| This paper starts with the divergence of the shape of saving rate-age curve in the literature.We explain the root of the disagreement.We investigate the effect of household structure on saving rate-age curve.In the exploration,we find and demonstrate that the tuition reform can explain the positive u-shape of saving rate.In addition,this paper studies the impact of tuition fees reform of higher education on intergenerational mobility.According to the life cycle hypothesis proposed by Modigliani and Brumberg(1954),the saving rate–age curve of consumers presents an inverted u-shape.This classic prediction of the life cycle hypothesis has been validated in many European and American literatures.Zhou et al.(2009)analyze the data of China urban household survey(UHS)of six provinces from 1988 to 2003.They adopt the Age-Period-Cohort decomposition method in Deaton and Paxson(1994).They find that household consumption,household income and household saving rate all increase with the age of head of household.Chamon and Prasad(2010)analyze data from the China urban household survey of 10 provinces from 1990 to 2005,using Age-Period-Cohort decomposition method.However,Chamon and Prasad(2010)find that the consumption-age curve and the income-age curve are inverted u-shaped.In especial,they find that saving rate-age curve shows a positive U shape,which characterized by a low saving rate among middle-aged households.This raises two questions.Why do the two papers adopt similar methods and data,but get completely different results? Why are the household saving rate-age curves obtained in both papers not the inverted u-shape predicted by the life cycle hypothesis?Besides,Song and Yang(2010),Chamon et al.(2013),Fan and Zhu(2013),Rosenzweig and Zhang(2014),Li and Wu(2014)and the third chapter find that household saving rate-head age curve has a positive U shape in the data of UHS,CHIPS and CFPS.Its salient feature is that middle-aged households have the lowest saving rate.The second chapter explains the first question above.We find that the difference on collinearity identification method of APC decomposition is the root of the saving rate-age pattern divergence.APC decomposition is the breaking down of variables we care about(such as consumption,income,and savings)into three parts:age effects,cohort effects,period or year effects.The three parts are related to the age of the respondents,the year of birth and the year of the survey.When using data for APC decomposition,It is common practice to set up a group of dummy variables for each of the three categorical variables,a(age),c(cohort),and p/t(year).After removing one dummy variable from each group,put them into the regression equation.As age = year – cohort,there is collinearity between these three sets of dummy variables.Without constraints,OLS estimators cannot be obtained.Yang et al.(2008)proposed an IE estimation method based on principal component regression to solve this problem.Economic literature mostly follow Deaton and Paxson(1994)and Deaton(1997),and assume that the year effect adds up to 0 and that the year effect has no trend,which is the hypothesis of Zhou et al.(2009).Chamon and Prasad(2010)point out that the household income,consumption and saving rate in the data of China all showed a significant upward trend.Income,consumption and savings rates have risen rapidly over time in almost all groups.This should be thought of as year effect rather than age effect.Therefore,the assumption that the year effect has no trend is not consistent with the reality.Instead,they assume the cohort effects have no trend.Heathcote et al.(2010)believed that if the purpose of researchers is mainly to obtain age effect,instead of putting age,year,and cohort into the regression equation,it is better to discard the relatively insignificant group of year and cohort.In this chapter,the above four methods are applied to the analysis of UHS data.From the results of these four methods,a pattern emerges.If the time trend is not controlled in APC decomposition,the age effect obtained must be a straight line with an upward slope,such as assuming the year effect has no trend or discarding year effect.If the time trend is controlled in the APC decomposition,the age effect of household consumption and household income is inverted u-shaped,the age effect of the household saving rate is the lowest among middle-aged households,such as IE method,assuming the cohort effect have no trend,discarding cohort effect.This explains the divergence between Zhou et al.(2009)and Chamon and Prasad(2010).The APC decomposition of Zhou et al.(2009)do not control the time trend,but constrain the year effect to periodic fluctuation.Chamon and Prasad(2010)directly control the time trend.In the second chapter,the importance of collinearity identification method for APC decomposition has been explained.Although different collinearity identification methods can get different life cycle patterns of saving rate,the identification method alone cannot explain the puzzle of positive u-shape of saving rate-age curve.In the third chapter,APC decomposition is applied to investigate the impact of household structure on savings rate.The analysis based on UHS data shows that,the co-residence of the elder and adult offspring with the middle age house head emphasized by Rosenzweig and Zhang(2014),Li and Wu(2014)may not be the key to explain the abnormal phenomenon of the low saving rate of middle age households.The number of offspring aged 22 and under may be even more important.After controlling for the number of children aged 22 and under,the resulting household saving rate-age curve is close to an inverted U.The number of adult offspring and elders do not have the same effect.In this chapter,the same collinearity identification method and the same regression equation are used,but in the regressions of UHS data from 2002 to 2009,the number of offspring aged 22 and under are more powerful.Analysis based on CFPS data shows that the phenomenon of the lowest saving rate among middle-aged households is also confirmed in CFPS data from 2010 to 2016.It shows that this anomaly is not a short-term phenomenon.Controlling for the elder and adult offspring do not make the anomaly disappear,consistent with the findings of the UHS data.Controlling the number of college and high school students can make the anomaly of the lowest saving rate for middle-aged households mostly disappear.The number of college students,high school students,middle school students and babies all have a significant negative impact on the savings rate.Among them,the absolute value of the regression coefficient of the number of college students and high school students are the largest and have the strongest explanatory power.Inspired by the results of chapter three above,this paper carefully examines the household savings