Education and earnings in the Peoples Republic of China

THE WORLD BANK ECT5G Discussion Paper ED UCATION AND TRAINING SERIES Report No. EDT56 Educa Hon and Earnings in the People's Republic of China Dean T. Jamison Jacques van der Gaag January 1987 Education and Training Department Operations Policy Staff The views presented here are those of the author(s), and they should not be interpreted as reflecting those of the Worad Bank.  Discussion Paper Education and Training Series Report No. EDT56 EDUCATION AND EARNINGS IN'THE PEOPLE'S REPUBLIC OF CHINA Dean T. Jamison Jacques van der Gaag Education Policy Division Education and Training Department January 1987 The World Bank does not accept responsibility for the views expressed herein, which are those of the author(s) and should not be attributed to the World Bank or its affiliated organizations. The findings, interpretations, and conclusions are the results of research or analysis supported by the Bank; they do not necessarily represent official policy of the Bank. Copyright t 1987 The International Bank for Reconstruction and Development/ The World Bank  ABSTRACT Data from a household survey conducted in a relatively poor county in northwestern China allow assessment of the impact of education on the employment status and earnings of urban dwellers and on the value of output of small farms. The Mincerian rate-of-return to schooling was 4.5% for urban males and 5.5% for urban females; these are among the lowest level yet reported. The effect of experience on earnings was also unusually low for males and negligible for females. Findings from rural areas, where market forces play a more central role, were more consistent with findings from other countries. The education levels of adults in farm households had a strong impact on total farm earnings, mostly through its impact on "sideline"' production rather than grain output. Data on the educational attainment of children allowed an assessment of the determinants of schooling received; as in other developing countries, children of poorly educated parents do less well than those of better educated ones and rural dwellers do less well than those from urban areas. Surprisingly, female children seem to be receiving slightly more education than males.  Table of Contents Page No. INTRODUCTION ...................................... 1 I. DATA ...................................... 1 - 3 II. URBAN EMPLOYMENT AND EARNINGS ................... 3 - 4 III. THE INCOME OF FARM HOUSEHOLDS ................... 4 - 6 IV. DETERMINANTS OF SCHOOLING .. 6 NOTES ................................................. 7 REFERENCES ............................................ 8 TABLES Table 1: Variable Definitions, Means and Standard Deviations - Urban Sample .... ..... 9 Table 2: Variable Definitions, Means and Standard Deviations - Rural Sample .... ..... 10 Table 3: Probit Regression for Employment Status: Females in Urban Areas ...................... 11 Table 4: Determinants of Urban Earnings .............. 12 Table 5: Determinants of Agricultural Output Value... 13 Table 6: Determinants of Sideline Output Value ....... 14 Table 7: Determinants of Total Output Value .......... 15 Table 8: Years of School Attended, 10-17 Years. Old 16  Introduction In a recent issue of The Journal of Human Resources, Psacharopoulos (1985) reviewed evidence on the returns to education in 61 political entities that had a combined total population in 1983 of something over 1.6 billion persons. Our purpose in this note is to add an analysis of data from the People's Republic of China to this compilation -- bringing, thereby, the represented population total to over 2.6 billion. Our data are from only one of China's more than 2,000 counties, so any extrapolation of findings to the rest of the country must be done cautiously. Yet the mechanisms for determination of urban wages and the nature of the rural incentive structure have much in common across China while differing substantially from much of the rest of the world. It is thus of interest to assess the similarity of the effect of education on earnings in China with its effects elsewhere. This note begins with a brief description of our data then turns, in Section II, to assessments of the determinants of urban employment and wages. Section III deals with farm households, and Section IV deals briefly with the determinants of school attainment. I. Data In March, 1985, a small household survey was conducted in Hui County, Gansu Province, to provide detailed information on the use of health services in relation to urban/rural residency, insurance coverage, occupation, income, education, and several other variables of interest. I/ Households were the basic sampling units with all individuals in each sampled household to be interviewed. The