%) and long-range mobility (by 8%). The red class benefitted even more, with 30% overall upward net mobility and 9% long-range upward net mobility.

Downward mobility was prevalent in the bad class. For G2 from the bad class, the overall net downward mobility was 17%, and long-range net downward mobility was 4%. For the upper-middle-class G2, the overall net downward mobility was 10%, but long-range net downward mobility was very small, at 1%. Our interpretation is that the upper-middle class’s decline in advantage over a generation, shown in Fig. 1, did not result from radical intergenerational disruption caused by the revolution, as was the case for the bad class. Rather, it was due partly to the gradual erosion of their elite advantage through the steady expansion of education to serve previously underprivileged segments of Chinese society.

Three-Generational Mobility from G1 to G3

We now extend the analysis to educational outcomes of G3. In Table 1, we present the regression results, with G3’s education scores as the outcome variable and G2’s education score and G2’s education deviance from G1 as key independent variables. As we expected, G2’s education is positively associated with G3’s education. A more important finding is the negative effect of G2’s deviance on G3’s education, shown in model 1 of Table 1. We find that the greater G2’s education deviation from G1 is, the more G3’s education bounced back to G1’s level, a non-Markovian grandparent effect on grandchild’s achievement. To check the robustness of the finding, model 2 further breaks down G2’s educational deviance into positive deviance and negative deviance, with zero being the reference category. Model 2 results support our interpretation: when G2’s education deviated negatively from G1’s, G3’s education went up; when G2’s education positively deviated from G1’s, G3’s education declined.

Table 1.

Ordinary least square regression of G3’s educational attainment on G2’s educational deviation from G1, n = 9,023

Variables	Model 1	Model 2
G2 education percentile	0.610***	0.609***
	(0.022)	(0.022)
G2 educational deviance from G1	−0.160***
	(0.022)
Positive deviance of G2 from G1		−0.152***
		(0.032)
Negative deviance of G2 from G1		−0.167***
		(0.030)
G3 sibling size	−4.160***	−4.159***
	(0.273)	(0.273)
G3 male	−1.006	−1.004
	(0.581)	(0.581)
Constant	38.761***	38.639***
	(1.447)	(1.486)
R2	0.206	0.206
F test: model 1 nested in model 2		0.130

SEs are in parentheses. The dependent variable of models 1 and 2 is G3’s education percentile score. The reference category of G3’s sex is females. Model 1 (restricted model) is nested in model 2 (unrestricted model). The F test for the nested models shows that model 1 is preferred. ***P < 0.001.

Regression results in Table 1 suggest that grandparents’ (G1) non-Markovian effects may have worked to weaken the long-term impact of the revolution on G3. Because the government’s class-based policies helped the red and worker classes but suppressed the bad and upper-middle classes for G2’s education, the direct, non-Markovian effects of G1 on G3 counterbalanced the policy effects on G2. To what extent did the grandparent effects mute the long-term impact of the revolution on G3?

To answer this question, we conducted an exercise in which we compared class differences in educational outcomes of G3 under 2 counterfactual scenarios () using our regression estimates. The first scenario assumed no educational deviation of G2 from G1; that is, the revolution did not happen or never interrupted educational reproduction between G1 and G2, so we assigned to G2 the class-specific median. Using the observed G3 education percentile as the final outcome, we found that in this case, G3 of peasant, worker, and red-class origins would be much less educated than was actually observed. In contrast, G3 of bad-class and upper-middle-class origins would achieve higher levels of education than observed. The differences were all substantial, between 6 and 14 percentage points (Fig. 3).

Fig. 3.

The observed and counterfactual educational attainment of G3. The observed G3 education percentile scores, marked by open circles, are based on the G3 sample without missing values in covariates. One counterfactual scenario, marked by black squares, is the predicted G3 education percentile score if G2’s education followed G1’s father’s class median education percentile score (“if immobility from G1 to G2”). The other counterfactual scenario, marked by red squares, is the predicted G3 education percentile score if the effect of regression toward G1 was removed (“if no effect of G1 on G3”).

In the second scenario, we assumed there was no non-Markovian grandparent effect, that is, no direct effect of G1 on G3, while holding other things unchanged. For G3 of red class and worker origins, removal of grandparent effects increased their counterfactual education scores, as their G1 low education held them back. The magnitude of the differences, however, was small, about 2 points. Further, the grandparent effect was also small for the bad class, as its removal reduced their G3’s education by about 2 points (Fig. 3). The removal of non-Markovian grandparent effects had no effect on G3’s education in families of poor peasants, middle peasants, or the upper-middle class. In summary, the non-Markovian effects of G1 on G3 did mitigate somewhat the effects of the revolution, in the direction of disadvantaging G3 of the red and worker classes and helping G3 of the bad class, but the size of the impact of this reversion was small, insufficient to compensate for the very large, class-based effects of the revolution on G3’s education.

The Changing Impact of the Revolution

G2 in our study grew up in different periods after the revolution, and thus, some were more affected by it than others. G2 born between 1940 and 1962 were exposed to many postrevolution political events, such as the Land Reform, the Anti-Rightist Movement, and the Great Leap Forward. By comparison, the last cohort, born in 1963 to 1969, did not suffer from these turbulent events or the Great Famine. They were too young to be affected by the send-down movement (1, 4). Is there a variation across the cohorts in the impact of the revolution, operationalized as the class-based difference in G1–G2 social mobility?

In Fig. 4, we present the results of the cohort variation in G1–G2 education difference by family class. The y axis represents net mobility, the average difference in education percentile scores between G1 and G2. For the first cohort, born between 1940 and 1947, we observe large differences in net mobility by class, with the red, worker, and middle-peasant classes at the top, the bad class at the bottom, and the other 2 classes in the middle, close to zero. The class differences in net mobility persisted into the second cohort, born between 1948 and 1962. However, class differences in net mobility are less pronounced for the last cohort than for the previous 2 cohorts, except for the upper-middle class. This supports the interpretation that the impact of the revolution on G1–G2 social mobility waned over time, although the cohort trend could be confounded by important political events, such as the Cultural Revolution (1966 to 1976).

Fig. 4.

(A–F) The average G2 − G1 difference in education percentile by G2’s birth cohort and family class origin.

Conclusion

In this study, we analyzed recently collected data on 3-generational social mobility in contemporary China to answer the question of whether the Chinese Communist Revolution fundamentally altered the prerevolutionary social stratification order over the long term. Our answer is a qualified yes. We showed that the revolution disrupted the cross-generational reproduction of the prevailing social stratification at the time of the revolution in promoting the social status of children of the red, poor-peasant, and worker classes and disadvantaging those from privileged classes. We also found a tendency for reversion mitigating the revolution’s effects in the third generation toward the grandparents’ generation in social status, but it was insufficient to compensate for the class-based preferential policies implemented by the government. In other words, grandchildren of the bad class and the upper-middle class were helped by better-educated grandparents than grandchildren of the worker and peasant classes. However, the impact of the revolution on the educational attainment of descendants by class classification immediately following the revolution remains strong even after 6 decades have passed. There is also variation in the impact of the revolution by cohort, as it was most pronounced for the birth cohorts immediately following the revolution, attenuating for more recently born cohorts.