Discussion Paper

No. 2019-7 | February 05, 2019
The only child, birth order and educational outcomes


The one-child policy was implemented in September 1980 and abolished in late 2015. With this change in the demographic policy, the fertility decision of families also changed. Such decisions can result in an increase in the number of siblings in a family. Individuals’ educational outcomes may be affected by a change in their parents’ fertility decision. The objective of this paper is to provide evidence of the difference of educational outcomes between the only child and the first born. The authors try to estimate the change of educational outcomes when the only child of a family turns to the first born of a family. Moreover, they estimate different channels to interpret these effects. They employ the dataset of China Education Panel data in this paper. In the part of mechanism check, the Sobel-Good test is used for checking the mediation effects of different channels. They found the only child has significant higher educational outcomes comparing to a child who has siblings. Furthermore, the middle child has the lowest educational outcomes of a family. The last born has higher educational outcomes than his or her siblings. To explain these effects, the authors use three channels to interpret: (1) money resource, (2) parenting time, and (3) closeness of parent-child relationships. The policy implication is to help the policymaker estimate and predict the impact of the new demographic policy.

JEL Classification:

I20, J13


  • Downloads: 259


Cite As

Yehui Lao and Zhiqiang Dong (2019). The only child, birth order and educational outcomes. Economics Discussion Papers, No 2019-7, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2019-7

Comments and Questions

Anonymous - Referee report 1
March 18, 2019 - 13:50

This paper uses a school based survey and reports an interesting phenomenon: when a child has no siblings, his/her academic performance is significantly better, and this is channeled through more parental investment in the only child.

The authors then try to instrument the only child variable with a host ...[more]

... of parental characteristics: birth age, ethnicity and political party membership of each parent. I do not like this set of instrument – they can have independent effects on children’s school performance other than affecting the number of children. And I do not see a possibility of finding any plausible instrument in this case. Traditional instruments such as sex composition of the first two children do not apply because they are looking at whether the family has one or more children; family planning policy variations do not apply either because they rely on a school fixed effect estimation; the only possible choice of instrument is the occurrence of twins at first birth but this may be too rare.

I think the study can still be interesting without identifying the causal effect of the only-child. The proportion of such children is 46% in the sample, which is very significant. How the only-children fare in school and beyond is a big issue for China. Some hypothesized that these children are treated as “little emperors” by parents and grandparents and are too self-centered; others hypothesized that these children are under too much pressure that their mental health suffers. The current paper shows the positive side of this phenomenon – that their school achievements are better. Instead of going the direction of causal identification, the paper can be enriched by looking at more dimensions of child development – Does the survey contain any psycho-social indicators?

Along with the story about only-children, the second focus of the paper is sibling order effect. They conclude that the middle children are the worse achievers. This is far less interesting than the only-children analysis and the research is not well implemented. If you look at the data summary table (page 7), only 3.2% of all children belong to his category (sibr2), and the next category (sibr3) is 22.7%. There must be some mistakes in handling the data - How can there be so many third born children when the second born is so few? Even if data processing is corrected, I think this data set is not suitable for studying the sibling effect - You would need family fixed effect models. Since children from same families can not be identified, this is impossible.

There are some other data processing problems. Many categorical variables (ethnicity, party membership. Hukou place and others) that take values from 1 to N are treated as continuous variables in regressions – they should be converted to dummy variables with values of 0 and 1.

Minor problems:
1. You need better data description. In what provinces are these schools chosen? Within a school, how are classes and students chosen? Are these places mainly urban or rural, poor or rich areas?
2. Variable definitions are sometimes unclear: sibr1-3 are not defined.
3. Sample choice: why do you drop 9th grade students? You are not doing student individual fixed effect so there is no need to drop them.
4. Why don’t you use class fixed effects?
5. I suggest that you report control variables in the regressions.
6. Use better variable naming in the regressions so that they are self explanatory.
7. Literature section is too long.

Yehui Lao - Author´s response to the reviewer 1
March 20, 2019 - 16:40

We thank the reviewer for the thoughtful review and comments. Please see the attached file.

Yehui Lao - Revised version
March 31, 2019 - 19:44

See attached file