Educational expenditure in urban China: income effects, family characteristics and the demand for domestic and overseas education

Abstract

Analysing survey data from 32 selected cities across China in 2003, this article examines parents’ expenditure on their children's education from two aspects: factors affecting domestic education expenditure and factors affecting expenditure on overseas education. The main findings that emerge from this study are as follows. First, household income has significant effects on the magnitude of the domestic and overseas educational expenditures. Second, households where mothers have senior secondary school or college education, and fathers are working in professional occupations are likely to spend more on education for their children. Third, being in the highest income category, having a college-educated father, having a mother who is a cadre or middle professional and living in a coastal area significantly enhances the probabilities for the households sending their children overseas for education.

Notes

¹ In 2000, the gross school enrolment rate at primary, secondary and tertiary levels in China was 114%, 68% and 13%, respectively. These figures compare to the corresponding world average rates: 103%, 70% and 24% (WDI online, World Bank, 2009).

² The 27 provincial capital cities sampled were Shijiazhuang, Taiyuan, Huhehaote, Shenyang, Changchun, Harbin, Nanjing, Hangzhou, Hefei, Fuzhou, Nanchang, Jinan, Zhengzhou, Wuhan, Changsha, Guangzhou, Nanning, Haikou, Chengdu, Guiyang, Kunming, Lasa, Xi’an, Lanzhou, Xining, Yinchuan and Urumqi.

³ Of the 10 793 distributed surveys, some respondents did not answer all questions or did not fill in the questionnaire properly for some questions, so these observations were removed prior to analysis. Thus, depending on the specific empirical model employed, corresponding to a specific question of interest, we had a different number of observations. Instead of only using the data for which we had valid data for all models, we used the maximum number of respondents with valid data for each model.

⁴ The reason that we did not raise the age for the analysis of domestic education expenditure is that there are more financial resources for students if they choose postgraduate study at home. For example, postgraduate students can study part-time while working full-time, which is generally not possible if they study abroad, due to visa restrictions on foreign students. Alternatively employees can get financial assistance from the organizations where they work to allow them to study for a postgraduate degree in China. Studying abroad, though, mainly depends on getting financial assistance from one's family.

⁵ The income variable in this study is categorical. On a scale of 20 income categories ranging from the lowest income group to the highest income group, the median income category is the ninth, skewed to the lower end. The mean monthly household income of the sample is not necessarily equal to the median value and more than likely, it will be smaller than the median value.

⁶ Exceptions are as follows: (a) Ethnic minorities can have more than one child irrespective of whether they live in urban or rural areas. (b) Rural residents can have two children. If rural residents migrate to the cities and acquire an urban household registration they will be counted as urban residents with more than one child. (c) The one-child policy does not apply to children born to Chinese nationals abroad who come to China with their parents when they return to China.

⁷ This variable was also initially included in the sample used to analyse the determinants of domestic education expenditure. The results show that the coefficient on the variable measuring confidence in the pension insurance system was insignificant in all models and that the t-values were very close to zero. Hence, this variable was omitted from the reported models estimating domestic education expenditure.

⁸ Four of 35 cities are not included in this survey; namely, Dalian, Ningbo, Shenzhen and Qingdao. Lhasa is also not listed among the 35 cities, but it is included in the survey employed here.

⁹ The structural equation for the Tobit model is (Long, 1997, p. 196): where is the latent variable, and x _i is a vector of household characteristics, β is the vector of parameters to be estimated and ϵ is the normally and independently distributed error term. The observed y _i is defined as

¹⁰ To assess potential bias due to multicollinearity, we examine the pairwise correlation coefficients between each pair of variables. All the correlations between each pair of variables are smaller than 0.5 for both domestic and overseas sample. The problem of multicollinearity is not well-defined (Wooldridge, 2006). Gujarati (1995, pp. 335–6) suggests that multicollinearity is of concern if the simple correlation is higher than 0.6 and a serious problem if the simple correlation is higher than 0.8.

¹¹ http://www.stata.com/faqs/stat/pseudor2.html (accessed 14 November 2008).