The Evolution of Educational Inequality between Rural and Urban Residents in China: A Cohort Analysis

This study focuses on how social reform and social changes aﬀect the educational inequality between social groups, speciﬁcally the rural and urban Chinese. Existing sociology theories address the class diﬀerence drives educational inequality, and treat social group diﬀerences as a byproduct of class diﬀerences. By analyzing the evolution and components of the educational gap between rural and urban residents in China, this study found that the main driver of the educational gap between rural and urban residents varies in diﬀerent historical periods. In the low social stratiﬁcation period, if the eﬀect of institutional discrimination is small, the educational gap is small; in the high social stratiﬁcation period with more signiﬁcant institutional discrimination, the inferior rural residents are double jeopardized. These results imply the educational inequality between social groups is driven by both structural changes and discriminatory policies, but if institutional discrimination is the main driver, policies aim to reduce the structural inequality will have little eﬀect on reducing inequality.

Appendix 1: Data Description Appendix 2: Ordered Probit Model of Educational Attainment The educational attainments in the sample is recoded into six categories in each wave, the educational attainments are explained according to the following modeE u = α υ +Ξ υ β υ +s u E r =α ρ +Ξ ρ β ρ +s r Where E represents the educational attainment, α is the constant term, X is the vector of explanatory variables, β are the coefficients of X which generalized by OP process, and srepresents the error term.The educational attainment E i = 1, 2, 6 , E * is the observed outcome given by a set of threshold levels μ κ , the threshold level of follows the well-defined order, where:μ1 The conditional probability that individual i will be located in the educational level can be represented as : Where X is the vector of characteristics of individual i, the δis the the returns of the independent variables, μ is the scale which represents the observing schooling level, and Φ means the cumulative density function follows the standard normal distribution.
Appendix 3: Decomposition Analysis: at One Period and Overtime Four-Fold Division Decomposition at one periodThe overall differences between urban and rural residents in a time point can be written as:tt u,r = E u -E -r = (α υ -α ρ ) + (X ujβ υ -Ξ r jβ ρ ) I denote tt u,r as the overall difference in the educational attainment between urban and rural hukou holders.The relationship of the educational attainment and other characteristics is presented in the linear regression model in the paper, decompose the gap of educational attainment between urban and rural residence could be written as follows: 5 which refers the outcome to be "explained" by the difference of the characteris-tics(Jann 2008).The U is unexplained part U is the "unexplained" part, many scholars interpret it as discrimination, it also generate the effects of difference in unobserved variables (Oaxaca 1972 ;Blinder 1976;Jones and Kelley 1984).Decomposition: Changes OvertimeWhen investigate the details of the evolution of the gap, we apply the Welling-ton's decomposition(1993) and add time t and cohort c to the previous model.Here t refers to the year of major change happening, and c refers to the cohort of birth who are affected by the changes.For example, if I want to investigate the effects of 1978 reforms on the gap of educational attainment, I need to compare the difference of the links between residence types and other characteristics for cohort born between 1955-1964(cohort 1) and cohort born between 1965-1974 (cohort 2).
The equation (1) on the right hand side measures the changes of the gap in the mean educational attainments of urban and rural forc1 andc2.The equation (2) of the right hand side is the decomposition of the change.
The first term in the square brackets is the unexplained changes for rural and urban residents.The second term measures the portion of changes links to the changes of char-acteristics of urban and rural residents (e.g., the average measured of father's education of rural increases relative to that of urban).The last term in square bracket measures the portion of changes due to changes of coefficients (or it could be interpreted as the changes due to difference in the return of the independent variable).Using same mode decompose the urban-rural gap of educational at-tainment between cohort 1975-1984 (c3) and cohort 1964-1975 (c2) to investigate the influence of educational expansion.

Table 1 :
Description of Data

Table 2 :
Graduate from high school: Logistical Regression of Difference between Rural and Urban for Urban (All cohorts) Poll Urban Rural Decomposition Difference% * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: the positive numbers of decomposition results indicates advantage for urban hukou holders; the negative numbers indicates advantage for ruralhukouholders 7

Table 3 :
Graduate from high school: Logistical Regression of Difference between Rural and Urban for Urban < 0.05, ** p < 0.01, *** p < 0.001 Notes: the positive numbers of decomposition results indicates advantage for urban hukou holders; the negative numbers indicates advantage for rural hukou holders 8 Table4: Graduate from high school: Logistical Regression of Difference between Rural and Urban for Urban( Cohort: 1965( Cohort:  -1974) )Poll Urban Rural Decomposition Difference% * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: the positive numbers of decomposition results indicates advantage for urban hukou holders; the negative numbers indicates advantage for rural hukou holders 9

Table 5 :
Graduate from high school: Logistical Regression of Difference between Rural and Urban for Urban