sample was designed to provide information from 300 rural  -2- and 200 urban households in the county. Because of logistical, financial, and time constraints, data were obtained on a stratified, rather than on a random sample. The households of the county were divided into three groups. The first group constitutes the urban sample; the second and third groups form the rural sample - one from valley areas, the other from hill and forest areas. Due to refusals, the fact that some households were not present at the time of the interview, and missing data in some of the filled-in questionnaires, the available data are from 481 households of which 290 are in rural areas. Urban households have on average 4.42 members per household; rural household are larger, with 5.97 members on average. The sample has information on 2154 individuals, 49.7% of whom are males. The average age of the head of a household is 46.2 years, with not much difference between urban and rural households. Table 1 describes the data we use for analysis of urban earnings, and Table 2 describes the farm sample. Adults in both samples are defined as those over 17 years of age, and the urban analysis includes all adults not indicating themselves to be students or retired. Virtually all males in this group are employed, but a significant fraction of females (28%) indicated that their work was principally in the home. For this reason the data for females in Table l are broken down by employment status. Definitions of the variables in Tables 1 and 2 are self-explanatory with the exception of rural income measures. The value of agricultural output, YAG, is an aggregate of the quantities of cereal output of the farm with prevailing prices serving as the weights. 2/ The "sideline" income of a farm, SY, is the value of production of hogs, vegetables, handicrafts and farm implements, etc. The dismantling of  -3- communal production in rural China, which started in 1979, has resulted in a blossoming of sideline production along with major increases in grain output. The Chinese "yuan" at the time of the survey was worth about $.50; the "mu" is a land area measure equal to 1/15 hectare. II. Urban Employment and Earnings Virtually all working-age urban males are employed either as enterprise (factory) workers or as civil servants. This is not true for females and, as Table 1 indicates, urban females who are not employed have almost no earnings. Table 3 reports a probit regression analysis of the determinants of employment among females aged 17 to 55 who were neither students nor retired. Older women are less likely to be employed, but the number of children or other adults in the household has little effect. The women's education leveL, however, has a striking effect on employment; having 5 years of education rather than none increases the probability of being employed from 38% to 71%, if other variables are assumed to equal their sample means. Table 4 reports earnings functions for employed males and females. The earnings function is standard in including estimates both of experience and the square of experience; the results are surprising in that the effect of experience is slight, relative to what is normally found. For females, experience is estimated to have no effect at all. This, rather than differences in the rate of return to schooling, accounts for the the lower observed average wages of employed females. Since it is possible that our measure of experience overestimates the experience of females (who may have been out of the labor force due to child rearing), we attempt to control for  -4- this possibility by entering the number of children in the household into the regression. The effect of this variable, although larger for women than for men, is small and statistically insignificant; this is consistent with stated current practice in China, which has the women only briefly away from employment after childbirth. The coefficient on years of schooling in earnings functions with the log of earnings as the dependent variable can be interpreted as the rate-of- return to education, if the direct costs of education are ignored (Mincer, 1974). 