rate-head age curve in the UHS data from 1986 to 2009.This chapter finds that 1996 seems to be the cut-off point.Before that,the household savings-head age curve are roughly an inverted U shape,while after that,the urban household savings-head age curve begin to show a positive U shape with a lowest household saving rate for middle age(around 40-45 years old),and more and more obvious.This fact has not been highlighted in the literature.This could be some sort of policy shock that has shifted the saving rate-age curve.Combined with the historical background,this chapter holds that the increase of educational burden caused by the reform of college tuitions between 1994 and 1997 may be an important reason for the change of saving rate-age curve.Before 1994,most of China's ordinary institutions of higher education and technical secondary school were funded by the government.Most students did not need to pay tuition fees,the government also gave some living allowance to students.In addition,this chapter find that in the UHS data,the education expenditure-age curve is inverted u-shaped,and the curve moves up over time.This means that education spending has an increasing impact on household saving rate,suggesting that policy shocks to education may explain the positive u-shape of saving rate.The fourth chapter demonstrates that the tuition reform of college and technical secondary school around 1997 can explain the positive u-shape puzzle of saving rate.In theory,household consumption is smooth,the savings rate-age curve may be inverted U-shaped,either at public expense or without borrowing constraints.At one's own expense and with lending constraints,before high school and college,families may save for education.In high school and college,families face high education expenditure,and low income of child,household savings rates are likely to be low.After graduation,household income rises and education expenditure falls,and the household savings rate may rebound.This may result in a positive U-shape of the savings-age curve.Based on the data of urban household survey,taking the tuition reform as a quasi-natural experiment,we employ a difference in difference(DID)estimation method.We find that the saving rates of households with children aged 15-23 years old fall significantly in 1997 and later,relative to other households.The dynamic analysis shows that the negative effect of children aged 18-23 on the household savings rate increased rapidly from 1997 to 2002 and gradually decreased from 2002 to 2009.This is to some extent consistent with the trend of the average tuition burden of college students.The dynamic analysis also found that the cross term coefficients before 1997 were not significant,which also verified the common trend hypothesis needed for DID estimation.In the regression analysis,the year effect of the household savings rate obtained was almost in a straight line from 1997 onwards,which meant that most households experienced a long-term upward trend of the savings rate from 1997 to 2009,which is the embodiment of the mystery of China's constantly rising savings rate.In the regression analysis,the household saving rate-head age curve obtained generally presents an inverted U-shape,which is close to the prediction result of life cycle theory.According to the data,middle-aged families have the largest number of children aged 15-23,and the negative impact on the household savings rate is greater,which may lead to a relatively low savings rate.Mechanism analysis shows that households with children aged 15-23 years old have relatively higher consumption due to education expenditure,and relatively lower income due to the low employment of children,which leads to low saving rates.Studies on placebo test,sample adjustment,co-residence bias,exclusion of other policies in the same period,show that the results of this chapter are relatively robust.Heterogeneity analysis show that low-income households are subjected to greater negative impact.In the fifth chapter,we study the impact of a rise of higher education tuition on the intergenerational mobility.Existing literature focuses on analyzing the impact of college expansion since 1999 on intergenerational mobility.Most of they found that the college expansion did not improve but reduced the intergenerational mobility and the fairness of educational opportunities.Some of these literatures compare the intergenerational mobility before and after enrollment expansion to study the impact of enrollment expansion.Given that the expansion of college enrollment is accompanied by a rapid increase in tuition fees,this paper suggests that such estimates may reflect the mixed effect of education expansion and tuition increase.If the expansion were to take place under a publicly funded system,the children of low-income families might not be deterred by high tuition fees from giving up or reducing higher education,the expansion is likely to lead to an improvement in intergenerational mobility.This chapter develops a theoretical model on the parental decision of children's education investment to illustrate the impact from the perspective of budget constraint.In the model of this paper,the parents with high education background have higher income and can pay for higher education tuition without borrowing money.The borrowing constraint is not binding,and tuition payment has little impact on their consumption.The parents with low education background have low income and cannot afford the higher education tuition by themselves.In the face of borrowing constraints,paying tuition has a great impact on their own consumption,and their children may give up higher education.When tuition fees rise,more parents will have binding borrowing constraints,resulting in a smaller intergenerational mobility of education.In the empirical study,this chapter takes the college tuition and subsidies reform in 1986 as a quasi-natural experiment,to identify the policy effect of college tuition reform on the intergenerational mobility of education,using the 2000 census data and the China Family Panel Studies(CFPS)data.Through baseline regression and multiple grouping regressions,this chapter finds that the increase of the education burden induced from the reform of college tuition has increased intergenerational relevance of education,and reduced the intergenerational mobility of education,and this effect is more pronounced in areas with higher increase in the tuition.The analysis of urban and rural heterogeneity shows that the decline of intergenerational mobility is more significant in the high tuition burden urban areas,which is consistent with the incremental nature of tuition fee reform.Our results are robust with consideration of co-residence bias,government investment in basic education and enrollment rate of higher education. |
PDF Full Text Request