31 Table 4 thus indicates Mincerian rates-of-return of 4.5% for males and 5.6% for females. The returns for males are lower than are any of the 60 Mincerian returns from 34 countries cited by Psacharopoulos (1985), and only studies from Cyprus and Canada found returns lower than these for the females. It is of interest to note, however, that of the 12 Asian studies that Psacharopoulos cites (from 9 political entities), the ones from Taiwan province in 1972 (6.0%) and from Hong Kong in 1981 (6.1%) are the lowest. A more recent Taiwan province study (Fannicott, 1984), using a different methodology with 1982 data, found as did we, higher returns for females than for males - although the levels and differences were higher (16.1% for females versus 8.4% for males). Overall our results suggest that the wage system for civil servants and urban factory workers fails to reflect the productivity *they acquire either through education or experience. III. The Income of Farm Households Table 5 reports Cobb-DougLas production function estimates of the determinants of agricultural output value. The first of the equations is quite good, and land and labor show the expected sign and are highly  significant. Perhaps the most surprising result is that schooling of the head of the household does not seem to have any influence (equation 2) on agricultural output, and the average education level of household members is only suggestively important (equation 1). This suggests that peasants may still follow "guidelines" regarding planting, fertilizing, cultivating and harvesting the major crops, thus reducing whatever effect education could have on output. To test this latter hypothesis, we estimated determinants of income from sideline activities, which constitutes about 25 percent of total income, as the dependent variable (Table 6). The impact of schooling is remarkable:- a 10% increase in income from sideline activities results from every extra year of schooling of the head of the household, for example. Clearly, as initiative and entrepreneurship have become important to take advantage of the emergence of free markets in rural China, education plays a very significant role. So does labor: an extra adult increases the income from sideline activities on average by about 35%. The negative effect of land (i.e., the total area of land under cultivation) suggests some type of specialization: peasant households with large plots for agricultural production have little time for sideline activities. Table 7 shows determinants of total income, reflecting thereby the quality of decisions allocating household labor by amount and quality between sideline and agricultural activities. Here, again, the schooling of the household head seems to be very important, and the average schooling of the household's adults even more so. The importance of education in dealing with new and changing opportunities for farm households has been documented elsewhere (Jamison and Lau, 1982), and these results from China fit well with  -6- earlier findings. We conclude that the normal, positive effect of education on rural earnings holds in China, as long as market forces rather than government regulations determine economic outcomes. 41 IV. Determinants of Schooling Since in this Chinese sample as well as elsewhere, schooling appears to be a powerful determinant of income, it is worth assessing the factors shaping its distribution. Important inequalities persist: Among 14-year olds in the households in this sample, for example, 100% of urban boys were enrolled in school, but only 68% of rural females were. While it is beside the point of this paper to address determinants of schooling in any detail, Table 8 reports the results of an analysis of determinants of the years of schooling completed by the 374 10 to 17 year olds in our sample. The child's age in this age range is, of course, the strongest determinant of the amount of schooling completed; almost 3/4 of a year more schooling is acquired with every additional year of age. Girls on average have 1/4 of a year more schooling than boys, and urban dwellers have 1.6 years more than rural ones. The average amount of schooling of adults in the household is also important; an increase of one standard deviation (3.2 years) in this variable increases the expected amount of the child's schooling by about 1/3 of a year.5/ None of this is surprising; but it does indicate continued inequity among children in the distribution of a resource key in determining future income.  -7- NOTES: / See van der Gaag, J. et al. (1986) for the analyses of the health care utilization data. 2/ For every crop planted, we asked the total yield (in jin), how much was sold to the government, and for which price, and how much was sold at the market, for which price. The quantity of agricultural products kept for household consumption was also asked, and later evaluated at the market price. 3/ We also estimated earnings functions using indicator variables to represent different levels of education; this resulted in a suggestion of some non-linearity in returns, with relatively smaller returns to low levels of education. 4/ Recent unpublished study from China (Hao, 1986) refers to results quite consistent with our based on a cross-tabulation (using data from 36,000 Chinese peasant households) of the highest education level of a household member with not income per household member. Compared to illiterate households, those having a primary graduate as the most educated had 35% higher incomes; the comparable figure for lower secondary graduates was 64%, and, for upper secondary graduates, 95%. When we ran regressions including income as an explanatory variabLe (separately for the rural and urban samples), its effects were small and statistically insignificant.  -8- REFERENCES Cannicott, K., 1984. "Male and Female Earnings in a Developing Economy: The Case of Taiwan". Department of Economics, University of New South Wales, Mimeo. 8ao, Keming, 1986. "Tasks and Policies for Educational Development". Mimeo. Jamison, D. T., and Lawrence J. Lau, 1982. Farmer Education and Farm Efficiency. Baltimore and London: Johns Hopkins University, Press for the World Bank. Mincer, J., 1974. Schooling, Experience and Earnings. New York, Columbia University Press for the National Bureau of Economic Research. Psacharopoulos, C., 1985. "Returns to Education: A Further International Update and Implications." Journal of Human Resources XX: 583-604 van der Caag, J., M. Young, J. So, 1986. Health and Medical Care in Gansu Providence: Evidence from a Household Survey. The World Bank, Mimeo.  -9- Table 1: Variable Definitions, Means and Standard Deviations - Urban Sample (Non-Retired Adults) Variable Definition Mean Standard Deviation Employed Unemployed Employed Unemployed Males Females Females Males Females Females y Individual income, yuan per month 75.50 63.47 4.39 25.40 21.54 15.34 LNY Natural logarithm of Y 4.25 4.02 .36 .48 .73 1.12 SCHOOL Amount of schooling received, in years 8.7 8.3 3.0 3.5 3.3 3.5 AGE Age in years - 39.0 33.9 45.3 12.6 10.4 8.3 EXP Work experience in years, defined to equal AGE minus SCHOOL minus 6 24.3 19.7 36.3 13.5 11.3 10.5 EXPSQ EXP squared 772 511 1426 7.4 536.5 648.4 NCHILD Number of children in the household of the respondent 2.2 1.7 2.2 1.3 1.3 1.4 Sample Size 145 110 41 - - -  - 10 - Table 2: Variable Definitions, Means and Standard Deviations - Rural Sample Standard Variable Definition Mean Deviation LAND Amount of land available to household, in mu 13.20 7.69 LNLAND Natural logarithm of LAND 2.38 .71 LABOR Number of adults in household 3.33 1.44 LN LABOR Natural Logarithm of LABOR 1.10 .46 SCHAVE Average number of years of schooling of adults in household 2.57 2.01 SCHHH Number of years of schooling of household head 2.97 3.18 YAG Total value of agricultural output of household in 1984, yuan 659.44 518.78 LNYAG Natural Logarithm of YAG 6.19 .87 YS Value of sideline output of households in 1984, yuan 219.10 337.85 LNYS Natural logarithm of YS 3.76 2.57 Y YAG + YS 878.54 663.78 LNY Natural Logarithm of Y 6.57 .67  - 11 - Table 3: Probit Regression for Employment Status: Females in Urban Areas (T-Values in Parenthesis) Number of Number of Adults Children Constant AGE SCHOOL (LABOR) (NCHILD) 1.534 -.046 .169 -.076 .053 (1.77) (2.90) (3.88) (.68) (.61)  - 12 - Table 4: Determinants of Urban Earnings* Determinants of Logarithm of Earnings of Determining Variable Males Females SCHOOL .045 .056 (3.7) (2.4) EXP .019 .004 (1.7) (.178) EXPSQ nil nil NCHILD -.03 -.07 (-.92) (-1.3) Intercept 3.45 3.45 N 145 110 2 R .27 .08 R .25 .05 * Table shows regression coefficients; t-vaLues are in parenthesis.  - 13 - Table 5: Determinants of Agricultural Output Value a' Independent Variable Dependent Variable: LNYAG (1) (2) LNLAND .88 .87 (16.1) (16.1) LNLABOR .19 .20 (2.3) (2.4) SCHAVE .02 (1.3) SCHHH .004 (.38) INTERCEPT 3.8 3.9 -2 R .60 .60 N 284 284 a/ Table shows regression coefficiencs; t-values are in parenthesis.  - 14 - Table 6: Determinants of Sideline Output Value a' Independent Variable Dependent Variable: LNYS (1) (2) LNLAND -1.10 -1.13 (8.3) (-4.66) LNLABOR 1.11 1.205 (2.98) (3.22) SCHAVE .22 (3.0) SCHHH .115 (2.51) INTERCEPT 4.57 4.77 - 2 R 0.09 .08 N 284 284 - a! Table shows regression coefficients; t-values are in parenthesis.  - 15 - Table 7: Determinants of Total Output Value a' Independent Variable Dependent Variable: LNY (1) ~~(2) LNLAND .59 .58 (13.3) (12.6) LNLABOR .20 .23 (3.0) (3.3) SCHAVE .079 (6.0) SCHHH .028 (3.2) INTERCEPT 4.8 4.9 R2 .56 .52 .56 .52 N 284 284 a! Table shows regression coefficients; t-values are in parenthesis.  - 16 - Table 8: Years of School Attended, 10 - 17 Years Old a' Standard Regression Variable Definition Mean Deviation Coefficient t-value SCHOOL Years of school attended 5.1 2.3 dependent variable AGE Age of child in years 13.1 2.0 .73 23.2 SEX Indicator variable taking the value "1" for male, "0"?. for female .49 .50 -.23 -1.9 URBAN Indicator variable taking the value "1" for urban children, "0" for ruraL ones .38 .48 1.6 8.9 PSCH Average number of years of schooling of adults in household 4.6 3.2 .08 3.0 INTERCEPT - -5.4 - The number of observations was 374; the r-squared and adjusted r-squared were .74.